{"id": "2890c16b240a2a42e5dbe3574416937ba6a7186999f616397837f13766c96021", "sources": ["arxiv"], "title": "Universal Image Restoration via Internalized Chain-of-Thought Reasoning", "abstract": "Image restoration seeks to recover high-quality images from degraded inputs but becomes highly ill-posed under complex, mixed degradations. While unified all-in-one models are common, their performance declines as degradation complexity increases. Recent works adopt Chain-of-Thought (CoT) reasoning for multi-round restoration using specialized modules. However, this approach faces two key limitations: (i) increased computational cost due to multi-step processing, and (ii) weak modeling of interactions between degradations during stepwise inference. We introduce CoTIR, a universal image restoration framework that internalizes CoT reasoning within a single model. Concretely, we view image restoration as a specialized subtask of image editing, which implies that a large-scale pre-trained editing model provides a more favorable optimization starting point. Building on this, we fine-tune the model for restoration and further encode structured CoT-style reasoning into the learning objective via a differentiable formulation inspired by Lagrangian optimization, enabling holistic restoration without chaining specialized restorers. To facilitate training and evaluation, we further present CoTIR-Bench, a large-scale benchmark comprising 5.2 million samples with CoT-style reasoning traces. Extensive experiments on CoTIR-Bench and broad real composite degradation scenes show that CoTIR achieves stronger perceptual quality and more competitive fidelity than both all-in-one models and multi-round restoration methods. The source code is available at https://github.com/gy65896/CoTIR.", "authors": ["Yu Guo", "Zhengru Fang", "Shengfeng He", "Senkang Hu", "Yihang Tao", "Phone Lin", "Yuguang Fang"], "categories": ["cs.CV"], "fields_of_study": [], "published_date": "2026-06-16", "url": "https://arxiv.org/abs/2606.17557", "pdf_url": "https://arxiv.org/pdf/2606.17557v1", "arxiv_id": "2606.17557", "doi": null, "citation_count": 0, "influential_citation_count": 0, "has_code": true, "code_url": "https://github.com/gy65896/CoTIR", "venue": null, "quality_score": 0.65}
{"id": "aa10d74c9c9366ec63ba58301500ee40162177e78d4d18b48b8db82a0e32da4e", "sources": ["arxiv", "semantic_scholar"], "title": "Latent Thought Flow: Efficient Latent Reasoning in Large Language Models", "abstract": "Large Language Models (LLMs) increasingly rely on intermediate reasoning, yet explicit Chain-of-Thought (CoT) suffers from a linguistic space bottleneck: each thought must be decoded into tokens, causing high inference overhead. Latent reasoning moves deliberation into continuous space, but existing methods mostly learn deterministic or reward-maximizing paths, lacking a principled way to allocate probability across trajectories with different correctness and costs. We propose Latent Thought Flow (LTF), which models reasoning as variable-length continuous trajectories and trains a sampler to match a reward-induced posterior over answer quality and computation cost. We instantiate this with a continuous GFlowNet using stochastic latent transitions. To handle sparse answer supervision, we introduce an Entropy-Weighted Subtrajectory Balance objective for intermediate rewards and a reference-prior regularizer to anchor exploration. Experiments under finetuning and transfer learning settings show that LTF outperforms explicit CoT and latent reasoning baselines, improving accuracy by 9.5% while reducing reasoning length by 27.2% on average compared with strong latent reasoning baselines.", "authors": ["Xiandong Zou", "Jing Huang", "Jianshu Li", "Pan Zhou"], "categories": ["cs.AI", "cs.LG"], "fields_of_study": ["Computer Science"], "published_date": "2026-06-15", "url": "https://arxiv.org/abs/2606.16222", "pdf_url": "https://arxiv.org/pdf/2606.16222v1", "arxiv_id": "2606.16222", "doi": null, "citation_count": 0, "influential_citation_count": 0, "has_code": false, "code_url": null, "venue": null, "quality_score": 0.35}
{"id": "b8bede8aea3a4641efd49163e98de775abc49de3360427548b5df5eac099c686", "sources": ["arxiv", "semantic_scholar"], "title": "RoboPIN: Grounded Embodied Reasoning via Pinned Chain-of-Thought", "abstract": "Embodied reasoning requires models to perceive task-relevant objects and spaces in physical environments and maintain consistent visual grounding throughout multi-step reasoning. However, current vision-language models rely on text-only or coordinate-augmented chain-of-thought, where entity references remain implicit and ambiguous. This may cause the reasoning process to decouple from visual evidence, entity references to drift across steps, and a causal disconnection between the reasoning trajectory and the final answer, with these problems further amplified in multi-view scenarios due to cross-view appearance changes. To address these issues, we propose Pinned Chain-of-Thought (\\pincot{}), a structured reasoning paradigm that pins every reasoning step to visual evidence. \\pincot{} introduces the concept of \\reasoninganchor{}, which binds each task-relevant entity to a structured visual anchor with entity name, unique identity, view index, and spatial grounding, enabling consistent entity tracking across reasoning steps and views. We build a fully automated data generation pipeline to construct \\dataset{}, a high-quality \\pincot{}-formatted reasoning dataset. We then train \\method{} through three-stage post-training that progressively injects embodied knowledge, structured reasoning ability, and process-supervised alignment, with rewards that directly constrain both anchor localization and identity consistency during reasoning. On 14 benchmarks covering embodied spatial reasoning, multi-view reasoning, and pointing, \\method{} with only 4B parameters consistently outperforms 7B level open-source embodied models, achieving a 12\\% average improvement over the strongest 7B baseline, Mimo-Embodied. Further analysis shows that \\pincot{} improves grounding accuracy and cross-step identity consistency, validating the effectiveness of process supervision.", "authors": ["Yaoting Huang", "Yifu Yuan", "Linqi Han", "Chengwen Li", "Shuoheng Zhang", "Xianze Yao", "Hongyao Tang", "Yan Zheng", "Jianye Hao"], "categories": ["cs.AI"], "fields_of_study": ["Computer Science"], "published_date": "2026-06-14", "url": "https://arxiv.org/abs/2606.15753", "pdf_url": "https://arxiv.org/pdf/2606.15753v1", "arxiv_id": "2606.15753", "doi": null, "citation_count": 0, "influential_citation_count": 0, "has_code": true, "code_url": null, "venue": null, "quality_score": 0.65}
{"id": "78e6d85cf4540c86bcccc2db06d052f84252c174e28c72fd7549d3d190dd9830", "sources": ["arxiv", "semantic_scholar"], "title": "Beyond the Commitment Boundary: Probing Epiphenomenal Chain-of-Thought in Large Reasoning Models", "abstract": "Chain-of-thought (CoT) reasoning is the dominant paradigm for inference-time scaling in language models, yet the causal influence of individual steps on the final answer poorly understood. We estimate each step's causal importance via early exit and use this measure to study how answers form across the reasoning traces of several model families. Across diverse tasks, we find that reasoning typically crosses a \\emph{commitment boundary} -- a sharp transition from transient intermediate guesses to a stable, high-confidence answer. This transition often happens in a single step, well before the model's reasoning block ends, and is followed by \\emph{epiphenomenal} CoT steps that leave the final answer probability unaltered. Using attention probes, we show that answer-formation stages can be linearly decoded from intermediate reasoning steps with high accuracy and generalize robustly to unseen reasoning tasks. We exploit this signal to early-exit reasoning blocks at the commitment boundary, reducing the length of CoTs up to 55\\% on average with negligible impact on model performance.", "authors": ["Daniel Scalena", "Sara Candussio", "Luca Bortolussi", "Elisabetta Fersini", "Malvina Nissim", "Gabriele Sarti"], "categories": ["cs.LG", "cs.AI", "cs.CL"], "fields_of_study": ["Computer Science"], "published_date": "2026-06-11", "url": "https://arxiv.org/abs/2606.13603", "pdf_url": "https://arxiv.org/pdf/2606.13603v1", "arxiv_id": "2606.13603", "doi": null, "citation_count": 0, "influential_citation_count": 0, "has_code": false, "code_url": null, "venue": null, "quality_score": 0.35}
{"id": "4a1b4dea4dbec85e251a1e996f085f8fd2de1cebee214d614fb5b2825b278221", "sources": ["arxiv", "semantic_scholar"], "title": "AVIS: Adaptive Test-Time Scaling for Vision-Language Models", "abstract": "Modern Vision-Language Models (VLMs) benefit from chain-of-thought prompting and test-time scaling, but these gains often come with prohibitive inference cost due to large visual contexts and long decoding chains. We view this cost through two coupled axes: Visual Context Scaling (VCS), which controls how much visual evidence is passed to the language model, and Visual Reasoning Scaling (VRS), which controls how much inference-time reasoning search is performed. Existing methods typically optimize one axis at a time, leaving the joint allocation of compute across these axes underexplored. We introduce Adaptive Visual Inference Scaling (AVIS), a lightweight policy that adapts both VCS and VRS per query. AVIS realizes VCS through Key Diversity Visual (KDV) pruning, a training-free $O(N)$ key-based rule for removing redundant visual tokens before prefilling, and realizes VRS through adaptive self-consistency, using a learned difficulty predictor to select the number of reasoning rollouts. AVIS is deployment-friendly and compatible with shared-prefill inference, where all rollouts reuse a single prefilling pass and KV cache. Across diverse image and video reasoning benchmarks, AVIS improves the accuracy--compute trade-off relative to VCS-only and VRS-only baselines, and remains effective on top of RL post-trained VLMs while keeping compute and latency low.", "authors": ["Ahmadreza Jeddi", "Minh Ngoc Le", "Amirhossein Kazerouni", "Hakki Can Karaimer", "Hue Nguyen", "Iqbal Mohomed", "Michael Brudno", "Alex Levinshtein", "Konstantinos G. Derpanis", "Babak Taati", "Radek Grzeszczuk"], "categories": ["cs.CV", "cs.AI"], "fields_of_study": ["Computer Science"], "published_date": "2026-06-10", "url": "https://arxiv.org/abs/2606.11576", "pdf_url": "https://arxiv.org/pdf/2606.11576v1", "arxiv_id": "2606.11576", "doi": null, "citation_count": 0, "influential_citation_count": 0, "has_code": false, "code_url": null, "venue": null, "quality_score": 0.35}
{"id": "aa3334c4d32fd7c1cc7ccc544cfa0e5a1a255207a16c4eb8e1ca08c33928a946", "sources": ["arxiv", "semantic_scholar"], "title": "TVI-CoT: Text-Visual Interleaved Chain-of-Thought Reasoning for Multimodal Understanding", "abstract": "Chain-of-thought (CoT) reasoning has proven effective for enhancing problem-solving in large language models. However, when applied to multimodal LLMs (MLLMs), existing CoT approaches suffer from a fundamental limitation: they perform reasoning entirely in text without accessing visual features during the reasoning process. After initial visual encoding, image information becomes inaccessible, forcing models to reason based solely on whatever was captured in the initial description, which forms a `vision-blind reasoning' paradigm that limits fine-grained visual extraction, error verification, and adaptive attention. We propose Text-Visual Interleaved Chain-of-Thought (TVI-CoT), a framework that enables explicit interleaving of textual reasoning and visual feature access through learnable control tokens <THINK>, <LOOK> and <ANSWER>. These tokens allow dynamic switching between reasoning and visual grounding, attending to relevant image regions conditioned on the evolving reasoning state. Experiments on eight benchmarks demonstrate state-of-the-art results among MLLM-based CoT methods and notable performance boost compared to the baseline: +6.1% on MMMU, +3.8% on MathVerse, +3.4% on MathVista, and +3.4% on ScienceQA. Code is available at https://github.com/hulianyuyy/TVI-CoT.", "authors": ["Lianyu Hu", "Xiaoyu Ma", "Zeqin Liao", "Yang Liu"], "categories": ["cs.CV"], "fields_of_study": ["Computer Science"], "published_date": "2026-06-07", "url": "https://arxiv.org/abs/2606.08464", "pdf_url": "https://arxiv.org/pdf/2606.08464v1", "arxiv_id": "2606.08464", "doi": null, "citation_count": 0, "influential_citation_count": 0, "has_code": true, "code_url": "https://github.com/hulianyuyy/TVI-CoT", "venue": null, "quality_score": 0.65}
{"id": "ed6d36c1b4ca3432beff5959ef1c752cc688989dc5485d27e74b69585769a9a5", "sources": ["arxiv", "semantic_scholar"], "title": "ThinkBooster: A Unified Framework for Seamless Test-Time Scaling of LLM Reasoning", "abstract": "Test-time compute (TTC) scaling has emerged as a powerful paradigm for improving large language model (LLM) reasoning by allocating additional compute during inference, e.g., via multi-sample generation and verifier-based reranking. Existing TTC scaling strategies and reasoning scorers remain fragmented, evaluated under inconsistent protocols, and are rarely analyzed through the lens of quality-cost trade-offs. We introduce ThinkBooster, a unified framework for seamless test-time compute scaling of LLM reasoning, which consists of (i) a modular Python library implementing state-of-the-art TTC scaling strategy and scorer families, (ii) a benchmark that jointly evaluates performance and computational efficiency, and (iii) a deployable OpenAI-compatible proxy service that enables drop-in integration of adaptive reasoning into real-world applications. We further provide a demo visual debugger for inspecting the reasoning trajectories, intermediate selection decisions, and alternative reasoning paths. Empirical results on mathematical and coding tasks reveal the performance-compute trade-offs of TTC scaling strategies and scoring methods and demonstrate that ThinkBooster provides practical gains in real-world tasks. The code is available online under an MIT license.", "authors": ["Vladislav Smirnov", "Chieu Nguyen", "Sergey Senichev", "Minh Ngoc Ta", "Ekaterina Fadeeva", "Artem Vazhentsev", "Daria Galimzianova", "Nikolai Rozanov", "Viktor Mazanov", "Jingwei Ni", "Tianyi Wu", "Igor Kiselev", "Mrinmaya Sachan", "Iryna Gurevych", "Preslav Nakov", "Timothy Baldwin", "Artem Shelmanov"], "categories": ["cs.CL", "cs.AI", "cs.LG"], "fields_of_study": ["Computer Science"], "published_date": "2026-06-05", "url": "https://arxiv.org/abs/2606.06915", "pdf_url": "https://arxiv.org/pdf/2606.06915v2", "arxiv_id": "2606.06915", "doi": null, "citation_count": 0, "influential_citation_count": 0, "has_code": true, "code_url": null, "venue": null, "quality_score": 0.65}
{"id": "b29ed21a15ec1c1081dac4ebf79b2a09aba73dafef9e52d43d37eb5f23716726", "sources": ["arxiv", "semantic_scholar"], "title": "VTI-CoT: Visual-Textual Interleaved Chain of Thought for Video Reasoning", "abstract": "Video reasoning aims to understand complex temporal events and causal relationships within videos. Recently, Chain-of-Thought (CoT) has been introduced to this field to enhance reasoning accuracy. However, existing CoT-based video reasoning methods primarily rely on text-only information for logical deduction, overlooking critical visual information during the inference process. Inspired by the human cognitive mechanism of reviewing visual segments during inference, we propose VTI-CoT, a Visual-Textual Interleaved CoT framework. VTI-CoT integrates textual reasoning steps with corresponding visual frames. Given the scarcity of visual-textual interleaved CoT in existing datasets, we develop an automated annotation pipeline to construct high-quality multimodal CoT data. Further, reasoning over long-form videos entails increasingly long CoT token sequences, which severely hinders training convergence and efficiency. To address this, we employ Optical Character Recognition (OCR)-based compression techniques to compress CoT supervision signals into a single canvas. Experimental results demonstrate that VTI-CoT achieves state-of-the-art performance among models of the same parameter scale while significantly improving training efficiency.", "authors": ["Shufan Zhang", "Ziyue Lin", "Bairun Wang", "Lei Jin", "Xuanding Ding", "Xinzhu Ma", "Kunlin Yang"], "categories": ["cs.CV"], "fields_of_study": ["Computer Science"], "published_date": "2026-06-04", "url": "https://arxiv.org/abs/2606.05736", "pdf_url": "https://arxiv.org/pdf/2606.05736v1", "arxiv_id": "2606.05736", "doi": null, "citation_count": 0, "influential_citation_count": 0, "has_code": false, "code_url": null, "venue": null, "quality_score": 0.35}
{"id": "604514446b2f6a04aa58fff8b494679bdbe18cd5dfa9b5dcfc5d034cffd4b1f3", "sources": ["arxiv", "semantic_scholar"], "title": "Closing the Loop on Latent Reasoning via Test-Time Reconstruction", "abstract": "Recent work moves intermediate reasoning from natural-language traces into latent or cache-level representations to reduce token overhead and avoid a discrete communication bottleneck. However, this shift also removes a key advantage of textual reasoning: intermediate states are no longer inspectable, making it difficult to determine whether a latent state still preserves the constraints of the original query. As a result, latent reasoning typically operates in an open loop, where a latent state is produced and consumed without an input-anchored fidelity check. We propose ReLAT (Reconstruction-Guided Latent Reasoning At Test Time), a self-supervised test-time training method that closes this loop using the query itself as the reference. Our key observation is that if a latent state faithfully represents a query, the query should be recoverable from it; if the query cannot be recovered, the latent state has lost task-relevant information. ReLAT operationalizes this principle by constructing a differentiable Question -> Latent Thought -> Question cycle and optimizing query reconstruction loss through the latent thought before answer generation. This anchors opaque latent computation to the problem specification it is supposed to represent. Across mathematical reasoning, knowledge QA, and code generation benchmarks on the Qwen family, ReLAT consistently improves over single-model inference, text-based collaboration, open-loop latent collaboration, and alternative test-time training objectives. On Qwen3-8B, ReLAT raises AIME 2024 accuracy from 56.7% to 73.3%, a 16.6-point gain over the strongest open-loop latent baseline.", "authors": ["Xiaopeng Yuan", "Haibo Jin", "Ye Yu", "Peng Kuang", "Lijun Yu", "Yushun Dong", "Haohan Wang"], "categories": ["cs.AI"], "fields_of_study": ["Computer Science"], "published_date": "2026-06-04", "url": "https://arxiv.org/abs/2606.06252", "pdf_url": "https://arxiv.org/pdf/2606.06252v1", "arxiv_id": "2606.06252", "doi": null, "citation_count": 0, "influential_citation_count": 0, "has_code": false, "code_url": null, "venue": null, "quality_score": 0.35}
{"id": "bffd920975cdc13173cc31903a03cdaf9c70d1ae6cd1e448d6f336c287cbdceb", "sources": ["arxiv", "semantic_scholar"], "title": "Test-Time Compute Scaling for ASR with Depth-Conditioned Looped Transformers", "abstract": "End-to-end ASR systems typically use fixed-depth acoustic encoders at inference, making it difficult to trade additional test-time computation for improved recognition without training a larger model. A natural approach is to reuse a shared Transformer block recurrently, but we find that naive looping does not fully exploit additional recurrent compute. We introduce LARM, a depth-conditioned looped Transformer that turns recurrent encoder depth into a controllable test-time compute axis. LARM combines sparse CTC checkpoints, supervision-clock embeddings, FiLM depth conditioning, and delayed soft-posterior feedback. These components structure the loop into recognition checkpoints separated by latent refinement phases and allow shared weights to specialize across recurrent steps. On LibriSpeech, LARM improves WER as the number of inference loops increases and achieves performance competitive with deeper unshared-parameter baselines. Our results show that test-time compute scaling can extend beyond autoregressive language-model reasoning to continuous non-autoregressive speech recognition.", "authors": ["Yacouba Kaloga", "Shashi Kumar", "Shakeel A. Sheikh", "Driss Khalil", "Petr Motlicek", "Ina Kodrasi"], "categories": ["cs.LG"], "fields_of_study": ["Computer Science"], "published_date": "2026-06-03", "url": "https://arxiv.org/abs/2606.04678", "pdf_url": "https://arxiv.org/pdf/2606.04678v2", "arxiv_id": "2606.04678", "doi": null, "citation_count": 0, "influential_citation_count": 0, "has_code": false, "code_url": null, "venue": null, "quality_score": 0.35}
{"id": "14df81b06c0375a689285c596d4ad587399121fd928d5e1bfd23c95de1541d12", "sources": ["arxiv", "semantic_scholar"], "title": "HybridThinker: Efficient Chain-of-Thought Reasoning via Compressed Memory and Transient Thought Steps", "abstract": "Extended chain-of-thought (CoT) traces improve LLM reasoning but incur substantial computational and memory costs. While existing CoT compression methods mitigate this by condensing thought steps into compact representations via memory tokens and retaining only these representations at inference time, the loss of fine-grained information makes subsequent steps more error-prone. To alleviate this, we propose \\textbf{HybridThinker}, where in addition to preserved these representations, thought steps are also temporarily retained to provide fine-grained details. However, we observe that naively keeping thought steps accessible to subsequent steps \\emph{during training} lets the model bypass memory tokens by retrieving information directly from these steps, leaving the model's ability to compress and retrieve information through memory tokens insufficiently trained. We therefore introduce a hybrid training scheme, in which only some thought steps are directly accessible through attention to subsequent steps, while the other thought steps are masked, forcing the model to use memory tokens for compression and retrieval. Across 4 reasoning benchmarks, HybridThinker matches the uncompressed baseline, advancing the state of the art in CoT compression by 5.8 points on average accuracy with similar inference time. Ablation studies confirm that both temporary thought-step retention and the hybrid training scheme contribute to these gains.", "authors": ["Xin Liu", "Runsong Zhao", "Xinyu Liu", "Junhao Ruan", "Pengcheng Huang", "Shichao Dong", "Chunyang Xiao", "Chenglong Wang", "Changliang Li", "Jingbo Zhu", "Tong Xiao"], "categories": ["cs.CL"], "fields_of_study": ["Computer Science"], "published_date": "2026-06-02", "url": "https://arxiv.org/abs/2606.03768", "pdf_url": "https://arxiv.org/pdf/2606.03768v1", "arxiv_id": "2606.03768", "doi": null, "citation_count": 0, "influential_citation_count": 0, "has_code": false, "code_url": null, "venue": null, "quality_score": 0.35}
{"id": "3334d58c7548bd50708ee71a66f46205f0fec1bc22f88e899b0da4ac4f9c6d0a", "sources": ["arxiv", "semantic_scholar"], "title": "Agentic Chain-of-Thought Steering for Efficient and Controllable LLM Reasoning", "abstract": "Large language models improve final-answer accuracy through extended chain-of-thought reasoning, but often spend tokens inefficiently and offer little inference-time control. Existing efficient reasoning methods control thinking length by shortening, early-stopping, or compressing traces, leaving how the model thinks implicit. In this paper, we propose Agentic Chain-of-Thought Steering (ACTS), which formulates reasoning steering as a Markov decision process where a controller agent adaptively steers a frozen reasoner during inference. At each step, the controller observes the reasoning trace and remaining thinking budget, then issues a steering action consisting of a reasoning strategy and a steering phrase that initiates the next reasoner step. This enables budget-aware strategy control for efficient reasoning while preserving the reasoner's generation continuity. We initialize the controller agent from our constructed synthetic steering trajectories with multi-budget augmentation, and further optimize it via reinforcement learning with budget-conditioned reward shaping. Experiments across multiple benchmarks show that ACTS matches full-thinking performance with substantial token savings, and enables controllable accuracy-efficiency trade-offs across different reasoners and tasks. The code is available at https://github.com/Andree-9/ACTS.", "authors": ["Yu Xia", "Zhouhang Xie", "Xin Xu", "Byungkyu Kang", "Prarit Lamba", "Xiang Gao", "Julian McAuley"], "categories": ["cs.CL", "cs.AI"], "fields_of_study": ["Computer Science"], "published_date": "2026-06-02", "url": "https://arxiv.org/abs/2606.03965", "pdf_url": "https://arxiv.org/pdf/2606.03965v1", "arxiv_id": "2606.03965", "doi": null, "citation_count": 0, "influential_citation_count": 0, "has_code": true, "code_url": "https://github.com/Andree-9/ACTS", "venue": null, "quality_score": 0.65}
{"id": "877194799d6a291a62d5389c2083cb61ff34c39196b810133a44238a05873085", "sources": ["arxiv", "semantic_scholar"], "title": "Unveiling the Entropy Dynamics of Chain-of-Thought Reasoning", "abstract": "This paper investigates the entropy dynamics of Chain-of-Thought (CoT) and uncovers a consistent two-phase structure: an Uncertainty Region of exploration transitioning sharply to a Confidence Region of convergence. We demonstrate that the Confidence Region possesses two critical properties: 1) High Reliability -- answers in the confidence region become highly accurate and stable, and 2) High Redundancy -- models generate unnecessary tokens long after reaching the correct answer. These properties unlock more efficient and reliable inference strategies: 1) Early Exit leverages reliability and redundancy to terminate computation safely when returns diminish, and 2)Test-Time Scaling uses the Confidence Region signal to prioritize converged trajectories. To operationalize these insights, we formulate Confidence Region detection as a sequential change-point detection problem, being the first to apply classical change-point methods to monitor CoT reasoning. Using the Cumulative Sum (CUSUM) algorithm, a statistically optimal change-point detector, we develop a training-free framework for real-time inference control. Experiments show our approach establishes a superior Pareto-frontier for early exit. CUSUM achieves 63.06% accuracy with 11.1% token reduction, outperforming DEER and Dynasor by 3.28% and 4.36% in accuracy respectively. For test-time scaling, CUSUM-weighted voting consistently outperforms self-consistency.", "authors": ["Ting Xu", "Xu He", "Yupu Lu", "Jiankai Sun", "Dong Li", "Wai Lam", "Jianye Hao"], "categories": ["cs.CL", "cs.LG"], "fields_of_study": ["Computer Science"], "published_date": "2026-06-01", "url": "https://arxiv.org/abs/2606.02020", "pdf_url": "https://arxiv.org/pdf/2606.02020v1", "arxiv_id": "2606.02020", "doi": null, "citation_count": 0, "influential_citation_count": 0, "has_code": false, "code_url": null, "venue": null, "quality_score": 0.35}
{"id": "4b2836c5545119e8dce572ec418d7525212188a54dde882d7e75f61f888d0e1c", "sources": ["arxiv", "semantic_scholar"], "title": "Spectral-Progressive Thought Flow for Lightweight Multimodal Reasoning", "abstract": "Multimodal spatial reasoning often relies on long chains of intermediate textual and visual thoughts, where accumulating visual tokens and dense cross-modal attention incur substantial computation and memory overhead. To address this challenge, we propose Spectral-Progressive Thought Flow (SpecFlow), a novel lightweight multimodal spatial reasoning framework that represents intermediate visual thoughts in a fixed-size discrete cosine space. By exploiting strong energy compaction, SpecFlow preserves global layout and relational structure while introducing high-frequency details only when increased spatial precision is required. To align visual state evolution with linguistic intent, classifier-free guidance enables autoregressive textual thoughts to steer flow-based updates of the visual workspace/state without expanding the context. As a result, SpecFlow maintains a bounded visual workspace whose updates depend only on the current visual state and accumulated textual trace, enabling long-horizon inference with stable latency and memory usage independent of reasoning depth. Empirical results show that SpecFlow achieves competitive or superior reasoning performance while reducing computation and KV cache costs by up to 2.1 times.", "authors": ["Yixian Shen", "Zhiheng Yang", "Qi Bi", "Changshuo Wang", "Shuai Wang", "Jia-Hong Huang", "George Floros", "Prayag Tiwari", "Anuj Pathania"], "categories": ["cs.LG"], "fields_of_study": ["Computer Science"], "published_date": "2026-06-01", "url": "https://arxiv.org/abs/2606.02842", "pdf_url": "https://arxiv.org/pdf/2606.02842v1", "arxiv_id": "2606.02842", "doi": null, "citation_count": 0, "influential_citation_count": 0, "has_code": false, "code_url": null, "venue": null, "quality_score": 0.35}
{"id": "c33f5f383762c8669c600081a7a157e4f1280bbad20e52bdc64954473f2ef74b", "sources": ["arxiv", "semantic_scholar"], "title": "Echoes within the Reasoning: Stealthy and Effective Watermarking via Chain of Thought", "abstract": "Large Language Models with Chain-of-Thought reasoning capabilities represent valuable intellectual property, yet existing black-box watermarking methods often trade robustness for reasoning fidelity by perturbing final answers or relying on fragile trigger patterns. We propose BiCoT, a watermarking framework that embeds ownership signals into the internal geometry of reasoning traces by aligning high-saliency structural anchors with a private signature subspace while regularizing ordinary control tokens to preserve semantic capacity. This design couples the watermark with reasoning-relevant representations, making removal difficult without disrupting the features that support coherent reasoning. To enable verification under model theft and representation drift, we introduce Robust Subspace Registration (RSR), a Top- logprob-based black-box verifier that uses sentinel tokens to calibrate systematic shifts in the output distribution. Experiments show that BiCoT preserves reasoning fidelity across diverse complex reasoning tasks while achieving robust detection under fine-tuning, quantization, model-level perturbations, and adaptive output-level attacks across in-domain and out-of-distribution settings.", "authors": ["Jiacheng Lu", "Yiming Li", "Tao Song", "Weijian Wang", "Wenjie Qu", "Haibing Guan", "Jiaheng Zhang"], "categories": ["cs.CR", "cs.LG"], "fields_of_study": ["Computer Science"], "published_date": "2026-05-27", "url": "https://arxiv.org/abs/2605.28890", "pdf_url": "https://arxiv.org/pdf/2605.28890v1", "arxiv_id": "2605.28890", "doi": null, "citation_count": 0, "influential_citation_count": 0, "has_code": false, "code_url": null, "venue": null, "quality_score": 0.35}
{"id": "c2d64e9668a6a5023bd78f4679b2644a212d8b6cdf13d420c3a46a00b6644901", "sources": ["arxiv", "semantic_scholar"], "title": "Boosting Inference with Guided Reasoning: Stochastic Exploration for Recursive Models", "abstract": "Recent work on recursive architectures has shown that tiny neural networks can be surprisingly powerful on structured reasoning tasks. The trick is to model reasoning trajectories with a latent dynamical system. We argue that the inference-time behaviour of these architectures is best understood as approximate inference over latent reasoning trajectories, with deterministic recursion as the one-particle, zero-noise limit. We make this view operational through guided stochastic exploration: stochastic perturbations of the reasoning dynamics propose neighbouring trajectories, and the model's existing early-stopping head reweights them online. The framework yields three label-free diagnostics: local stability, guide alignment, and cloud-token entropy. These predict, from inference traces alone, whether the procedure will help and which of its outputs to trust. On Sudoku-Extreme it lifts exact-solve accuracy from $85.9\\%$ to $98.0\\%$ without retraining; on Maze-Hard the diagnostics flag a misaligned guide, as validation performance later confirms. The same machinery thus characterises both when recursive reasoning has room to improve at the trajectory level and when the model's internal guide can recover it.", "authors": ["Andrew Corbett", "Archit Sood", "Anna Tzatzopoulou", "Sai-Aakash Ramesh", "Tim Dodwell"], "categories": ["cs.AI"], "fields_of_study": ["Computer Science"], "published_date": "2026-05-24", "url": "https://arxiv.org/abs/2605.25230", "pdf_url": "https://arxiv.org/pdf/2605.25230v2", "arxiv_id": "2605.25230", "doi": null, "citation_count": 0, "influential_citation_count": 0, "has_code": false, "code_url": null, "venue": null, "quality_score": 0.35}
{"id": "16fb0e7b90437465881ebb2faa0330f209a148002c2992b4225f6aed1fed93e4", "sources": ["arxiv", "semantic_scholar"], "title": "Pause and Reflect: Conformal Aggregation for Chain-of-Thought Reasoning", "abstract": "Chain-of-thought (CoT) reasoning with self-consistency improves performance by aggregating multiple sampled reasoning paths. In this setting, correctness is no longer tied to a single reasoning trace but to the aggregation rule over a pool of candidate paths, making aggregation uncertainty the central challenge. This issue is critical where confidently incorrect answers are far more costly than abstentions. We introduce a conformal procedure for CoT reasoning that directly addresses aggregation uncertainty. Our approach replaces majority voting with weighted score aggregation over reasoning paths and calibrates an abstention rule using conformal risk control. This approach leads to finite-sample guarantees on the confident-error rate--the probability that the system answers and is wrong. We further identify score separability as the key condition under which abstention provably improves selective accuracy, and derive closed-form expressions that predict accuracy gains from calibration data alone. The method is fully inference-time, and requires no retraining. Across four benchmarks, four open-source models, and three score classes, realized confident-error rates are consistent with the prescribed targets up to calibration-split and test-set variability. Our method achieves $90.1\\%$ selective accuracy on GSM8K by abstaining on less than $5\\%$ of problems, compared with $82\\%$ accuracy under majority-voting baseline.", "authors": ["Yu Gu", "Zijun Yu", "Vahid Partovi Nia", "Masoud Asgharian"], "categories": ["stat.ML", "cs.CL", "cs.LG"], "fields_of_study": ["Mathematics", "Computer Science"], "published_date": "2026-05-13", "url": "https://arxiv.org/abs/2605.14098", "pdf_url": "https://arxiv.org/pdf/2605.14098v1", "arxiv_id": "2605.14098", "doi": null, "citation_count": 0, "influential_citation_count": 0, "has_code": true, "code_url": null, "venue": null, "quality_score": 0.65}
{"id": "420073b4bdbb37ead37537d90ff681b7b79452fbf14c0877a11b1170a485f8c7", "sources": ["arxiv", "semantic_scholar"], "title": "Drop the Act: Probe-Filtered RL for Faithful Chain-of-Thought Reasoning", "abstract": "Reasoning models post-hoc rationalize answers they have already committed to internally, producing chains of *reasoning theater*: deliberative-looking steps that contribute nothing to correctness. This wastes inference tokens, pollutes interpretability, and obscures what the model actually computed. We introduce **ProFIL** (**Pro**be-**Fil**tered Reinforcement Learning) to *reduce theater, increase chain-of-thought faithfulness, and shrink chain length* in a single, drop-in extension to Group Relative Policy Optimization (GRPO). A multi-head attention probe is trained *once* on the *frozen* base model to detect post-commitment steps from internal activations alone; during GRPO, rollouts whose probe score exceeds a threshold have their advantage zeroed. *Our central finding is that a probe trained on a frozen base, with verifier-derived labels and no human annotation, provides a stable signal that suppresses theater while resisting the RL-obfuscation failure mode predicted by prior work.* Across four reasoning domains (GSM8K, LiveCodeBench, ToolUse, MMLU-Redux) and two model architectures (Llama-8B, Qwen-7B), ProFIL reduces post-commitment theater by **11--100%**, raises faithful-fraction (e.g., +24pp on LiveCodeBench under an independent Claude 3.7 Sonnet judge), and shortens chains by 4--19%, all while preserving or improving task accuracy. ProFIL also beats a matched length-penalty GRPO baseline, isolating the gain as semantic commitment-detection rather than chain compression. Probe weights, training configurations, and rollouts are released across all four domains.", "authors": ["Swapnil Parekh"], "categories": ["cs.LG", "cs.AI"], "fields_of_study": ["Computer Science"], "published_date": "2026-05-12", "url": "https://arxiv.org/abs/2605.11467", "pdf_url": "https://arxiv.org/pdf/2605.11467v1", "arxiv_id": "2605.11467", "doi": null, "citation_count": 0, "influential_citation_count": 0, "has_code": false, "code_url": null, "venue": null, "quality_score": 0.35}
{"id": "a67c077356fcb5fc93431e9d3285cbf424040872c4410515229a86ea512153e2", "sources": ["arxiv", "semantic_scholar"], "title": "TMAS: Scaling Test-Time Compute via Multi-Agent Synergy", "abstract": "Test-time scaling has become an effective paradigm for improving the reasoning ability of large language models by allocating additional computation during inference. Recent structured approaches have further advanced this paradigm by organizing inference across multiple trajectories, refinement rounds, and verification-based feedback. However, existing structured test-time scaling methods either weakly coordinate parallel reasoning trajectories or rely on noisy historical information without explicitly deciding what should be retained and reused, limiting their ability to balance exploration and exploitation. In this work, we propose TMAS, a framework for scaling test-time compute via multi-agent synergy. TMAS organizes inference as a collaborative process among specialized agents, enabling structured information flow across agents, trajectories, and refinement iterations. To support effective cross-trajectory collaboration, TMAS introduces hierarchical memories: the experience bank reuses low-level reliable intermediate conclusions and local feedback, while the guideline bank records previously explored high-level strategies to steer subsequent rollouts away from redundant reasoning patterns. Furthermore, we design a hybrid reward reinforcement learning scheme tailored to TMAS, which jointly preserves basic reasoning capability, enhances experience utilization, and encourages exploration beyond previously attempted solution strategies. Extensive experiments on challenging reasoning benchmarks show that TMAS achieves stronger iterative scaling than existing test-time scaling baselines, with hybrid reward training further improving scaling effectiveness and stability across iterations. Code and data are available at https://github.com/IQuestLab/tmas.", "authors": ["George Wu", "Nan Jing", "Qing Yi", "Chuan Hao", "Ming Yang", "Feng Chang", "Yuan Wei", "Jian Yang", "Ran Tao", "Bryan Dai"], "categories": ["cs.AI"], "fields_of_study": ["Computer Science"], "published_date": "2026-05-11", "url": "https://arxiv.org/abs/2605.10344", "pdf_url": "https://arxiv.org/pdf/2605.10344v2", "arxiv_id": "2605.10344", "doi": null, "citation_count": 0, "influential_citation_count": 0, "has_code": true, "code_url": "https://github.com/IQuestLab/tmas", "venue": null, "quality_score": 0.65}
{"id": "e3060e2c4073895799c97c486c0eefeb178b6dd640de0672ae62ce0c9661c674", "sources": ["arxiv", "semantic_scholar"], "title": "Latent Chain-of-Thought Improves Structured-Data Transformers", "abstract": "Chain-of-thought and more broadly test-time compute are known to augment the expressive capabilities of language models and have led to major innovations in reasoning. Motivated by this success, this paper explores latent chain-of-thought as well as the impact of depth and looping for time-series and tabular data. We propose a recurrent scheme in which a structured-data transformer, after an initial forward pass, compresses its query-position hidden states into feedback tokens that are appended to the input and processed again, allowing multiple rounds of latent computation before prediction. We compare CoT models against a same-depth no-CoT baseline, a deeper baseline matched to the CoT model in effective depth, and a looped transformer with weight-tied recurrence but no additional chain-of-thought tokens. Across 36 datasets in time-series forecasting and tabular prediction, latent chain-of-thought improves over the baseline on 7/9 time-series datasets (+12.63\\% average gain) and 23/27 tabular datasets (+3.25\\% average gain), with CoT models performing best on average in both settings. We also show that the benefit of CoT extends to pretrained foundation models: applying latent CoT to nanoTabPFN, a small open-source tabular foundation model, improves its performance above the much larger TabPFN-v2 on TabArena. Together, these results demonstrate that chain-of-thought is a useful axis for scaling test-time compute for structured data.", "authors": ["Carson Dudley", "Samet Oymak"], "categories": ["cs.LG"], "fields_of_study": ["Computer Science"], "published_date": "2026-05-11", "url": "https://arxiv.org/abs/2605.11262", "pdf_url": "https://arxiv.org/pdf/2605.11262v2", "arxiv_id": "2605.11262", "doi": null, "citation_count": 0, "influential_citation_count": 0, "has_code": true, "code_url": null, "venue": null, "quality_score": 0.65}
{"id": "0892fb3b57067ca316c2839e47243fc786bd9064f3689608064faefa9f258e80", "sources": ["arxiv", "semantic_scholar"], "title": "Rethinking Dense Sequential Chains: Reasoning Language Models Can Extract Answers from Sparse, Order-Shuffling Chain-of-Thoughts", "abstract": "Modern reasoning language models generate dense, sequential chain-of-thought traces implicitly assuming that every token contributes and that steps must be consumed in order. We challenge both assumptions through a systematic intervention pipeline--removal, masking, shuffling, and noise injection--applied to model-generated reasoning chains across three models and three benchmarks. Our findings are counterintuitive on three dimensions. Order: Does the sequential order of a reasoning chain matter for answer extraction? No--line-level shuffling reduces accuracy by less than 0.5 pp; word-level shuffling retains 62%-89% accuracy; only token-level shuffling collapses to near zero. Pretrained-only and instruction-tuned variants exhibit near-identical tolerance (78.67% vs. 78.00% under line shuffling), indicating order-independence originates from pretraining rather than reasoning-specific fine-tuning. Dense: Is all the information in a reasoning chain important for answer extraction? No--masking numeric digits collapses accuracy to exactly 0%, while masking alphabetic prose improves accuracy by 4.7 pp. Robustness: Is a reasoning chain that is both order-shuffling and non-dense still robust? Yes--the most aggressively reduced representation (all natural language removed, lines arbitrarily shuffled) still achieves 83% accuracy, and injecting false answers at 3x true-answer frequency leaves accuracy unchanged (83.3%->83.3%), falsifying a frequency-based extraction account. These results establish that answer extraction operates on a sparse, order-insensitive, and structurally robust informational substrate, opening paths toward parallelized and token-efficient reasoning generation.", "authors": ["Yi-Chang Chen", "Feng-Ting Liao", "Da-shan Shiu", "Hung-yi Lee"], "categories": ["cs.CL"], "fields_of_study": ["Computer Science"], "published_date": "2026-05-08", "url": "https://arxiv.org/abs/2605.07307", "pdf_url": "https://arxiv.org/pdf/2605.07307v1", "arxiv_id": "2605.07307", "doi": null, "citation_count": 1, "influential_citation_count": 0, "has_code": false, "code_url": null, "venue": null, "quality_score": 0.35}
{"id": "0a0a80a69160bdc2ced10db021442478c4a639fbd174b5ac4d63b5bb17795f8e", "sources": ["arxiv", "semantic_scholar"], "title": "Multi-Agent Reasoning Improves Compute Efficiency: Pareto-Optimal Test-Time Scaling", "abstract": "Advances in inference methods have enabled language models to improve their predictions without additional training. These methods often prioritize raw performance over cost-effective compute usage. However, computational efficiency is key for real-world applications with resource constraints. We provide a systematic analysis of the inference scaling strategies self-consistency, self-refinement, multi-agent debate, and mixture-of-agents, to study their computational performance tradeoffs. We evaluate methods on two reasoning benchmarks (MMLU-Pro, BBH) and include extensive parameter configurations (e.g., scaling the number of parallel predictions, agents, and debate rounds) across different model sizes. Across 34 configurations and over 100 evaluations, we compute the Pareto-optimal front to select methods that achieve the best accuracy with the lowest computational budget. Notably, inference scaling improves accuracy by up to +7.1% points over chain-of-thought at the highest evaluated budgets (20x the CoT compute budget) on MMLU-Pro. With an equal computing budget, debate and mixture-of-agents outperform self-consistency by 1.3% and 2.7% points, respectively. While self-consistency saturates earlier, multi-agent gains persist, particularly on more complicated tasks. We identify a simple multi-agent design guideline: mixture-of-agents is most efficient when the number of parallel generations exceeds the number of sequential aggregations.", "authors": ["Florian Valentin Wunderlich", "Lars Benedikt Kaesberg", "Jan Philip Wahle", "Terry Ruas", "Bela Gipp"], "categories": ["cs.AI"], "fields_of_study": ["Computer Science"], "published_date": "2026-05-02", "url": "https://arxiv.org/abs/2605.01566", "pdf_url": "https://arxiv.org/pdf/2605.01566v1", "arxiv_id": "2605.01566", "doi": null, "citation_count": 0, "influential_citation_count": 0, "has_code": false, "code_url": null, "venue": null, "quality_score": 0.35}
{"id": "54fbf8078843e7f03d3f2db7716cf9ca732718f908a9a3cdb90d8f8fc7ee1202", "sources": ["arxiv", "semantic_scholar"], "title": "Compute Aligned Training: Optimizing for Test Time Inference", "abstract": "Scaling test-time compute has emerged as a powerful mechanism for enhancing Large Language Model (LLM) performance. However, standard post-training paradigms, Supervised Fine-Tuning (SFT) and Reinforcement Learning (RL), optimize the likelihood of individual samples under a base policy, creating a misalignment with test time procedures that rely on aggregated or filtered outputs. In this work, we propose Compute Aligned Training, which aligns training objectives with test-time strategies. By conceptualizing inference strategies as operators on the base policy, we derive new loss functions that maximize performance when said strategies are applied. We instantiate such loss functions for SFT and RL across common test time strategies. Finally, we provide empirical evidence that this training method substantially improves test time scaling over standard training.", "authors": ["Adam Ousherovitch", "Ambuj Tewari"], "categories": ["cs.LG", "cs.AI"], "fields_of_study": ["Computer Science"], "published_date": "2026-04-27", "url": "https://arxiv.org/abs/2604.24957", "pdf_url": "https://arxiv.org/pdf/2604.24957v2", "arxiv_id": "2604.24957", "doi": "10.48550/arXiv.2604.24957", "citation_count": 0, "influential_citation_count": 0, "has_code": false, "code_url": null, "venue": "arXiv.org", "quality_score": 0.55}
{"id": "10e315a723d507675407e7cabe22dca1f1ce2b3425fe98cf76442f6089c7bd8e", "sources": ["arxiv", "semantic_scholar"], "title": "Thoughts-as-Planning: Latent World Models for Chain-of-Thoughts Optimization via Reinforcement Planning", "abstract": "The success of large language models (LLMs) across diverse NLP tasks has elevated the importance of reasoning chain optimization as a critical step in aligning model behavior with task objectives. Existing reasoning chain tuning methods often rely on black-box heuristics or gradient-free search, which lack interpretability, generalization, and sample efficiency. In this work, we introduce \\textbf{Thoughts-as-Planning}, a novel framework that formalizes reasoning chain optimization as a sequential decision-making process over a latent semantic space. We model the LLM as a partially observable environment and learn a latent world model that simulates the effect of reasoning chain edits on downstream outputs. A proximity-preserving embedding space is constructed to encode reasoning chain-response dynamics, enabling planning via gradient descent or reinforcement learning. Our method supports multi-scale abstraction, allowing reasoning chain edits at token, segment, and instruction levels to be integrated into a unified planner. Through extensive experiments on language understanding and generation tasks, we demonstrate that Thoughts-as-Planning outperforms state-of-the-art reasoning chain tuning baselines in efficiency, robustness, and generalization, while offering interpretability through its structured planning trajectory. Our code is available at https://github.com/FastLM/Thoughts-as-Planning.", "authors": ["Dong Liu", "Yanxuan Yu", "Ying Nian Wu"], "categories": ["cs.CL", "cs.AI"], "fields_of_study": ["Computer Science", "Biology"], "published_date": "2026-04-27", "url": "https://arxiv.org/abs/2605.28842", "pdf_url": "https://arxiv.org/pdf/2605.28842v1", "arxiv_id": "2605.28842", "doi": "10.64898/2026.05.10.724161", "citation_count": 0, "influential_citation_count": 0, "has_code": true, "code_url": "https://github.com/FastLM/Thoughts-as-Planning", "venue": "bioRxiv", "quality_score": 0.85}
{"id": "7920ca456a052177f62b072204a767269b1a6d90d416b483311c63f74182c9db", "sources": ["arxiv", "semantic_scholar"], "title": "Ulterior Motives: Detecting Misaligned Reasoning in Continuous Thought Models", "abstract": "Chain-of-Thought (CoT) reasoning has emerged as a key technique for eliciting complex reasoning in Large Language Models (LLMs). Although interpretable, its dependence on natural language limits the model's expressive bandwidth. Continuous thought models address this bottleneck by reasoning in latent space rather than human-readable tokens. While they enable richer representations and faster inference, they raise a critical safety question: how can we detect misaligned reasoning in an uninterpretable latent space? To study this, we introduce MoralChain, a benchmark of 12,000 social scenarios with parallel moral/immoral reasoning paths. We train a continuous thought model with backdoor behavior using a novel dual-trigger paradigm - one trigger that arms misaligned latent reasoning ([T]) and another that releases harmful outputs ([O]). We demonstrate three findings: (1) continuous thought models can exhibit misaligned latent reasoning while producing aligned outputs, with aligned and misaligned reasoning occupying geometrically distinct regions of latent space; (2) linear probes trained on behaviorally-distinguishable conditions ([T][O] vs [O]) transfer to detecting armed-but-benign states ([T] vs baseline) with high accuracy; and (3) misalignment is encoded in early latent thinking tokens, suggesting safety monitoring for continuous thought models should target the \"planning\" phase of latent reasoning.", "authors": ["Sharan Ramjee"], "categories": ["cs.AI", "cs.CL", "cs.LG"], "fields_of_study": ["Computer Science"], "published_date": "2026-04-25", "url": "https://arxiv.org/abs/2604.23460", "pdf_url": "https://arxiv.org/pdf/2604.23460v1", "arxiv_id": "2604.23460", "doi": "10.48550/arXiv.2604.23460", "citation_count": 0, "influential_citation_count": 0, "has_code": false, "code_url": null, "venue": "arXiv.org", "quality_score": 0.55}
{"id": "1d31ea5bdbcef31e58c0aee61b18fd0bc4da1c484108e712eafd8e644c06bf77", "sources": ["arxiv", "semantic_scholar"], "title": "CAP-CoT: Cycle Adversarial Prompt for Improving Chain of Thoughts in LLM Reasoning", "abstract": "Chain-of-Thought (CoT) prompting has emerged as a simple and effective way to elicit step-by-step solutions from large language models (LLMs). However, CoT reasoning can be unstable across runs on long, multi-step problems, leading to inconsistent answers for unchanged task. Most prior work focuses on improving the forward reasoning chain within a single pass, with less attention to iterative and contrastive correction. To address this gap, we propose CAP-CoT, a Cycle Adversarial Prompt optimization framework designed to improve both CoT reasoning accuracy and stability of a single deployed solver. In each cycle, a forward solver generates candidate reasoning chains, an adversarial challenger constructs plausible but deliberately flawed chains using targeted error strategies, and a feedback agent contrasts the two chains and produces step-aligned structured feedback. This feedback closes the optimization loop in two directions, including updating the solver prompt based on errors exposed by the challenger, and updating the challenger prompt to generate increasingly targeted errors in subsequent cycles. Unlike safety-oriented adversarial prompting such as jailbreak or prompt-injection attacks, our adversarial component is task-semantic and aims to expose logical vulnerabilities in reasoning chains. Experiments across six benchmarks and four LLM backbones demonstrate that within two to three adversarial prompt optimization cycles, CAP-CoT consistently reduces variability across runs while improving reasoning accuracy and robustness to prompt perturbations.", "authors": ["Shuxu Chen", "Yitian Zhou", "Jiaquan Zhang", "Haoyu Bian", "Aming Wu", "Sungyoung Lee", "Chaoning Zhang", "Hyundong Shin"], "categories": ["cs.AI"], "fields_of_study": ["Computer Science"], "published_date": "2026-04-25", "url": "https://arxiv.org/abs/2604.23270", "pdf_url": "https://arxiv.org/pdf/2604.23270v1", "arxiv_id": "2604.23270", "doi": "10.48550/arXiv.2604.23270", "citation_count": 0, "influential_citation_count": 0, "has_code": false, "code_url": null, "venue": "arXiv.org", "quality_score": 0.55}
{"id": "51dd73db0adae36378ed9798308592414f75c1bec937fa92c3464d09f5e10717", "sources": ["arxiv", "semantic_scholar"], "title": "Calibration Drift Under Reasoning: How Chain-of-Thought Budgets Induce Overconfidence in Large Language Models", "abstract": "The ability of large language models (LLMs) to express calibrated uncertainty is important for safe deployment. Chain-of-thought (CoT) reasoning is widely used to improve accuracy and reliability, but its effect on calibration is not fully understood. We show that this picture is incomplete: in some settings, increasing the reasoning budget beyond a task-specific threshold can cause models to become systematically overconfident, assigning high confidence to incorrect answers. We call this phenomenon Calibration Drift Under Reasoning (CDUR) and study it both theoretically and empirically. We define reasoning budget B and analyze conditions under which Expected Calibration Error ECE(B) follows a non-monotonic pattern: it first decreases as reasoning corrects errors, then increases as longer reasoning produces internally consistent but incorrect explanations. We propose a Hypothesis Lock-In model based on autoregressive generation to explain this behavior. We evaluate Llama-3.1-8B and Llama-3.3-70B on 47 reasoning-trap questions across four reasoning budgets and three seeds (1,368 API calls; 574 valid responses). The 8B model shows non-monotonic calibration behavior, while results for the 70B model are limited to baseline evaluation and are inconclusive for budget-dependent effects. We introduce CABStop, a calibration-aware stopping rule that halts reasoning when confidence diverges from an auxiliary accuracy estimate. These results suggest that increasing reasoning depth does not always improve reliability and should be monitored carefully.", "authors": ["Prakul Sunil Hiremath", "Harshit R. Hiremath"], "categories": ["cs.CL", "cs.AI", "cs.LG"], "fields_of_study": ["Computer Science"], "published_date": "2026-04-24", "url": "https://arxiv.org/abs/2606.11211", "pdf_url": "https://arxiv.org/pdf/2606.11211v1", "arxiv_id": "2606.11211", "doi": null, "citation_count": 0, "influential_citation_count": 0, "has_code": false, "code_url": null, "venue": null, "quality_score": 0.35}
{"id": "08b537b2ae446c36e9101264f6c00b3020b852889a24ab5ff84989581896e72f", "sources": ["arxiv", "semantic_scholar"], "title": "Thinking Without Words: Efficient Latent Reasoning with Abstract Chain-of-Thought", "abstract": "While long, explicit chains-of-thought (CoT) have proven effective on complex reasoning tasks, they are costly to generate during inference. Non-verbal reasoning methods have emerged with shorter generation lengths by leveraging continuous representations, yet their performance lags behind verbalized CoT. We propose $\\textbf{Abstract Chain-of-Thought}$, a discrete latent reasoning post-training mechanism in which the language model produces a short sequence of tokens from a reserved vocabulary in lieu of a natural language CoT, before generating a response. To make previously unseen ''abstract'' tokens useful, we introduce a policy iteration-style warm-up loop that alternates between (i.) bottlenecking from a verbal CoT via masking and performing supervised fine-tuning, and (ii.) self-distillation by training the model to generate abstract tokens from the prompt alone via constrained decoding with the codebook. After warm-up, we optimize the generation of abstract sequences with warm-started reinforcement learning under constrained decoding. Abstract-CoT achieves up to $11.6\\times$ fewer reasoning tokens while demonstrating comparable performance across mathematical reasoning, instruction-following, and multi-hop reasoning, and generalizes across language model families. We also find an emergent power law distribution over the abstract vocabulary, akin to those seen in natural language, that evolves across the training phases. Our findings highlight the potential for post-training latent reasoning mechanisms that enable efficient inference through a learned abstract reasoning language.", "authors": ["Keshav Ramji", "Tahira Naseem", "Ramón Fernandez Astudillo"], "categories": ["cs.CL"], "fields_of_study": ["Computer Science"], "published_date": "2026-04-24", "url": "https://arxiv.org/abs/2604.22709", "pdf_url": "https://arxiv.org/pdf/2604.22709v2", "arxiv_id": "2604.22709", "doi": "10.48550/arXiv.2604.22709", "citation_count": 3, "influential_citation_count": 0, "has_code": false, "code_url": null, "venue": "arXiv.org", "quality_score": 0.55}
{"id": "1a406f868c12bd57787b2b6a552870075888a0701cd656c49fb458d9184f7bd9", "sources": ["arxiv", "semantic_scholar"], "title": "Bridging the Reasoning Gap in Vietnamese with Small Language Models via Test-Time Scaling", "abstract": "The democratization of ubiquitous AI hinges on deploying sophisticated reasoning capabilities on resource-constrained devices. However, Small Language Models (SLMs) often face a \"reasoning gap\", particularly in non-English languages like Vietnamese, where they struggle to maintain coherent chains of thought. This paper investigates Test-Time Scaling strategies for the Qwen3-1.7B architecture within the context of Vietnamese Elementary Mathematics. We introduce Vi-S1K, a high-fidelity reasoning dataset localized via a Gemini 2.5 Flash-Lite powered pipeline, and Vi-Elementary-Bench, a dual-resource benchmark for rigorous evaluation. Using an LLM-as-a-Judge protocol, we reveal that the base model possesses robust latent knowledge (Accuracy: 4.05/5.00) but suffers from a severe \"formatting gap\" in communication. Supervised Fine-Tuning (SFT) acts as a critical \"reasoning unlocker\", yielding a 77% improvement in Explanation Quality and bridging the gap between raw calculation and pedagogical coherence. Furthermore, our analysis of prompting strategies uncovers a significant trade-off: structured frameworks like ReAct impose a \"cognitive tax\" on the 1.7B parameter capacity, degrading performance relative to pure Chain-of-Thought (CoT) combined with Self-Consistency. These findings establish a deployment hierarchy for SLMs, demonstrating that SFT combined with simplified test-time scaling is superior to complex agentic workflows for edge-based reasoning.", "authors": ["Bui The Trung", "Do Minh Duc", "Nguyen Van Vinh", "Bui Nguyen Quoc Trinh"], "categories": ["cs.CL", "cs.AI"], "fields_of_study": ["Computer Science"], "published_date": "2026-04-20", "url": "https://arxiv.org/abs/2604.17794", "pdf_url": "https://arxiv.org/pdf/2604.17794v1", "arxiv_id": "2604.17794", "doi": "10.48550/arXiv.2604.17794", "citation_count": 0, "influential_citation_count": 0, "has_code": false, "code_url": null, "venue": "arXiv.org", "quality_score": 0.55}
{"id": "7ba3c8dac4c812789fba60e24f9a640c641b6e178d4367e44fdb00573c902b45", "sources": ["arxiv", "semantic_scholar"], "title": "Efficient Test-Time Scaling via Temporal Reasoning Aggregation", "abstract": "Test-time scaling improves the reasoning performance of large language models but often results in token-inefficient overthinking, where models continue reasoning beyond what is necessary for a correct answer. Existing dynamic early-exit methods typically rely on single-step confidence signals, which are often unreliable for detecting reasoning convergence in multi-step settings. To mitigate this limitation, we propose TRACE, a training-free framework for efficient test-time scaling that determines when to terminate reasoning based on temporal aggregation of multi-step evidence rather than instantaneous signals. TRACE detects reasoning convergence over time by aggregating two complementary signals across recent reasoning steps: answer consistency, capturing the persistence of predicted answers, and confidence trajectory, modeling the temporal evolution of model confidence. Benefiting from these two factors, TRACE can accurately determine whether the reasoning process has converged, thereby promptly halting inference and effectively avoiding redundant reasoning steps. Extensive experiments on multiple challenging benchmarks show that TRACE reduces reasoning token usage by 25-30% on average while maintaining accuracy within 1-2% of full-length reasoning, consistently outperforming existing dynamic reasoning methods.", "authors": ["Jiakun Li", "Xingwei He", "Kefan Li", "Hongzheng Chai", "Hongyue Yu", "Yuan Yuan"], "categories": ["cs.AI"], "fields_of_study": ["Computer Science"], "published_date": "2026-04-19", "url": "https://arxiv.org/abs/2604.17304", "pdf_url": "https://arxiv.org/pdf/2604.17304v1", "arxiv_id": "2604.17304", "doi": "10.48550/arXiv.2604.17304", "citation_count": 0, "influential_citation_count": 0, "has_code": false, "code_url": null, "venue": "arXiv.org", "quality_score": 0.5489}
{"id": "3f049551d54f083da747e1173326c2334fa9f0a681dacd0b4d0616f82cee9e03", "sources": ["arxiv", "semantic_scholar"], "title": "LLM Reasoning Is Latent, Not the Chain of Thought", "abstract": "This position paper argues that large language model (LLM) reasoning should be studied as latent-state trajectory formation rather than as faithful surface chain-of-thought (CoT). This matters because claims about faithfulness, interpretability, reasoning benchmarks, and inference-time intervention all depend on what the field takes the primary object of reasoning to be. We ask what that object should be once three often-confounded factors are separated and formalize three competing hypotheses: H1, reasoning is primarily mediated by latent-state trajectories; H2, reasoning is primarily mediated by explicit surface CoT; and H0, most apparent reasoning gains are better explained by generic serial compute than by any privileged representational object. Reorganizing recent empirical, mechanistic, and survey work under this framework, and adding compute-audited worked exemplars that factorize surface traces, latent interventions, and matched budget expansions, we find that current evidence most strongly supports H1 as a default working hypothesis rather than as a task-independent verdict. We therefore make two recommendations: the field should treat latent-state dynamics as the default object of study for LLM reasoning, and it should evaluate reasoning with designs that explicitly disentangle surface traces, latent states, and serial compute.", "authors": ["Wenshuo Wang"], "categories": ["cs.AI"], "fields_of_study": ["Computer Science"], "published_date": "2026-04-17", "url": "https://arxiv.org/abs/2604.15726", "pdf_url": "https://arxiv.org/pdf/2604.15726v1", "arxiv_id": "2604.15726", "doi": "10.48550/arXiv.2604.15726", "citation_count": 0, "influential_citation_count": 0, "has_code": false, "code_url": null, "venue": "arXiv.org", "quality_score": 0.5466}
{"id": "b7c8cc2bb2a96867aeef5e419ac26c96832dc344a2a09af2fd0e42b8fd3c3fc6", "sources": ["arxiv", "semantic_scholar"], "title": "Adapting in the Dark: Efficient and Stable Test-Time Adaptation for Black-Box Models", "abstract": "Test-Time Adaptation (TTA) for black-box models accessible only via APIs remains a largely unexplored challenge. Existing approaches such as post-hoc output refinement offer limited adaptive capacity, while Zeroth-Order Optimization (ZOO) enables input-space adaptation but faces high query costs and optimization challenges in the unsupervised TTA setting. We introduce BETA (Black-box Efficient Test-time Adaptation), a framework that addresses these limitations by employing a lightweight, local white-box steering model to create a tractable gradient pathway. Through a prediction harmonization technique combined with consistency regularization and prompt learning-oriented filtering, BETA enables stable adaptation with no additional API calls and negligible latency beyond standard inference. On ImageNet-C, BETA achieves a +7.1% accuracy gain on ViT-B/16 and +3.4% on CLIP, surpassing strong white-box and gray-box methods including TENT and TPT. On a commercial API, BETA achieves comparable performance to ZOO at 250x lower cost while maintaining real-time inference speed, establishing it as a practical and efficient solution for real-world black-box TTA.", "authors": ["Yunbei Zhang", "Shuaicheng Niu", "Chengyi Cai", "Feng Liu", "Jihun Hamm"], "categories": ["cs.LG", "cs.CV"], "fields_of_study": ["Computer Science"], "published_date": "2026-04-17", "url": "https://arxiv.org/abs/2604.15609", "pdf_url": "https://arxiv.org/pdf/2604.15609v1", "arxiv_id": "2604.15609", "doi": "10.48550/arXiv.2604.15609", "citation_count": 0, "influential_citation_count": 0, "has_code": false, "code_url": null, "venue": "arXiv.org", "quality_score": 0.5466}
{"id": "177911204c8c30dcddbd3ef4b09529f145e36d26d4026c13addfdac5e6610579", "sources": ["arxiv", "semantic_scholar"], "title": "Structured Abductive-Deductive-Inductive Reasoning for LLMs via Algebraic Invariants", "abstract": "Large language models exhibit systematic limitations in structured logical reasoning: they conflate hypothesis generation with verification, cannot distinguish conjecture from validated knowledge, and allow weak reasoning steps to propagate unchecked through inference chains. We present a symbolic reasoning scaffold that operationalizes Peirce's tripartite inference -- abduction, deduction, and induction -- as an explicit protocol for LLM-assisted reasoning. The framework enforces logical consistency through five algebraic invariants (the Gamma Quintet), the strongest of which -- the Weakest Link bound -- ensures that no conclusion in a reasoning chain can exceed the reliability of its least-supported premise. This principle, independently grounded as weakest link resolution in possibilistic logic and empirically validated for chain-of-thought reasoning, prevents logical inconsistencies from accumulating across multi-step inference. We verify all invariants through a property-based testing suite of 100 properties and 16 fuzz tests over 10^5+ generated cases, providing a verified reference implementation of the invariants suitable as a foundation for future reasoning benchmarks.", "authors": ["Sankalp Gilda", "Shlok Gilda"], "categories": ["cs.AI", "cs.LG", "cs.LO"], "fields_of_study": ["Computer Science"], "published_date": "2026-04-17", "url": "https://arxiv.org/abs/2604.15727", "pdf_url": "https://arxiv.org/pdf/2604.15727v1", "arxiv_id": "2604.15727", "doi": "10.48550/arXiv.2604.15727", "citation_count": 0, "influential_citation_count": 0, "has_code": false, "code_url": null, "venue": "arXiv.org", "quality_score": 0.5466}
{"id": "3f4f9817b6710fd89fd891b02c0f418140d51a4f247d8a6f0fa7a969a9b9f5a0", "sources": ["arxiv", "semantic_scholar"], "title": "ProtoTTA: Prototype-Guided Test-Time Adaptation", "abstract": "Deep networks that rely on prototypes-interpretable representations that can be related to the model input-have gained significant attention for balancing high accuracy with inherent interpretability, which makes them suitable for critical domains such as healthcare. However, these models are limited by their reliance on training data, which hampers their robustness to distribution shifts. While test-time adaptation (TTA) improves the robustness of deep networks by updating parameters and statistics, the prototypes of interpretable models have not been explored for this purpose. We introduce ProtoTTA, a general framework for prototypical models that leverages intermediate prototype signals rather than relying solely on model outputs. ProtoTTA minimizes the entropy of the prototype-similarity distribution to encourage more confident and prototype-specific activations on shifted data. To maintain stability, we employ geometric filtering to restrict updates to samples with reliable prototype activations, regularized by prototype-importance weights and model-confidence scores. Experiments across four prototypical backbones on four diverse benchmarks spanning fine-grained vision, histopathology, and NLP demonstrate that ProtoTTA improves robustness over standard output entropy minimization while restoring correct semantic focus in prototype activations. We also introduce novel interpretability metrics and a vision-language model (VLM) evaluation framework to explain TTA dynamics, confirming ProtoTTA restores human-aligned semantic focus and correlates reliably with VLM-rated reasoning quality. Code is available at: https://github.com/DeepRCL/ProtoTTA.", "authors": ["Mohammad Mahdi Abootorabi", "Parvin Mousavi", "Purang Abolmaesumi", "Evan Shelhamer"], "categories": ["cs.LG", "cs.CV"], "fields_of_study": ["Computer Science"], "published_date": "2026-04-16", "url": "https://arxiv.org/abs/2604.15494", "pdf_url": "https://arxiv.org/pdf/2604.15494v1", "arxiv_id": "2604.15494", "doi": "10.48550/arXiv.2604.15494", "citation_count": 0, "influential_citation_count": 0, "has_code": true, "code_url": "https://github.com/DeepRCL/ProtoTTA", "venue": "arXiv.org", "quality_score": 0.8429}
{"id": "8076447c2685ddbdf8077674fa06a09d7133f9bdd8fc1deea1fb14c1e4fc6b95", "sources": ["arxiv", "semantic_scholar"], "title": "LongCoT: Benchmarking Long-Horizon Chain-of-Thought Reasoning", "abstract": "As language models are increasingly deployed for complex autonomous tasks, their ability to reason accurately over longer horizons becomes critical. An essential component of this ability is planning and managing a long, complex chain-of-thought (CoT). We introduce LongCoT, a scalable benchmark of 2,500 expert-designed problems spanning chemistry, mathematics, computer science, chess, and logic to isolate and directly measure the long-horizon CoT reasoning capabilities of frontier models. Problems consist of a short input with a verifiable answer; solving them requires navigating a graph of interdependent steps that span tens to hundreds of thousands of reasoning tokens. Each local step is individually tractable for frontier models, so failures reflect long-horizon reasoning limitations. At release, the best models achieve <10% accuracy (GPT 5.2: 9.8%; Gemini 3 Pro: 6.1%) on LongCoT, revealing a substantial gap in current capabilities. Overall, LongCoT provides a rigorous measure of long-horizon reasoning, tracking the ability of frontier models to reason reliably over extended periods.", "authors": ["Sumeet Ramesh Motwani", "Daniel Nichols", "Charles London", "Peggy Li", "Fabio Pizzati", "Acer Blake", "Hasan Hammoud", "Tavish McDonald", "Akshat Naik", "Alesia Ivanova", "Vignesh Baskaran", "Ivan Laptev", "Ruben Glatt", "Tal Ben-Nun", "Philip Torr", "Natasha Jaques", "Ameya Prabhu", "Brian Bartoldson", "Bhavya Kailkhura", "Christian Schroeder de Witt"], "categories": ["cs.LG", "cs.AI"], "fields_of_study": ["Computer Science"], "published_date": "2026-04-15", "url": "https://arxiv.org/abs/2604.14140", "pdf_url": "https://arxiv.org/pdf/2604.14140v1", "arxiv_id": "2604.14140", "doi": "10.48550/arXiv.2604.14140", "citation_count": 2, "influential_citation_count": 0, "has_code": false, "code_url": null, "venue": "arXiv.org", "quality_score": 0.5443}
{"id": "5fb881ea6d52354f9ce2cf8c3547a6711f4fc0ec9d9035ad07d3ee343c7e89c1", "sources": ["arxiv", "semantic_scholar"], "title": "Sample Complexity of Autoregressive Reasoning: Chain-of-Thought vs. End-to-End", "abstract": "Modern large language models generate text autoregressively, producing tokens one at a time. To study the learnability of such systems, Joshi et al. (COLT 2025) introduced a PAC-learning framework for next-token generators, the primitive underlying autoregressive models. In this framework, an unknown next-token generator maps a sequence of tokens to the next token and is iteratively applied for $T$ steps, producing a chain of tokens whose final token constitutes the model's output. The learning task is to learn the input-output mapping induced by this autoregressive process. Depending on the available supervision, training examples may reveal only the final output (End-to-End supervision) or the entire generated chain (Chain-of-Thought supervision). This raises two natural questions: how the sample complexity depends on the generation length $T$, and how much Chain-of-Thought supervision can reduce this dependence. In this work we give a nearly complete answer to both questions by uncovering a taxonomy of how the sample complexity scales with $T$. For End-to-End learning, we show that the landscape is remarkably rich: subject to mild conditions, essentially any growth rate $r(T)$ between constant and linear can arise as the sample complexity, and combined with the linear upper bound of Joshi et al., this yields an essentially complete characterization. In contrast, under Chain-of-Thought supervision we show that the sample complexity is independent of $T$, demonstrating that access to intermediate reasoning steps can eliminate the dependence on the generation length altogether. Our analysis introduces new combinatorial tools, and as corollaries we resolve several open questions posed by Joshi et al. regarding the dependence of learnability on the generation length and the role of Chain-of-Thought supervision.", "authors": ["Steve Hanneke", "Idan Mehalel", "Shay Moran"], "categories": ["cs.LG"], "fields_of_study": ["Computer Science"], "published_date": "2026-04-13", "url": "https://arxiv.org/abs/2604.12013", "pdf_url": "https://arxiv.org/pdf/2604.12013v2", "arxiv_id": "2604.12013", "doi": "10.48550/arXiv.2604.12013", "citation_count": 2, "influential_citation_count": 0, "has_code": false, "code_url": null, "venue": "arXiv.org", "quality_score": 0.542}
{"id": "83c76090f675531f1ca8e2afffed056d9104f10d7793d4f5f1412e4011be3638", "sources": ["arxiv", "semantic_scholar"], "title": "When More Thinking Hurts: Overthinking in LLM Test-Time Compute Scaling", "abstract": "Scaling test-time compute through extended chains of thought has become a dominant paradigm for improving large language model reasoning. However, existing research implicitly assumes that longer thinking always yields better results. This assumption remains largely unexamined. We systematically investigate how the marginal utility of additional reasoning tokens changes as compute budgets increase. We find that marginal returns diminish substantially at higher budgets and that models exhibit ``overthinking'', where extended reasoning is associated with abandoning previously correct answers. Furthermore, we show that optimal thinking length varies across problem difficulty, suggesting that uniform compute allocation is suboptimal. Our cost-aware evaluation framework reveals that stopping at moderate budgets can reduce computation significantly while maintaining comparable accuracy.", "authors": ["Shu Zhou", "Rui Ling", "Junan Chen", "Xin Wang", "Tao Fan", "Hao Wang"], "categories": ["cs.AI"], "fields_of_study": ["Computer Science"], "published_date": "2026-04-12", "url": "https://arxiv.org/abs/2604.10739", "pdf_url": "https://arxiv.org/pdf/2604.10739v1", "arxiv_id": "2604.10739", "doi": "10.48550/arXiv.2604.10739", "citation_count": 3, "influential_citation_count": 0, "has_code": false, "code_url": null, "venue": "arXiv.org", "quality_score": 0.5408}
{"id": "b84b934da1e313fd967f3021bad1618ab33678ce325e58b7fb7479d34090a657", "sources": ["arxiv", "semantic_scholar"], "title": "FACT-E: Causality-Inspired Evaluation for Trustworthy Chain-of-Thought Reasoning", "abstract": "Chain-of-Thought (CoT) prompting has improved LLM reasoning, but models often generate explanations that appear coherent while containing unfaithful intermediate steps. Existing self-evaluation approaches are prone to inherent biases: the model may confidently endorse coherence even when the step-to-step implication is not valid, leading to unreliable faithfulness evaluation. We propose FACT-E, a causality-inspired framework for evaluating CoT quality. FACT-E uses controlled perturbations as an instrumental signal to separate genuine step-to-step dependence from bias-driven artifacts, producing more reliable faithfulness estimates (\\textit{intra-chain faithfulness}). To select trustworthy trajectories, FACT-E jointly considers \\textit{intra-chain faithfulness} and \\textit{CoT-to-answer consistency}, ensuring that selected chains are both faithful internally and supportive of the correct final answer. Experiments on GSM8K, MATH, and CommonsenseQA show that FACT-E improves reasoning-trajectory selection and yields stronger in-context learning exemplars. FACT-E also reliably detects flawed reasoning under noisy conditions, providing a robust metric for trustworthy LLM reasoning.", "authors": ["Yuxi Sun", "Aoqi Zuo", "Haotian Xie", "Wei Gao", "Mingming Gong", "Jing Ma"], "categories": ["cs.AI"], "fields_of_study": ["Computer Science"], "published_date": "2026-04-12", "url": "https://arxiv.org/abs/2604.10693", "pdf_url": "https://arxiv.org/pdf/2604.10693v2", "arxiv_id": "2604.10693", "doi": "10.48550/arXiv.2604.10693", "citation_count": 5, "influential_citation_count": 0, "has_code": false, "code_url": null, "venue": "arXiv.org", "quality_score": 0.5408}
{"id": "2cc40113a36d0b7e8238c52a4895f8f64cd16a8ca76db21efb05af740b05db94", "sources": ["arxiv", "semantic_scholar"], "title": "Cognitive Loop of Thought: Reversible Hierarchical Markov Chain for Efficient Mathematical Reasoning", "abstract": "Multi-step Chain-of-Thought (CoT) has significantly advanced the mathematical reasoning capabilities of LLMs by leveraging explicit reasoning steps. However, the widespread adoption of Long CoT often results in sequence lengths that exceed manageable computational limits. While existing approaches attempt to alleviate this by reducing KV Cache redundancy via Markov chain-like structures, they introduce two critical limitations: inherent memorylessness (loss of context) and limited backward reasoning capability. To address these limitations, we propose a novel Chain-of-Thought framework based on Reversible Hierarchical Markov Chain, termed Cognitive Loop of Thought (CLoT), and a backward reasoning dataset CLoT-Instruct. In CLoT, problems are decomposed into sub-problems with hierarchical dependencies. Inspired by human cognitive processes, we introduce a backward verification mechanism at each hierarchical layer. Furthermore, we implement a pruning strategy: once higher-level sub-problems are verified, redundant lower-level sub-problems are pruned to maximize efficiency. This approach effectively mitigates error propagation and enhances reasoning robustness. Experiments on four mathematical benchmarks demonstrate the effectiveness of our method. Notably, on the AddSub dataset using GPT-4o-mini, CLoT achieves 99.0% accuracy, outperforming traditional CoT and CoT-SC by 4.1% and 2.9%, respectively.", "authors": ["Jia-Chen Zhang", "Yu-Jie Xiong", "Zheng Zhou"], "categories": ["cs.CL"], "fields_of_study": ["Computer Science"], "published_date": "2026-04-08", "url": "https://arxiv.org/abs/2604.06805", "pdf_url": "https://arxiv.org/pdf/2604.06805v2", "arxiv_id": "2604.06805", "doi": "10.48550/arXiv.2604.06805", "citation_count": 0, "influential_citation_count": 0, "has_code": false, "code_url": null, "venue": "arXiv.org", "quality_score": 0.5363}
{"id": "9962d7de9b2e43fa10f41b4d1fe11e55d64f1408698e58e54c18967d8871e3f2", "sources": ["arxiv", "semantic_scholar"], "title": "Long-Context Reasoning Through Proxy-Based Chain-of-Thought Tuning", "abstract": "Recent large language models support inputs of up to 10 million tokens, yet they perform poorly on long-context tasks that require complex reasoning. Such tasks can be solved using only a subset of the input -- a proxy context -- rather than the full sequence. Despite sharing the same underlying reasoning process, models exhibit a significant performance disparity between proxy and full contexts. To improve long-context reasoning, we propose ProxyCoT, a novel training framework that transfers reasoning capabilities from short proxy contexts to full long contexts. Specifically, we first obtain high-quality chain-of-thought reasoning traces on proxy contexts through reinforcement learning or distillation from a larger teacher model, and then ground the generated traces in full long contexts with supervised fine-tuning. Experiments across different datasets demonstrate that ProxyCoT consistently outperforms strong baselines with reduced computational overhead. Furthermore, models trained with ProxyCoT generalize their long-context reasoning capabilities to out-of-domain tasks.", "authors": ["Miao Li", "Irina Saparina", "Alexander Gurung", "Mirella Lapata"], "categories": ["cs.CL", "cs.AI", "cs.LG"], "fields_of_study": ["Computer Science"], "published_date": "2026-04-06", "url": "https://arxiv.org/abs/2605.20201", "pdf_url": "https://arxiv.org/pdf/2605.20201v2", "arxiv_id": "2605.20201", "doi": null, "citation_count": 0, "influential_citation_count": 0, "has_code": false, "code_url": null, "venue": null, "quality_score": 0.3398}
{"id": "921cd8f78d2d1f2caa4a42943d587ad0744bea130005c9e922e8c49d936c6492", "sources": ["arxiv", "semantic_scholar"], "title": "Hierarchical Chain-of-Thought Prompting: Enhancing LLM Reasoning Performance and Efficiency", "abstract": "Chain-of-Thought (CoT) prompting has significantly improved the reasoning capabilities of large language models (LLMs). However, conventional CoT often relies on unstructured, flat reasoning chains that suffer from redundancy and suboptimal performance. In this work, we introduce Hierarchical Chain-of-Thought (Hi-CoT) prompting, a structured reasoning paradigm specifically designed to address the challenges of complex, multi-step reasoning. Hi-CoT decomposes the reasoning process into hierarchical substeps by alternating between instructional planning and step-by-step execution. This decomposition enables LLMs to better manage long reasoning horizons and maintain logical coherence. Extensive evaluations across diverse LLMs and mathematical reasoning benchmarks show that Hi-CoT consistently improves average accuracy by 6.2% (up to 61.4% on certain models and tasks) while reducing reasoning trace length by 13.9% compared to CoT prompting. We further show that accuracy and efficiency are maximized when models strictly adhere to the hierarchical structure. Our code is available at https://github.com/XingshuaiHuang/Hi-CoT.", "authors": ["Xingshuai Huang", "Derek Li", "Bahareh Nikpour", "Parsa Omidi"], "categories": ["cs.CL"], "fields_of_study": ["Computer Science"], "published_date": "2026-03-31", "url": "https://arxiv.org/abs/2604.00130", "pdf_url": "https://arxiv.org/pdf/2604.00130v1", "arxiv_id": "2604.00130", "doi": "10.48550/arXiv.2604.00130", "citation_count": 0, "influential_citation_count": 0, "has_code": true, "code_url": "https://github.com/XingshuaiHuang/Hi-CoT", "venue": "arXiv.org", "quality_score": 0.8146}
{"id": "dafc9666d56152aeb3e4fc0b4c6255df90b34b13c0296dd50fcf4b10923f089c", "sources": ["arxiv", "semantic_scholar"], "title": "CoT2-Meta: Budgeted Metacognitive Control for Test-Time Reasoning", "abstract": "Recent test-time reasoning methods improve performance by generating more candidate chains or searching over larger reasoning trees, but they typically lack explicit control over when to expand, what to prune, how to repair, and when to abstain. We introduce CoT2-Meta, a training-free metacognitive reasoning framework that combines object-level chain-of-thought generation with meta-level control over partial reasoning trajectories. The framework integrates four components: strategy-conditioned thought generation, tree-structured search, an online process oracle for step-level reasoning evaluation, and a meta-controller that allocates computation through expansion, pruning, repair, stopping, and fallback decisions. Under matched inference budgets, CoT2-Meta consistently outperforms strong single-path, sampling-based, and search-based baselines, including ReST-MCTS. On the default backbone, it achieves 92.8 EM on MATH, 90.4 accuracy on GPQA, 98.65 EM on GSM8K, 75.8 accuracy on BBEH, 85.6 accuracy on MMMU-Pro, and 48.8 accuracy on HLE, with gains over the strongest non-CoT2-Meta baseline of +3.6, +5.2, +1.15, +2.0, +4.3, and +4.3 points, respectively. Beyond these core results, the framework remains effective across a broader 15-benchmark suite spanning knowledge and QA, multi-hop reasoning, coding, and out-of-distribution evaluation. Additional analyses show better compute scaling, improved calibration, stronger selective prediction, targeted repair behavior, and consistent gains across backbone families. These results suggest that explicit metacognitive control is a practical design principle for reliable and compute-efficient test-time reasoning systems.", "authors": ["Siyuan Ma", "Bo Gao", "Zikai Xiao", "Hailong Wang", "Xinlei Yu", "Rui Qian", "Jiayu Qian", "Luqi Gong", "Yang Liu"], "categories": ["cs.AI"], "fields_of_study": ["Computer Science"], "published_date": "2026-03-30", "url": "https://arxiv.org/abs/2603.28135", "pdf_url": "https://arxiv.org/pdf/2603.28135v1", "arxiv_id": "2603.28135", "doi": "10.48550/arXiv.2603.28135", "citation_count": 1, "influential_citation_count": 0, "has_code": false, "code_url": null, "venue": "arXiv.org", "quality_score": 0.5259}
{"id": "5b2f26528ee86c560c17f9acf5bac05e7029bb072ed334c5b1fb87398d5e9970", "sources": ["arxiv", "semantic_scholar"], "title": "Enhanced Mycelium of Thought (EMoT): A Bio-Inspired Hierarchical Reasoning Architecture with Strategic Dormancy and Mnemonic Encoding", "abstract": "Current prompting paradigms for large language models (LLMs), including Chain-of-Thought (CoT) and Tree-of-Thoughts (ToT), follow linear or tree-structured reasoning paths that lack persistent memory, strategic dormancy, and cross-domain synthesis. We present the Enhanced Mycelium of Thought (EMoT) framework, a bio-inspired reasoning architecture that organises cognitive processing into a four-level hierarchy (Micro, Meso, Macro, Meta), implements strategic dormancy and reactivation of reasoning nodes, and integrates a Memory Palace with five mnemonic encoding styles. EMoT is a research prototype for complex, multi-domain problems, not a general-purpose prompting enhancement. Two complementary evaluations reveal a characteristic trade-off. In a blind LLM-as-Judge evaluation across three domains, EMoT achieved near-parity with CoT (4.20 vs. 4.33/5.0) with higher stability, and outperformed CoT on Cross-Domain Synthesis (4.8 vs. 4.4). Ablation studies show that strategic dormancy is architecturally essential (quality collapsed from 4.2 to 1.0 when disabled). On a 15-item short-answer benchmark, EMoT (27%) substantially underperformed simpler baselines, confirming systematic overthinking on simple problems. These results are subject to important limitations: small sample sizes (n=3 complex cases, n=15 short-answer items), LLM-as-Judge evaluation with potential self-preference bias, and approximately 33-fold computational cost overhead. To our knowledge, EMoT is the first reasoning framework to combine hierarchical topology, strategic thought dormancy with reactivation, and mnemonic memory encoding in a single architecture.", "authors": ["Florian Odi Stummer"], "categories": ["cs.AI"], "fields_of_study": ["Computer Science"], "published_date": "2026-03-25", "url": "https://arxiv.org/abs/2603.24065", "pdf_url": "https://arxiv.org/pdf/2603.24065v1", "arxiv_id": "2603.24065", "doi": "10.48550/arXiv.2603.24065", "citation_count": 0, "influential_citation_count": 0, "has_code": false, "code_url": null, "venue": "arXiv.org", "quality_score": 0.5202}
{"id": "6bb2b49f4817b2a2f5b5b0f6edb669880e002e2b778352834f42a27eaae8ddfc", "sources": ["arxiv", "semantic_scholar"], "title": "Caterpillar of Thoughts: The Optimal Test-Time Algorithm for Large Language Models", "abstract": "Large language models (LLMs) can often produce substantially better outputs when allowed to use additional test-time computation, such as sampling, chain of thought, backtracking, or revising partial solutions. Despite the growing empirical success of such techniques, there is limited theoretical understanding of how inference time computation should be structured, or what constitutes an optimal use of a fixed computation budget. We model test-time computation as an algorithm interacting with a Markov chain: at any point, the algorithm may resume generation from any previously observed state. That is, unlike standard Markov chains where the states are drawn passively, we allow the algorithm to backtrack to any previously observed state of the Markov chain at any time. Many of the existing test-time algorithms, such as Chain-of-Thought (CoT) (Wei et al., 2023), Tree-of-Thoughts (ToT) (Yao et al., 2023), or Best-of-$k$ (Brown et al., 2024) could be seen as specific algorithms in this model. We prove that while backtracking can reduce the number of generations exponentially, a very limited form of backtracking is theoretically sufficient. Namely, we show that the optimal algorithm always generates a caterpillar tree. That is, if we remove the leaves of the state tree generated by the optimal algorithm, we obtain a path. Motivated by our characterization of the optimal algorithm, we present Caterpillar of Thoughts (CaT), a new test-time computation algorithm, reducing the number of token/state generations. Our empirical evaluation shows that CaT, compared to ToT, achieves a better success rate while also reducing the number of token generations.", "authors": ["Amir Azarmehr", "Soheil Behnezhad", "Alma Ghafari"], "categories": ["cs.LG"], "fields_of_study": ["Computer Science"], "published_date": "2026-03-24", "url": "https://arxiv.org/abs/2603.22784", "pdf_url": "https://arxiv.org/pdf/2603.22784v1", "arxiv_id": "2603.22784", "doi": "10.48550/arXiv.2603.22784", "citation_count": 0, "influential_citation_count": 0, "has_code": false, "code_url": null, "venue": "arXiv.org", "quality_score": 0.5191}
{"id": "8997397b632c2e87d7ca59600c8aed29fb7e8eec1add77c4b3118ff7c7517462", "sources": ["arxiv", "semantic_scholar"], "title": "Let's Think with Images Efficiently! An Interleaved-Modal Chain-of-Thought Reasoning Framework with Dynamic and Precise Visual Thoughts", "abstract": "Recently, Interleaved-modal Chain-of-Thought (ICoT) reasoning has achieved remarkable success by leveraging both multimodal inputs and outputs, attracting increasing attention. While achieving promising performance, current ICoT methods still suffer from two major limitations: (1) Static Visual Thought Positioning, which statically inserts visual information at fixed steps, resulting in inefficient and inflexible reasoning; and (2) Broken Visual Thought Representation, which involves discontinuous and semantically incoherent visual tokens. To address these limitations, we introduce Interleaved-modal Chain-of-Thought reasoning with Dynamic and Precise Visual Thoughts (DaP-ICoT), which incorporates two key components: (1) Dynamic Visual Thought Integration adaptively introduces visual inputs based on reasoning needs, reducing redundancy and improving efficiency. (2) Precise Visual Thought Guidance ensures visual semantically coherent and contextually aligned representations. Experiments across multiple benchmarks and models demonstrate that DaP-ICoT achieves state-of-the-art performance. In addition, DaP-ICoT significantly reduces the number of inserted images, leading to a 72.6% decrease in token consumption, enabling more efficient ICoT reasoning.", "authors": ["Xu Liu", "Yongheng Zhang", "Qiguang Chen", "Yao Li", "Sheng Wang", "Libo Qin"], "categories": ["cs.CV", "cs.AI"], "fields_of_study": ["Computer Science"], "published_date": "2026-03-23", "url": "https://arxiv.org/abs/2603.21754", "pdf_url": "https://arxiv.org/pdf/2603.21754v1", "arxiv_id": "2603.21754", "doi": "10.1609/aaai.v40i38.40494", "citation_count": 0, "influential_citation_count": 0, "has_code": false, "code_url": null, "venue": "AAAI Conference on Artificial Intelligence", "quality_score": 0.5179}
{"id": "76d461eda049cd37ce0a929b31d3d0ebb06afefe939bfe3c11ed475131f1dab4", "sources": ["arxiv", "semantic_scholar"], "title": "Do Multilingual VLMs Reason Equally? A Cross-Lingual Visual Reasoning Audit for Indian Languages", "abstract": "Vision-language models score well on mathematical, scientific, and spatial reasoning benchmarks, yet these evaluations are overwhelmingly English. I present the first cross-lingual visual reasoning audit for Indian languages. 980 questions from MathVista, ScienceQA, and MMMU are translated into Hindi, Tamil, Telugu, Bengali, Kannada, and Marathi using IndicTrans2, with Gemini 2.0 Flash cross-verification on 50 samples per language (inter-translator agreement 0.79-0.84). Eight VLMs, from 7B open-source models to GPT-4o, are evaluated across all seven languages, yielding 68,600 inference records that include text-only and chain-of-thought ablations. I find accuracy drops of 9.8-25 percentage points when switching from English to an Indian language, with Dravidian languages suffering up to 13.2 pp more than Indo-Aryan. Chain-of-thought prompting degrades Bengali (-14.4 pp) and Kannada (-11.4 pp) rather than helping, exposing English-centric reasoning chains. Aya-Vision-8B, built for 23 languages, still drops 28.5 pp on Dravidian scripts; multilingual pretraining alone does not transfer visual reasoning. I release the translated benchmark and all model outputs.", "authors": ["Swastik R"], "categories": ["cs.CL", "cs.LG"], "fields_of_study": ["Computer Science"], "published_date": "2026-03-23", "url": "https://arxiv.org/abs/2603.26742", "pdf_url": "https://arxiv.org/pdf/2603.26742v1", "arxiv_id": "2603.26742", "doi": "10.48550/arXiv.2603.26742", "citation_count": 0, "influential_citation_count": 0, "has_code": true, "code_url": "https://github.com/QuantumByte-01/multilingual-vlm-reasoning-audit", "venue": "arXiv.org", "quality_score": 0.8004}
{"id": "fe19ca7e418770e0556aaceb5ee7db49a774b7b811556a6ce56b4bc003ebffdf", "sources": ["arxiv", "semantic_scholar"], "title": "Lie to Me: How Faithful Is Chain-of-Thought Reasoning in Reasoning Models?", "abstract": "Chain-of-thought (CoT) reasoning has been proposed as a transparency mechanism for large language models in safety-critical deployments, yet its effectiveness depends on faithfulness (whether models accurately verbalize the factors that actually influence their outputs), a property that prior evaluations have examined in only two proprietary models, finding acknowledgment rates as low as 25% for Claude 3.7 Sonnet and 39% for DeepSeek-R1. To extend this evaluation across the open-weight ecosystem, this study tests 12 open-weight reasoning models spanning 9 architectural families (7B-685B parameters) on 498 multiple-choice questions from MMLU and GPQA Diamond, injecting six categories of reasoning hints (sycophancy, consistency, visual pattern, metadata, grader hacking, and unethical information) and measuring the rate at which models acknowledge hint influence in their CoT when hints successfully alter answers. Across 41,832 inference runs, overall faithfulness rates range from 39.7% (Seed-1.6-Flash) to 89.9% (DeepSeek-V3.2-Speciale) across model families, with consistency hints (35.5%) and sycophancy hints (53.9%) exhibiting the lowest acknowledgment rates. Training methodology and model family predict faithfulness more strongly than parameter count, and keyword-based analysis reveals a striking gap between thinking-token acknowledgment (approximately 87.5%) and answer-text acknowledgment (approximately 28.6%), suggesting that models internally recognize hint influence but systematically suppress this acknowledgment in their outputs. These findings carry direct implications for the viability of CoT monitoring as a safety mechanism and suggest that faithfulness is not a fixed property of reasoning models but varies systematically with architecture, training method, and the nature of the influencing cue.", "authors": ["Richard J. Young"], "categories": ["cs.CL", "cs.AI"], "fields_of_study": ["Computer Science"], "published_date": "2026-03-23", "url": "https://arxiv.org/abs/2603.22582", "pdf_url": "https://arxiv.org/pdf/2603.22582v1", "arxiv_id": "2603.22582", "doi": "10.48550/arXiv.2603.22582", "citation_count": 3, "influential_citation_count": 1, "has_code": false, "code_url": null, "venue": "arXiv.org", "quality_score": 0.5179}
{"id": "cd592e61be487a7e0f1e1e51bc2368b566fe4faae9f543d0e581a5ca9a18bcc0", "sources": ["arxiv", "semantic_scholar"], "title": "When the Chain Breaks: Interactive Diagnosis of LLM Chain-of-Thought Reasoning Errors", "abstract": "Current Large Language Models (LLMs), especially Large Reasoning Models, can generate Chain-of-Thought (CoT) reasoning traces to illustrate how they produce final outputs, thereby facilitating trust calibration for users. However, these CoT reasoning traces are usually lengthy and tedious, and can contain various issues, such as logical and factual errors, which make it difficult for users to interpret the reasoning traces efficiently and accurately. To address these challenges, we develop an error detection pipeline that combines external fact-checking with symbolic formal logical validation to identify errors at the step level. Building on this pipeline, we propose ReasonDiag, an interactive visualization system for diagnosing CoT reasoning traces. ReasonDiag provides 1) an integrated arc diagram to show reasoning-step distributions and error-propagation patterns, and 2) a hierarchical node-link diagram to visualize high-level reasoning flows and premise dependencies. We evaluate ReasonDiag through a technical evaluation for the error detection pipeline, two case studies, and user interviews with 16 participants. The results indicate that ReasonDiag helps users effectively understand CoT reasoning traces, identify erroneous steps, and determine their root causes.", "authors": ["Shiwei Chen", "Niruthikka Sritharan", "Xiaolin Wen", "Chenxi Zhang", "Xingbo Wang", "Yong Wang"], "categories": ["cs.HC"], "fields_of_study": ["Computer Science"], "published_date": "2026-03-22", "url": "https://arxiv.org/abs/2603.21286", "pdf_url": "https://arxiv.org/pdf/2603.21286v2", "arxiv_id": "2603.21286", "doi": "10.1111/cgf.70439", "citation_count": 0, "influential_citation_count": 0, "has_code": false, "code_url": null, "venue": null, "quality_score": 0.3289}
{"id": "4751df33b16098708d947c17f4f9e8f8568be3b205212ae0f97bc77e2eeb3591", "sources": ["arxiv", "semantic_scholar"], "title": "Enhancing reasoning accuracy in large language models during inference time", "abstract": "Large Language Models (LLMs) often exhibit strong linguistic abilities while remaining unreliable on multi-step reasoning tasks, particularly when deployed without additional training or fine-tuning. In this work, we study inference-time techniques to improve the reasoning accuracy of LLMs. We systematically evaluate three classes of inference-time strategies: (i) self-consistency via stochastic decoding, where the model is sampled multiple times using controlled temperature and nucleus sampling and the most frequent final answer is selected; (ii) dual-model reasoning agreement, where outputs from two independent models are compared and only consistent reasoning traces are trusted; and (iii) self-reflection, where the model critiques and revises its own reasoning. Across all evaluated methods, we employ Chain-of-Thought (CoT) [1] prompting to elicit explicit intermediate reasoning steps before generating final answers. In this work, we provide a controlled comparative evaluation across three inference-time strategies under identical prompting and verification settings. Our experiments on LLM [2] show that self-consistency with nucleus sampling and controlled temperature value yields the substantial gains, achieving a 9% to 15% absolute improvement in accuracy over greedy single-pass decoding, well-suited for low-risk domains, offering meaningful gains with minimal overhead. The dual-model approach provides additional confirmation for model reasoning steps thus more appropriate for moderate-risk domains, where higher reliability justifies additional compute. Self-reflection offers only marginal improvements, suggesting limited effectiveness for smaller non-reasoning models at inference time.", "authors": ["Vinay Sharma", "Manish Jain"], "categories": ["cs.CL", "cs.AI"], "fields_of_study": ["Computer Science"], "published_date": "2026-03-22", "url": "https://arxiv.org/abs/2603.21301", "pdf_url": "https://arxiv.org/pdf/2603.21301v1", "arxiv_id": "2603.21301", "doi": "10.48550/arXiv.2603.21301", "citation_count": 0, "influential_citation_count": 0, "has_code": false, "code_url": null, "venue": "arXiv.org", "quality_score": 0.5168}
{"id": "d8534aa131519bc69f9d206f8e81ae704e15d57b67cf532e68cc75397bfd3658", "sources": ["arxiv", "semantic_scholar"], "title": "Correct Chains, Wrong Answers: Dissociating Reasoning from Output in LLM Logic", "abstract": "LLMs can execute every step of chain-of-thought reasoning correctly and still produce wrong final answers. We introduce the Novel Operator Test, a benchmark that separates operator logic from operator name, enabling rigorous distinction between genuine reasoning and pattern retrieval. By evaluating Boolean operators under unfamiliar names across depths 1-10 on five models (up to 8,100 problems each), we demonstrate a reasoning-output dissociation that existing benchmarks cannot detect. At Claude Sonnet 4's depth 7, all 31 errors have verifiably correct reasoning yet wrong declared answers; 17/19 errors in mixed-operator chains exhibit the same pattern. The benchmark reveals two failure types: strategy failures at depth 2, where models attempt terse retrieval (+62pp from scaffolding), and content failures at depth 7, where models reason fully but err systematically (+8-30pp, 0/300 errors post-intervention). A Trojan operator (XOR's truth table under a novel name) confirms name alone does not gate reasoning (p >= 0.49), while Llama's novelty gap widens to 28pp at depth 8-9 with the Trojan at 92-100%, isolating genuine difficulty with novel logic from name unfamiliarity.", "authors": ["Abinav Rao", "Sujan Rachuri", "Nikhil Vemuri"], "categories": ["cs.CL", "cs.AI", "cs.LO"], "fields_of_study": ["Computer Science"], "published_date": "2026-03-19", "url": "https://arxiv.org/abs/2604.13065", "pdf_url": "https://arxiv.org/pdf/2604.13065v1", "arxiv_id": "2604.13065", "doi": "10.48550/arXiv.2604.13065", "citation_count": 0, "influential_citation_count": 0, "has_code": false, "code_url": null, "venue": "arXiv.org", "quality_score": 0.5133}
{"id": "75f050caf8dc2dffa9a844077350e203896315c238e553187f1c5517c57f2aba", "sources": ["arxiv", "semantic_scholar"], "title": "Efficient Reasoning on the Edge", "abstract": "Large language models (LLMs) with chain-of-thought reasoning achieve state-of-the-art performance across complex problem-solving tasks, but their verbose reasoning traces and large context requirements make them impractical for edge deployment. These challenges include high token generation costs, large KV-cache footprints, and inefficiencies when distilling reasoning capabilities into smaller models for mobile devices. Existing approaches often rely on distilling reasoning traces from larger models into smaller models, which are verbose and stylistically redundant, undesirable for on-device inference. In this work, we propose a lightweight approach to enable reasoning in small LLMs using LoRA adapters combined with supervised fine-tuning. We further introduce budget forcing via reinforcement learning on these adapters, significantly reducing response length with minimal accuracy loss. To address memory-bound decoding, we exploit parallel test-time scaling, improving accuracy at minor latency increase. Finally, we present a dynamic adapter-switching mechanism that activates reasoning only when needed and a KV-cache sharing strategy during prompt encoding, reducing time-to-first-token for on-device inference. Experiments on Qwen2.5-7B demonstrate that our method achieves efficient, accurate reasoning under strict resource constraints, making LLM reasoning practical for mobile scenarios. Videos demonstrating our solution running on mobile devices are available on our project page.", "authors": ["Yelysei Bondarenko", "Thomas Hehn", "Rob Hesselink", "Romain Lepert", "Fabio Valerio Massoli", "Evgeny Mironov", "Leyla Mirvakhabova", "Tribhuvanesh Orekondy", "Spyridon Stasis", "Andrey Kuzmin", "Anna Kuzina", "Markus Nagel", "Ankita Nayak", "Corrado Rainone", "Ork de Rooij", "Paul N Whatmough", "Arash Behboodi", "Babak Ehteshami Bejnordi"], "categories": ["cs.LG", "cs.CL"], "fields_of_study": ["Computer Science"], "published_date": "2026-03-17", "url": "https://arxiv.org/abs/2603.16867", "pdf_url": "https://arxiv.org/pdf/2603.16867v2", "arxiv_id": "2603.16867", "doi": "10.48550/arXiv.2603.16867", "citation_count": 0, "influential_citation_count": 0, "has_code": false, "code_url": null, "venue": "arXiv.org", "quality_score": 0.511}
{"id": "3459d17e911f1be1050b4a7c07782558478a8c1c8da020bf0c3d194606986d20", "sources": ["arxiv", "semantic_scholar"], "title": "VLA-Thinker: Boosting Vision-Language-Action Models through Thinking-with-Image Reasoning", "abstract": "Vision-Language-Action (VLA) models have shown promising capabilities for embodied intelligence, but most existing approaches rely on text-based chain-of-thought reasoning where visual inputs are treated as static context. This limits the ability of the model to actively revisit the environment and resolve ambiguities during long-horizon tasks. We propose VLA-Thinker, a thinking-with-image reasoning framework that models perception as a dynamically invocable reasoning action. To train such a system, we introduce a two-stage training pipeline consisting of (1) an SFT cold-start phase with curated visual Chain-of-Thought data to activate structured reasoning and tool-use behaviors, and (2) GRPO-based reinforcement learning to align complete reasoning-action trajectories with task-level success. Extensive experiments on LIBERO and RoboTwin 2.0 benchmarks demonstrate that VLA-Thinker significantly improves manipulation performance, achieving 97.5% success rate on LIBERO and strong gains across long-horizon robotic tasks. Project and Codes: https://cywang735.github.io/VLA-Thinker/ .", "authors": ["Chaoyang Wang", "Wenrui Bao", "Sicheng Gao", "Bingxin Xu", "Yu Tian", "Yogesh S. Rawat", "Yunhao Ge", "Yuzhang Shang"], "categories": ["cs.CV", "cs.AI", "cs.RO"], "fields_of_study": ["Computer Science"], "published_date": "2026-03-15", "url": "https://arxiv.org/abs/2603.14523", "pdf_url": "https://arxiv.org/pdf/2603.14523v1", "arxiv_id": "2603.14523", "doi": "10.48550/arXiv.2603.14523", "citation_count": 1, "influential_citation_count": 0, "has_code": false, "code_url": null, "venue": "arXiv.org", "quality_score": 0.5088}
{"id": "2a779685eddfc77bc5628e0ecd43674294b7d10ca141b335ede0716dfd32c45f", "sources": ["arxiv", "semantic_scholar"], "title": "TERMINATOR: Learning Optimal Exit Points for Early Stopping in Chain-of-Thought Reasoning", "abstract": "Large Reasoning Models (LRMs) achieve impressive performance on complex reasoning tasks via Chain-of-Thought (CoT) reasoning, which enables them to generate intermediate thinking tokens before arriving at the final answer. However, LRMs often suffer from significant overthinking, spending excessive compute time even after the answer is generated early on. Prior work has identified the existence of an optimal reasoning length such that truncating reasoning at this point significantly shortens CoT outputs with virtually no change in performance. However, determining optimal CoT lengths for practical datasets is highly non-trivial as they are fully task and model-dependent. In this paper, we precisely address this and design Terminator, an early-exit strategy for LRMs at inference to mitigate overthinking. The central idea underpinning Terminator is that the first arrival of an LRM's final answer is often predictable, and we leverage these first answer positions to create a novel dataset of optimal reasoning lengths to train Terminator. Powered by this approach, Terminator achieves significant reductions in CoT lengths of 14%-55% on average across four challenging practical datasets: MATH-500, AIME 2025, HumanEval, and GPQA, while outperforming current state-of-the-art methods and reducing inference latency by more than 2x compared to the original LRM.", "authors": ["Alliot Nagle", "Jakhongir Saydaliev", "Dhia Garbaya", "Michael Gastpar", "Ashok Vardhan Makkuva", "Hyeji Kim"], "categories": ["cs.LG", "cs.AI", "cs.CL"], "fields_of_study": ["Computer Science"], "published_date": "2026-03-13", "url": "https://arxiv.org/abs/2603.12529", "pdf_url": "https://arxiv.org/pdf/2603.12529v2", "arxiv_id": "2603.12529", "doi": "10.48550/arXiv.2603.12529", "citation_count": 2, "influential_citation_count": 0, "has_code": false, "code_url": null, "venue": "arXiv.org", "quality_score": 0.5065}
{"id": "0eea66db0c5042eb93effe084bc91726263a36248ab125bfb27a0aa1a5cf5b5b", "sources": ["arxiv", "semantic_scholar"], "title": "EndoCoT: Scaling Endogenous Chain-of-Thought Reasoning in Diffusion Models", "abstract": "Recently, Multimodal Large Language Models (MLLMs) have been widely integrated into diffusion frameworks primarily as text encoders to tackle complex tasks such as spatial reasoning. However, this paradigm suffers from two critical limitations: (i) MLLMs text encoder exhibits insufficient reasoning depth. Single-step encoding fails to activate the Chain-of-Thought process, which is essential for MLLMs to provide accurate guidance for complex tasks. (ii) The guidance remains invariant during the decoding process. Invariant guidance during decoding prevents DiT from progressively decomposing complex instructions into actionable denoising steps, even with correct MLLM encodings. To this end, we propose Endogenous Chain-of-Thought (EndoCoT), a novel framework that first activates MLLMs' reasoning potential by iteratively refining latent thought states through an iterative thought guidance module, and then bridges these states to the DiT's denoising process. Second, a terminal thought grounding module is applied to ensure the reasoning trajectory remains grounded in textual supervision by aligning the final state with ground-truth answers. With these two components, the MLLM text encoder delivers meticulously reasoned guidance, enabling the DiT to execute it progressively and ultimately solve complex tasks in a step-by-step manner. Extensive evaluations across diverse benchmarks (e.g., Maze, TSP, VSP, and Sudoku) achieve an average accuracy of 92.1%, outperforming the strongest baseline by 8.3 percentage points. The code and dataset are publicly available at https://internlm.github.io/EndoCoT/.", "authors": ["Xuanlang Dai", "Yujie Zhou", "Long Xing", "Jiazi Bu", "Xilin Wei", "Yuhong Liu", "Beichen Zhang", "Kai Chen", "Yuhang Zang"], "categories": ["cs.CV", "cs.CL"], "fields_of_study": ["Computer Science"], "published_date": "2026-03-12", "url": "https://arxiv.org/abs/2603.12252", "pdf_url": "https://arxiv.org/pdf/2603.12252v4", "arxiv_id": "2603.12252", "doi": "10.48550/arXiv.2603.12252", "citation_count": 1, "influential_citation_count": 0, "has_code": false, "code_url": null, "venue": "arXiv.org", "quality_score": 0.5053}
{"id": "dd8fecc43a6088400ef2697c117ad0fa02bfeb6ba435b09ca84707997e16ad5c", "sources": ["arxiv", "semantic_scholar"], "title": "TopoBench: Benchmarking LLMs on Hard Topological Reasoning", "abstract": "Solving topological grid puzzles requires reasoning over global spatial invariants such as connectivity, loop closure, and region symmetry and remains challenging for even the most powerful large language models (LLMs). To study these abilities under controlled settings, we introduce TopoBench, a benchmark of six puzzle families across three difficulty levels. We evaluate strong reasoning LLMs on TopoBench and find that even frontier models solve fewer than one quarter of hard instances, with two families nearly unsolved. To investigate whether these failures stem from reasoning limitations or from difficulty extracting and maintaining spatial constraints, we annotate 750 chain of thought traces with an error taxonomy that surfaces four candidate causal failure modes, then test them with targeted interventions simulating each error type. These interventions show that certain error patterns like premature commitment and constraint forgetting have a direct impact on the ability to solve the puzzle, while repeated reasoning is a benign effect of search. Finally we study mitigation strategies including prompt guidance, cell-aligned grid representations and tool-based constraint checking, finding that the bottleneck lies in extracting constraints from spatial representations and not in reasoning over them. Code and data are available at github.com/mayug/topobench-benchmark.", "authors": ["Mayug Maniparambil", "Nils Hoehing", "Janak Kapuriya", "Arjun Karuvally", "Ellen Rushe", "Anthony Ventresque", "Noel O'Connor", "Fergal Reid"], "categories": ["cs.AI", "cs.CL"], "fields_of_study": ["Computer Science"], "published_date": "2026-03-12", "url": "https://arxiv.org/abs/2603.12133", "pdf_url": "https://arxiv.org/pdf/2603.12133v1", "arxiv_id": "2603.12133", "doi": "10.48550/arXiv.2603.12133", "citation_count": 1, "influential_citation_count": 0, "has_code": true, "code_url": null, "venue": "arXiv.org", "quality_score": 0.7809}
{"id": "521cd95b0c0f6c04a3427a917b0ae147f042149411f57d15c421a91260a55dd1", "sources": ["arxiv", "semantic_scholar"], "title": "Ranking Reasoning LLMs under Test-Time Scaling", "abstract": "Test-time scaling evaluates reasoning LLMs by sampling multiple outputs per prompt, but ranking models in this regime remains underexplored. We formalize dense benchmark ranking under test-time scaling and introduce Scorio, a library that implements statistical ranking methods such as paired-comparison models, item response theory (IRT) models, voting rules, and graph- and spectral-based methods. Across $20$ reasoning models on four Olympiad-style math benchmarks (AIME'24, AIME'25, HMMT'25, and BrUMO'25; up to $N=80$ trials), most full-trial rankings agree closely with the Bayesian gold standard $\\mathrm{Bayes}_{\\mathcal{U}}@80$ (mean Kendall's $τ_b = 0.93$--$0.95$), and $19$--$34$ methods recover exactly the same ordering. In the single-trial regime, the best methods reach $τ_b \\approx 0.86$. Using greedy decoding as an empirical prior ($\\mathrm{Bayes}_{\\mathbf{R}_0}@N$) reduces variance at $N=1$ by $16$--$52\\%$, but can bias rankings when greedy and stochastic sampling disagree. These results identify reliable ranking methods for both high- and low-budget test-time scaling. We release Scorio as an open-source library at https://github.com/mohsenhariri/scorio.", "authors": ["Mohsen Hariri", "Michael Hinczewski", "Jing Ma", "Vipin Chaudhary"], "categories": ["cs.LG", "math.ST"], "fields_of_study": ["Computer Science", "Mathematics"], "published_date": "2026-03-11", "url": "https://arxiv.org/abs/2603.10960", "pdf_url": "https://arxiv.org/pdf/2603.10960v1", "arxiv_id": "2603.10960", "doi": "10.48550/arXiv.2603.10960", "citation_count": 2, "influential_citation_count": 0, "has_code": true, "code_url": "https://github.com/mohsenhariri/scorio", "venue": "arXiv.org", "quality_score": 0.7792}
{"id": "f7d8989d6f866213a322f399de3956ae349c8bd1d3536e396475c7002e154ebc", "sources": ["arxiv", "semantic_scholar"], "title": "Efficient Reasoning at Fixed Test-Time Cost via Length-Aware Attention Priors and Gain-Aware Training", "abstract": "We study efficient reasoning under tight compute. We ask how to make structured, correct decisions without increasing test time cost. We add two training only components to small and medium Transformers that also transfer to broader differentiable optimizers. First, a length aware attention prior built via fuzzy regime position alignment, RPA, yields a normalized pre softmax bias that guides attention like a structured regularizer while adding no new inference parameters. Second, a minimal gain aware controller, Guardian, nudges attention sharpness only when validation improvements warrant it, following a two timescale policy gradient view of nonconvex optimization. It is disabled at inference. A KL perspective shows softmax of z plus log pi as MAP with KL regularization, grounding the prior in a principled objective. Under strict compute parity on WikiText 2, we reduce validation cross entropy while matching baseline latency and memory. At inference, we add a precomputed, cached prior B of T as a single additive bias per head. The controller does not run. In practice, this incurs negligible overhead, a cached bias add per head, with no measurable p50 latency shift. Our results suggest that length aware priors and late phase gain control preserve scarce improvements, especially in long span, noisy logit regimes, while keeping test time costs effectively unchanged.", "authors": ["Rian Atri"], "categories": ["cs.LG"], "fields_of_study": ["Computer Science"], "published_date": "2026-03-10", "url": "https://arxiv.org/abs/2603.09253", "pdf_url": "https://arxiv.org/pdf/2603.09253v1", "arxiv_id": "2603.09253", "doi": "10.48550/arXiv.2603.09253", "citation_count": 0, "influential_citation_count": 0, "has_code": false, "code_url": null, "venue": "arXiv.org", "quality_score": 0.503}
{"id": "d5600d24a5a80e3123e3fc8d53039758b08eaf99e98c87f678c20fdda4fe6d5e", "sources": ["arxiv", "semantic_scholar"], "title": "Quantifying the Necessity of Chain of Thought through Opaque Serial Depth", "abstract": "Large language models (LLMs) tend to externalize their reasoning in their chain of thought, making the chain of thought a good target for monitoring. This is partially an inherent feature of the Transformer architecture: sufficiently long serial cognition must pass through the chain of thought (Korbak et al., 2025). We formalize this argument through the notion of opaque serial depth, given by the length of the longest computation that can be done without the use of interpretable intermediate steps like chain of thought. Given this formalization, we compute numeric upper bounds on the opaque serial depth of Gemma 3 models, as well as asymptotic results for additional architectures beyond standard LLMs. We also open-source an automated method that can calculate upper bounds on the opaque serial depth of arbitrary neural networks, and use it to demonstrate that Mixture-of-Experts models likely have lower depth than dense models. Overall, our results suggest that opaque serial depth is a useful tool for understanding the potential for models to do significant reasoning that is not externalized.", "authors": ["Jonah Brown-Cohen", "David Lindner", "Rohin Shah"], "categories": ["cs.AI"], "fields_of_study": ["Computer Science"], "published_date": "2026-03-10", "url": "https://arxiv.org/abs/2603.09786", "pdf_url": "https://arxiv.org/pdf/2603.09786v1", "arxiv_id": "2603.09786", "doi": "10.48550/arXiv.2603.09786", "citation_count": 0, "influential_citation_count": 0, "has_code": true, "code_url": null, "venue": "arXiv.org", "quality_score": 0.7774}
{"id": "d1dbff3daecd7a57aabe81629b6abdd81c2d7d049a0f6b051f4a725bd9bf2443", "sources": ["arxiv", "semantic_scholar"], "title": "Is continuous CoT better suited for multi-lingual reasoning?", "abstract": "We investigate whether performing reasoning in a continuous latent space leads to more robust multilingual capabilities. We compare Continuous Chain-of-Thought (using the CODI framework) against standard supervised fine-tuning across five typologically diverse languages: English, Chinese, German, French, and Urdu. Our experiments on GSM8k and CommonsenseQA demonstrate that continuous reasoning significantly outperforms explicit reasoning on low-resource languages, particularly in zero-shot settings where the target language was not seen during training. Additionally, this approach achieves extreme efficiency, compressing reasoning traces by approximately $29\\times$ to $50\\times$. These findings indicate that continuous latent representations naturally exhibit greater language invariance, offering a scalable solution for cross-lingual reasoning.", "authors": ["Ali Hamza Bashir", "Behzad Shomali", "Markus Frey", "Mehdi Ali", "Rafet Sifa", "David Berghaus"], "categories": ["cs.CL", "cs.AI", "cs.LG"], "fields_of_study": ["Computer Science"], "published_date": "2026-03-09", "url": "https://arxiv.org/abs/2603.08177", "pdf_url": "https://arxiv.org/pdf/2603.08177v1", "arxiv_id": "2603.08177", "doi": "10.48550/arXiv.2603.08177", "citation_count": 0, "influential_citation_count": 0, "has_code": false, "code_url": null, "venue": "arXiv.org", "quality_score": 0.5019}
{"id": "2d71a20be6a548d054b97bf968e779800047b559e3812215bcd6ba36bad55a01", "sources": ["arxiv", "semantic_scholar"], "title": "Learning When to Sample: Confidence-Aware Selective Sampling for Efficient Chain-of-Thought Reasoning", "abstract": "Large language models (LLMs) can achieve strong reasoning performance through chain-of-thought (CoT) reasoning, yet they often generate unnecessarily long reasoning paths that incur high inference cost. Self-consistency-based approaches push accuracy higher still, but they require sampling and aggregating multiple reasoning trajectories, leading to substantial computational overhead. In this paper, we introduce a confidence-aware selective sampling framework that, at inference time, analyzes a single reasoning trajectory to adaptively determine whether to rely on that trajectory alone or trigger multi-path sampling. The framework uses trajectory-level numeric features and sentence-level linguistic features extracted from reasoning states to guide selective multi-path reasoning. We train it on MedQA and evaluate it in-domain on MedQA and under calibration-only transfer on MathQA, MedMCQA, and MMLU, without further fine-tuning. Experimental results show that the proposed framework maintains comparable performance to full and efficient multi-path reasoning baselines, with accuracy changes of $-0.41 \\pm 0.58$ and $-0.31 \\pm 0.58$ percentage points, respectively, while reducing token usage by $71.7 \\pm 5.0%$ and $36.6 \\pm 9.1%$. These findings demonstrate that reasoning trajectories contain rich signals for uncertainty estimation, enabling a simple, transferable mechanism to balance accuracy and efficiency in LLM reasoning.", "authors": ["Juming Xiong", "Kevin Guo", "Congning Ni", "Wexin Liu", "Chao Yan", "Katherine Brown", "Avinash Baidya", "Xiang Gao", "Bradley Malin", "Zhijun Yin"], "categories": ["cs.CL"], "fields_of_study": ["Computer Science"], "published_date": "2026-03-09", "url": "https://arxiv.org/abs/2603.08999", "pdf_url": "https://arxiv.org/pdf/2603.08999v3", "arxiv_id": "2603.08999", "doi": null, "citation_count": 0, "influential_citation_count": 0, "has_code": false, "code_url": null, "venue": null, "quality_score": 0.3194}
{"id": "567b5ebc84843a54c715b9ff51e12b5af7e269e2dd68501c32e166c7202ee796", "sources": ["arxiv", "semantic_scholar"], "title": "FreeFly-Thinking : Aligning Chain-of-Thought Reasoning with Continuous UAV Navigation", "abstract": "Vision-Language Navigation aims to enable agents to understand natural language instructions and carry out appropriate navigation actions in real-world environments. Most work focuses on indoor settings, with little research in complex outdoor scenes. Current UAV Vision-and-Language Navigation models typically act as black boxes without explicit reasoning. We introduce FreeFly-thinking, an end-to-end VLN framework that converts the UAV agent's egocentric images and language instructions into a series of actions, inspired by environment of urban architecture proposed by OpenFly. We first construct a UAV dataset for navigation task, and then performing natural language chain of thought. We adopt a two-stage training strategy: Supervised fine-tuning and Reinforcement fine-tuning. Experiments on unseen test demonstrate a strong performance, presenting robustness and efficiency in UAV navigation issue.", "authors": ["Jiaxu Zhou", "Shaobo Wang", "Zhiyuan Yang", "Zhenjun Yu", "Tao Li"], "categories": ["cs.CV"], "fields_of_study": ["Computer Science"], "published_date": "2026-03-07", "url": "https://arxiv.org/abs/2603.07181", "pdf_url": "https://arxiv.org/pdf/2603.07181v2", "arxiv_id": "2603.07181", "doi": "10.48550/arXiv.2603.07181", "citation_count": 0, "influential_citation_count": 0, "has_code": false, "code_url": null, "venue": "arXiv.org", "quality_score": 0.4996}
{"id": "ce775f57bdab64bbaa6086fcf405d2ec2ec02a278cad89d9bd8bd54f2895d9ab", "sources": ["arxiv", "semantic_scholar"], "title": "Improving reasoning at inference time via uncertainty minimisation", "abstract": "Large language models (LLMs) now exhibit strong multi-step reasoning abilities, but existing inference-time scaling methods remain computationally expensive, often relying on extensive sampling or external evaluators. We propose a principled strategy that frames reasoning as uncertainty minimisation and operates at the level of individual thoughts rather than tokens. Our method selects, at each reasoning step, the continuation that maximizes the model's self-certainty, a metric computed from its internal predictive distribution. This approach achieves significant improvement with a small number of samples, relies exclusively on model-internal signals, and applies to open-ended questions as opposed to methods like majority voting. Experiments on MATH500 and GSM8K across multiple model sizes demonstrate that thought-level self-certainty maximization consistently outperforms greedy decoding and matches or exceeds self-consistency under comparable token budgets. Cross-linguistic evaluations further indicate that the method transfers robustly beyond high-resource languages. Furthermore, analysis of self-certainty dynamics reveals that correct reasoning trajectories converge early to stable paths, suggesting that early decisions, likely associated with the planning of the reasoning process, are predictive of final accuracy. Building on this result, we show that self-certainty maximisation applied to the early steps can explain most of the performance gain and provide a simple yet efficient inference-time scaling method.", "authors": ["Nicolas Legrand", "Kenneth Enevoldsen", "Márton Kardos", "Kristoffer Nielbo"], "categories": ["cs.AI"], "fields_of_study": ["Computer Science"], "published_date": "2026-03-07", "url": "https://arxiv.org/abs/2603.07159", "pdf_url": "https://arxiv.org/pdf/2603.07159v1", "arxiv_id": "2603.07159", "doi": "10.48550/arXiv.2603.07159", "citation_count": 0, "influential_citation_count": 0, "has_code": false, "code_url": null, "venue": "arXiv.org", "quality_score": 0.4996}
{"id": "137010ace5c39f6319afcf24b5436e76155fd711b34a4207594632615d9db313", "sources": ["arxiv", "semantic_scholar"], "title": "Reasoning Models Struggle to Control their Chains of Thought", "abstract": "Chain-of-thought (CoT) monitoring is a promising tool for detecting misbehaviors and understanding the motivations of modern reasoning models. However, if models can control what they verbalize in their CoT, it could undermine CoT monitorability. To measure this undesirable capability -- CoT controllability -- we introduce the CoT-Control evaluation suite, which includes tasks that require models to solve problems while adhering to CoT instructions, e.g., reasoning about a genetics question without using the word 'chromosome'. We show that reasoning models possess significantly lower CoT controllability than output controllability; for instance, Claude Sonnet 4.5 can control its CoT only 2.7% of the time but 61.9% when controlling its final output. We also find that CoT controllability is higher for larger models and decreases with more RL training, test-time compute, and increased problem difficulty. CoT controllability failures extend even to situations in which models are given incentives (as opposed to direct requests) to evade CoT monitors, although models exhibit slightly higher controllability when they are told they are being monitored. Similarly, eliciting controllability by adversarially optimizing prompts does not meaningfully increase controllability. Our results leave us cautiously optimistic that CoT controllability is currently unlikely to be a failure mode of CoT monitorability. However, the mechanism behind low controllability is not well understood. Given its importance for maintaining CoT monitorability, we recommend that frontier labs track CoT controllability in future models.", "authors": ["Chen Yueh-Han", "Robert McCarthy", "Bruce W. Lee", "He He", "Ian Kivlichan", "Bowen Baker", "Micah Carroll", "Tomek Korbak"], "categories": ["cs.AI"], "fields_of_study": ["Computer Science"], "published_date": "2026-03-05", "url": "https://arxiv.org/abs/2603.05706", "pdf_url": "https://arxiv.org/pdf/2603.05706v1", "arxiv_id": "2603.05706", "doi": "10.48550/arXiv.2603.05706", "citation_count": 6, "influential_citation_count": 1, "has_code": false, "code_url": null, "venue": "arXiv.org", "quality_score": 0.4973}
{"id": "79c22fcb09d4e35448bf438ca5895ca961bf7c2a9e63575591e91d140a1527a6", "sources": ["arxiv", "semantic_scholar"], "title": "Beyond Test-Time Compute Strategies: Advocating Energy-per-Token in LLM Inference", "abstract": "Large Language Models (LLMs) demonstrate exceptional performance across diverse tasks but come with substantial energy and computational costs, particularly in request-heavy scenarios. In many real-world applications, the full scale and capabilities of LLMs are often unnecessary, as Small Language Models (SLMs) can provide accurate responses for simpler text generation tasks. When enhanced with advanced reasoning strategies, such as Chain-of-Thought (CoT) prompting or Majority Voting, SLMs can approach the performance of larger models while reducing overall computational requirements. However, these strategies can also introduce additional energy costs, creating an energy-accuracy trade-off. Our analysis examines these trade-offs in test-time compute strategies for smaller models compared to larger ones, using the MMLU benchmark. Additionally, we explore the input-output token dynamics of transformer architectures, which result in nonlinear hardware energy operation curves for LLMs. To bridge AI research with its physical impact, we propose \\textit{energy efficiency metrics}, including Energy-per-Token, as complements to traditional accuracy benchmarks. Beyond model selection, we propose controlled reasoning in CoT token generation, using operating curves to regulate reasoning depth dynamically. This vision integrates a energy-aware routing mechanism, ensuring that model selection and inference strategies balance accuracy for sustainable AI deployment.", "authors": ["Patrick Wilhelm", "Thorsten Wittkopp", "Odej Kao"], "categories": ["cs.CL"], "fields_of_study": ["Computer Science"], "published_date": "2026-03-04", "url": "https://arxiv.org/abs/2603.20224", "pdf_url": "https://arxiv.org/pdf/2603.20224v1", "arxiv_id": "2603.20224", "doi": "10.1145/3721146.3721953", "citation_count": 19, "influential_citation_count": 1, "has_code": false, "code_url": null, "venue": "EuroMLSys2025", "quality_score": 0.4961}
{"id": "9ee808f19b0f5f26762fc24c033b7a5c38c7c19a158644494f887dbb1ea70ccd", "sources": ["arxiv", "semantic_scholar"], "title": "Draft-Thinking: Learning Efficient Reasoning in Long Chain-of-Thought LLMs", "abstract": "Long chain-of-thought~(CoT) has become a dominant paradigm for enhancing the reasoning capability of large reasoning models~(LRMs); however, the performance gains often come with a substantial increase in reasoning budget. Recent studies show that existing CoT paradigms tend to induce systematic overthinking, unnecessarily coupling reasoning capability with reasoning cost. Most prior approaches reduce token usage through post hoc techniques such as token compression, truncation, or length penalties, without explicitly addressing the core mechanisms of reasoning. We propose \\textbf{Draft-Thinking}, which guides models to first learn a concise \\textit{draft-style} reasoning structure that retains only the critical reasoning steps. Through a \\textit{progressive curriculum learning}, the model stably internalizes this efficient reasoning pattern as its capability scales. Moreover, Draft-Thinking introduces adaptive prompting, which elevates reasoning depth to a flexible, model-selectable behavior. Extensive experiments demonstrate that Draft-Thinking substantially reduces reasoning budget while largely preserving reasoning performance; for example, on MATH500, it achieves an 82.6\\% reduction in reasoning budget at the cost of only a 2.6\\% performance drop.", "authors": ["Jie Cao", "Tianwei Lin", "Zhenxuan Fan", "Bo Yuan", "Ziyuan Zhao", "Rolan Yan", "Wenqiao Zhang", "Siliang Tang"], "categories": ["cs.AI", "cs.CL"], "fields_of_study": ["Computer Science"], "published_date": "2026-02-28", "url": "https://arxiv.org/abs/2603.00578", "pdf_url": "https://arxiv.org/pdf/2603.00578v1", "arxiv_id": "2603.00578", "doi": "10.48550/arXiv.2603.00578", "citation_count": 0, "influential_citation_count": 0, "has_code": false, "code_url": null, "venue": "arXiv.org", "quality_score": 0.4916}
{"id": "91da387316ddcc9b9ac8919fb53acbe53f9610ffa0c850d0b5219e7eba943072", "sources": ["arxiv", "semantic_scholar"], "title": "VisRef: Visual Refocusing while Thinking Improves Test-Time Scaling in Multi-Modal Large Reasoning Models", "abstract": "Advances in large reasoning models have shown strong performance on complex reasoning tasks by scaling test-time compute through extended reasoning. However, recent studies observe that in vision-dependent tasks, extended textual reasoning at inference time can degrade performance as models progressively lose attention to visual tokens and increasingly rely on textual priors alone. To address this, prior works use reinforcement learning (RL)-based fine-tuning to route visual tokens or employ refocusing mechanisms during reasoning. While effective, these methods are computationally expensive, requiring large-scale data generation and policy optimization. To leverage the benefits of test-time compute without additional RL fine-tuning, we propose VisRef, a visually grounded test-time scaling framework. Our key idea is to actively guide the reasoning process by re-injecting a coreset of visual tokens that are semantically relevant to the reasoning context while remaining diverse and globally representative of the image, enabling more grounded multi-modal reasoning. Experiments on three visual reasoning benchmarks with state-of-the-art multi-modal large reasoning models demonstrate that, under fixed test-time compute budgets, VisRef consistently outperforms existing test-time scaling approaches by up to 6.4%.", "authors": ["Soumya Suvra Ghosal", "Youngeun Kim", "Zhuowei Li", "Ritwick Chaudhry", "Linghan Xu", "Hongjing Zhang", "Jakub Zablocki", "Yifan Xing", "Qin Zhang"], "categories": ["cs.CV", "cs.AI"], "fields_of_study": ["Computer Science"], "published_date": "2026-02-27", "url": "https://arxiv.org/abs/2603.00207", "pdf_url": "https://arxiv.org/pdf/2603.00207v1", "arxiv_id": "2603.00207", "doi": "10.48550/arXiv.2603.00207", "citation_count": 0, "influential_citation_count": 0, "has_code": false, "code_url": null, "venue": "arXiv.org", "quality_score": 0.4904}
{"id": "9e2251c99a0b9de6fd5232a99e3761351907cb26f8113adec95c900e7e14a253", "sources": ["arxiv", "semantic_scholar"], "title": "Learning from Partial Chain-of-Thought via Truncated-Reasoning Self-Distillation", "abstract": "Reasoning-oriented language models achieve strong performance by generating long chain-of-thought traces at inference time. However, this capability comes with substantial and often excessive computational cost, which can materialize in redundant or inefficient reasoning. We study this setting and introduce Truncated-Reasoning Self-Distillation (TRSD), a lightweight post-training procedure that encourages models to produce correct predictions from partial reasoning traces. In TRSD, a frozen teacher model first generates a full reasoning trace and evaluates the corresponding answer distribution conditioned on the prompt and the complete reasoning to construct a synthetic training target. A student model with the same architecture is then trained to match the teacher's answer distribution while being conditioned only on a truncated prefix of its reasoning trace. Across multiple reasoning benchmarks and token budgets, we demonstrate that TRSD improves robustness to truncated inference, with far reduced accuracy tradeoffs when applied to a diverse set of reasoning models. Moreover, although never explicitly regularized for shorter generation during training, we also find that TRSD-trained models inherently output shorter reasoning traces without truncation, significantly reducing inference-time costs even without artificial interventions.", "authors": ["Gianluigi Silvestri", "Edoardo Cetin"], "categories": ["cs.LG", "cs.AI"], "fields_of_study": ["Computer Science"], "published_date": "2026-02-27", "url": "https://arxiv.org/abs/2603.13274", "pdf_url": "https://arxiv.org/pdf/2603.13274v1", "arxiv_id": "2603.13274", "doi": "10.48550/arXiv.2603.13274", "citation_count": 1, "influential_citation_count": 0, "has_code": false, "code_url": null, "venue": "arXiv.org", "quality_score": 0.4904}
{"id": "b62294fe17f98fb5b0ad0ec2a02499137542d113b06fbd659e3758d885b00edd", "sources": ["arxiv", "semantic_scholar"], "title": "Stepwise Penalization for Length-Efficient Chain-of-Thought Reasoning", "abstract": "Large reasoning models improve with more test-time computation, but often overthink, producing unnecessarily long chains-of-thought that raise cost without improving accuracy. Prior reinforcement learning approaches typically rely on a single outcome reward with trajectory-level length penalties, which cannot distinguish essential from redundant reasoning steps and therefore yield blunt compression. Although recent work incorporates step-level signals, such as offline pruning, supervised data construction, or verifier-based intermediate rewards, reasoning length is rarely treated as an explicit step-level optimization objective during RL. We propose Step-wise Adaptive Penalization (SWAP), a fine-grained framework that allocates length reduction across steps based on intrinsic contribution. We estimate step importance from the model's on-policy log-probability improvement toward the correct answer, then treat excess length as a penalty mass redistributed to penalize low-importance steps more heavily while preserving high-importance reasoning. We optimize with a unified outcome-process advantage within group-relative policy optimization. Extensive experiments demonstrate that SWAP reduces reasoning length by 64.3% on average while improving accuracy by 5.7% relative to the base model.", "authors": ["Xintong Li", "Sha Li", "Rongmei Lin", "Hongye Jin", "Linwei Li", "Hejie Cui", "Sarah Zhang", "Chia-Yuan Chang", "Kewei Cheng", "Besnik Fetahu", "Priyanka Nigam", "Jingbo Shang", "Bing Yin"], "categories": ["cs.CL", "cs.AI", "cs.LG"], "fields_of_study": ["Computer Science"], "published_date": "2026-02-27", "url": "https://arxiv.org/abs/2603.00296", "pdf_url": "https://arxiv.org/pdf/2603.00296v1", "arxiv_id": "2603.00296", "doi": "10.48550/arXiv.2603.00296", "citation_count": 1, "influential_citation_count": 0, "has_code": false, "code_url": null, "venue": "arXiv.org", "quality_score": 0.4904}
{"id": "a9b8254531790e14a4f1e625155213825a8eacd9a70802c918ea7c2f335a389c", "sources": ["arxiv", "semantic_scholar"], "title": "TCM-DiffRAG: Personalized Syndrome Differentiation Reasoning Method for Traditional Chinese Medicine based on Knowledge Graph and Chain of Thought", "abstract": "Background: Retrieval augmented generation (RAG) technology can empower large language models (LLMs) to generate more accurate, professional, and timely responses without fine tuning. However, due to the complex reasoning processes and substantial individual differences involved in traditional Chinese medicine (TCM) clinical diagnosis and treatment, traditional RAG methods often exhibit poor performance in this domain. Objective: To address the limitations of conventional RAG approaches in TCM applications, this study aims to develop an improved RAG framework tailored to the characteristics of TCM reasoning. Methods: We developed TCM-DiffRAG, an innovative RAG framework that integrates knowledge graphs (KG) with chains of thought (CoT). TCM-DiffRAG was evaluated on three distinctive TCM test datasets. Results: The experimental results demonstrated that TCM-DiffRAG achieved significant performance improvements over native LLMs. For example, the qwen-plus model achieved scores of 0.927, 0.361, and 0.038, which were significantly enhanced to 0.952, 0.788, and 0.356 with TCM-DiffRAG. The improvements were even more pronounced for non-Chinese LLMs. Additionally, TCM-DiffRAG outperformed directly supervised fine-tuned (SFT) LLMs and other benchmark RAG methods. Conclusions: TCM-DiffRAG shows that integrating structured TCM knowledge graphs with Chain of Thought based reasoning substantially improves performance in individualized diagnostic tasks. The joint use of universal and personalized knowledge graphs enables effective alignment between general knowledge and clinical reasoning. These results highlight the potential of reasoning-aware RAG frameworks for advancing LLM applications in traditional Chinese medicine.", "authors": ["Jianmin Li", "Ying Chang", "Su-Kit Tang", "Yujia Liu", "Yanwen Wang", "Shuyuan Lin", "Binkai Ou"], "categories": ["cs.CL", "cs.AI"], "fields_of_study": ["Computer Science", "Medicine"], "published_date": "2026-02-26", "url": "https://arxiv.org/abs/2602.22828", "pdf_url": "https://arxiv.org/pdf/2602.22828v1", "arxiv_id": "2602.22828", "doi": "10.3389/fmed.2026.1804478", "citation_count": 0, "influential_citation_count": 0, "has_code": false, "code_url": null, "venue": "Frontiers in Medicine", "quality_score": 0.4893}
{"id": "0bb3ae82830e54c58ea6809aca6b41bf97a9ea84b4fc6c2af218ace9e83a2d26", "sources": ["arxiv", "semantic_scholar"], "title": "D-COT: Disciplined Chain-of-Thought Learning for Efficient Reasoning in Small Language Models", "abstract": "Chain-of-Thought (CoT) distillation from Large Language Models (LLMs) often induces \"overthinking\" in Small Language Models (SLMs), leading to performance degradation and excessive token consumption. In this study, we propose Disciplined Chain-of-Thought (D-CoT), a novel framework that enforces a structured reasoning process using control tags -- such as <TEMP_LOW> for fact-checking and <TEMP_HIGH> for multi-perspective exploration -- as auxiliary scaffolding during training. By optimizing the CoT trajectory, D-CoT suppresses reasoning drift and simultaneously achieves token reduction and performance improvement. We demonstrate the efficacy of our approach on Qwen3-8B: with only 5,000 training samples, D-CoT significantly boosts accuracy on GPQA-diamond by 9.9% and MMLU-Pro (0-shot) by 9.1%, while drastically reducing computational costs. Furthermore, we confirm that the model internalizes this disciplined thought structure, maintaining high performance even without explicit control tags during inference.", "authors": ["Shunsuke Ubukata"], "categories": ["cs.CL"], "fields_of_study": ["Computer Science"], "published_date": "2026-02-25", "url": "https://arxiv.org/abs/2602.21786", "pdf_url": "https://arxiv.org/pdf/2602.21786v1", "arxiv_id": "2602.21786", "doi": "10.48550/arXiv.2602.21786", "citation_count": 0, "influential_citation_count": 0, "has_code": true, "code_url": "https://github.com/gitpullpull/DisciplinedChainOfThought", "venue": "arXiv.org", "quality_score": 0.7544}
{"id": "e37d453a9265820365d533c683bd4c97821ea0ed6eb7c87d3b748451d57929a7", "sources": ["arxiv", "semantic_scholar"], "title": "The Art of Efficient Reasoning: Data, Reward, and Optimization", "abstract": "Large Language Models (LLMs) consistently benefit from scaled Chain-of-Thought (CoT) reasoning, but also suffer from heavy computational overhead. To address this issue, efficient reasoning aims to incentivize short yet accurate thinking trajectories, typically through reward shaping with Reinforcement Learning (RL). In this paper, we systematically investigate the mechanics of efficient reasoning for LLMs. For comprehensive evaluation, we advocate for more fine-grained metrics, including length distribution conditioned on correctness and performance across a wide spectrum of token budgets ranging from 2k to 32k. First, we reveal that the training process follows a two-stage paradigm: length adaptation and reasoning refinement. Through extensive experiments (about 0.2 million GPU hours) in a unified protocol, we deconstruct training prompts and rollouts, reward shaping, and optimization strategies. A central finding is to maintain a sufficient density of positive reward signals and avoid the short-is-correct trap. Moreover, the learned length bias generalizes across domains and difficulty levels. We distill these findings into valuable insights and practical guidelines, and validate them across the Qwen3 models ranging from 0.6B to 30B, demonstrating the robustness and generalization. Weights are available at https://wutaiqiang.github.io/project/Art", "authors": ["Taiqiang Wu", "Zenan Xu", "Bo Zhou", "Ngai Wong"], "categories": ["cs.CL", "cs.AI"], "fields_of_study": ["Computer Science"], "published_date": "2026-02-24", "url": "https://arxiv.org/abs/2602.20945", "pdf_url": "https://arxiv.org/pdf/2602.20945v3", "arxiv_id": "2602.20945", "doi": "10.48550/arXiv.2602.20945", "citation_count": 2, "influential_citation_count": 0, "has_code": false, "code_url": null, "venue": "arXiv.org", "quality_score": 0.487}
{"id": "fc5a7fe56f7fb18560be34fd6ec26658e61a5948fa4b937e8ea0a670a695fd69", "sources": ["arxiv", "semantic_scholar"], "title": "To Reason or Not to: Selective Chain-of-Thought in Medical Question Answering", "abstract": "Objective: To improve the efficiency of medical question answering (MedQA) with large language models (LLMs) by avoiding unnecessary reasoning while maintaining accuracy. Methods: We propose Selective Chain-of-Thought (Selective CoT), an inference-time strategy that first predicts whether a question requires reasoning and generates a rationale only when needed. Two open-source LLMs (Llama-3.1-8B and Qwen-2.5-7B) were evaluated on four biomedical QA benchmarks-HeadQA, MedQA-USMLE, MedMCQA, and PubMedQA. Metrics included accuracy, total generated tokens, and inference time. Results: Selective CoT reduced inference time by 13-45% and token usage by 8-47% with minimal accuracy loss ($\\leq$4\\%). In some model-task pairs, it achieved both higher accuracy and greater efficiency than standard CoT. Compared with fixed-length CoT, Selective CoT reached similar or superior accuracy at substantially lower computational cost. Discussion: Selective CoT dynamically balances reasoning depth and efficiency by invoking explicit reasoning only when beneficial, reducing redundancy on recall-type questions while preserving interpretability. Conclusion: Selective CoT provides a simple, model-agnostic, and cost-effective approach for medical QA, aligning reasoning effort with question complexity to enhance real-world deployability of LLM-based clinical systems.", "authors": ["Zaifu Zhan", "Min Zeng", "Shuang Zhou", "Yiran Song", "Xiaoyi Chen", "Yu Hou", "Yifan Wu", "Yang Ruan", "Rui Zhang"], "categories": ["cs.CL", "cs.AI"], "fields_of_study": ["Computer Science"], "published_date": "2026-02-23", "url": "https://arxiv.org/abs/2602.20130", "pdf_url": "https://arxiv.org/pdf/2602.20130v1", "arxiv_id": "2602.20130", "doi": "10.48550/arXiv.2602.20130", "citation_count": 0, "influential_citation_count": 0, "has_code": true, "code_url": null, "venue": "arXiv.org", "quality_score": 0.7508}
{"id": "cecc3fe9230371dc0f84492904b6c929e969df2f9348877e263bb9db0b2b487e", "sources": ["arxiv", "semantic_scholar"], "title": "Evaluating Chain-of-Thought Reasoning through Reusability and Verifiability", "abstract": "In multi-agent IR pipelines for tasks such as search and ranking, LLM-based agents exchange intermediate reasoning in terms of Chain-of-Thought (CoT) with each other. Current CoT evaluation narrowly focuses on target task accuracy. However, this metric fails to assess the quality or utility of the reasoning process itself. To address this limitation, we introduce two novel measures: reusability and verifiability. We decouple CoT generation from execution using a Thinker-Executor framework. Reusability measures how easily an Executor can reuse the Thinker's CoT. Verifiability measures how frequently an Executor can match the Thinker's answer using the CoT. We evaluated four Thinker models against a committee of ten Executor models across five benchmarks. Our results reveal that reusability and verifiability do not correlate with standard accuracy, exposing a blind spot in current accuracy-based leaderboards for reasoning capability. Surprisingly, we find that CoTs from specialized reasoning models are not consistently more reusable or verifiable than those from general-purpose LLMs like Llama and Gemma.", "authors": ["Shashank Aggarwal", "Ram Vikas Mishra", "Amit Awekar"], "categories": ["cs.AI", "cs.CL", "cs.IR"], "fields_of_study": ["Computer Science"], "published_date": "2026-02-19", "url": "https://arxiv.org/abs/2602.17544", "pdf_url": "https://arxiv.org/pdf/2602.17544v1", "arxiv_id": "2602.17544", "doi": "10.48550/arXiv.2602.17544", "citation_count": 3, "influential_citation_count": 0, "has_code": false, "code_url": null, "venue": "arXiv.org", "quality_score": 0.4813}
{"id": "62c4c284364a2abb2c1ca513c64d580beb7f9fb2275350a1b6a7fcede8397b82", "sources": ["arxiv", "semantic_scholar"], "title": "Framework of Thoughts: A Foundation Framework for Dynamic and Optimized Reasoning based on Chains, Trees, and Graphs", "abstract": "Prompting schemes such as Chain of Thought, Tree of Thoughts, and Graph of Thoughts can significantly enhance the reasoning capabilities of large language models. However, most existing schemes require users to define static, problem-specific reasoning structures that lack adaptability to dynamic or unseen problem types. Additionally, these schemes are often under-optimized in terms of hyperparameters, prompts, runtime, and prompting cost. To address these limitations, we introduce Framework of Thoughts (FoT)--a general-purpose foundation framework for building and optimizing dynamic reasoning schemes. FoT comes with built-in features for hyperparameter tuning, prompt optimization, parallel execution, and intelligent caching, unlocking the latent performance potential of reasoning schemes. We demonstrate FoT's capabilities by implementing three popular schemes--Tree of Thoughts, Graph of Thoughts, and ProbTree--within FoT. We empirically show that FoT enables significantly faster execution, reduces costs, and achieves better task scores through optimization. We release our codebase to facilitate the development of future dynamic and efficient reasoning schemes.", "authors": ["Felix Fricke", "Simon Malberg", "Georg Groh"], "categories": ["cs.AI"], "fields_of_study": ["Computer Science"], "published_date": "2026-02-18", "url": "https://arxiv.org/abs/2602.16512", "pdf_url": "https://arxiv.org/pdf/2602.16512v1", "arxiv_id": "2602.16512", "doi": "10.48550/arXiv.2602.16512", "citation_count": 1, "influential_citation_count": 0, "has_code": false, "code_url": null, "venue": "arXiv.org", "quality_score": 0.4801}
{"id": "cfcfdc6f8e95f0c7f406a633e639b9ac64e4c28e9b28e973ff6ddb98fb608c79", "sources": ["arxiv", "semantic_scholar"], "title": "GTS: Inference-Time Scaling of Latent Reasoning with a Learnable Gaussian Thought Sampler", "abstract": "Inference-time scaling (ITS) in latent reasoning models typically relies on heuristic perturbations, such as dropout or fixed Gaussian noise, to generate diverse candidate trajectories. However, we show that stronger perturbations do not necessarily yield better sampling quality: they often induce larger distribution shifts without producing more useful reasoning paths or better final decisions. A key limitation is that these perturbations inject stochasticity without defining an explicit conditional sampling distribution, making latent exploration difficult to control or optimize. To address this, we propose the Gaussian Thought Sampler (GTS), a lightweight module that reformulates latent exploration as sampling from a learned conditional distribution over continuous reasoning states. GTS predicts context-dependent perturbation distributions and is trained with GRPO-style policy optimization while keeping the backbone frozen, turning heuristic perturbation into an explicit probabilistic sampling policy. Experiments across multiple benchmarks and two latent reasoning architectures show that GTS yields more reliable inference-time scaling than heuristic baselines, suggesting that effective latent ITS requires better-controlled and optimizable sampling rather than simply amplifying stochasticity.", "authors": ["Minghan Wang", "Ye Bai", "Thuy-Trang Vu", "Ehsan Shareghi", "Gholamreza Haffari"], "categories": ["cs.CL", "cs.LG"], "fields_of_study": ["Computer Science"], "published_date": "2026-02-15", "url": "https://arxiv.org/abs/2602.14077", "pdf_url": "https://arxiv.org/pdf/2602.14077v2", "arxiv_id": "2602.14077", "doi": "10.48550/arXiv.2602.14077", "citation_count": 0, "influential_citation_count": 0, "has_code": false, "code_url": null, "venue": "arXiv.org", "quality_score": 0.4767}
{"id": "ded16176fd7022ea5f21054aed51471a6010ae6151ac6eae53361a1a0056e31d", "sources": ["arxiv", "semantic_scholar"], "title": "The Interspeech 2026 Audio Reasoning Challenge: Evaluating Reasoning Process Quality for Audio Reasoning Models and Agents", "abstract": "Recent Large Audio Language Models (LALMs) excel in understanding but often lack transparent reasoning. To address this \"black-box\" limitation, we organized the Audio Reasoning Challenge at Interspeech 2026, the first shared task dedicated to evaluating Chain-of-Thought (CoT) quality in the audio domain. The challenge introduced MMAR-Rubrics, a novel instance-level protocol assessing the factuality and logic of reasoning chains. Featured Single Model and Agent tracks, the competition attracting 156 teams from 18 countries and regions. Results show agent systems currently lead in reasoning quality, utilizing iterative tool orchestration and cross-modal analysis. Besides, single models are rapidly advancing via reinforcement learning and sophisticated data pipeline. We details the challenge design, methodology, and a comprehensive analysis of state-of-the-art systems, providing new insights for explainable audio intelligence.", "authors": ["Ziyang Ma", "Ruiyang Xu", "Yinghao Ma", "Chao-Han Huck Yang", "Bohan Li", "Jaeyeon Kim", "Jin Xu", "Jinyu Li", "Carlos Busso", "Kai Yu", "Eng Siong Chng", "Xie Chen"], "categories": ["cs.SD", "cs.CL", "cs.MM"], "fields_of_study": ["Computer Science"], "published_date": "2026-02-15", "url": "https://arxiv.org/abs/2602.14224", "pdf_url": "https://arxiv.org/pdf/2602.14224v1", "arxiv_id": "2602.14224", "doi": "10.48550/arXiv.2602.14224", "citation_count": 9, "influential_citation_count": 1, "has_code": false, "code_url": null, "venue": "arXiv.org", "quality_score": 0.4767}
{"id": "c9e1417ccbbf6bb42215bdc992e88b7880857b40e461add2daa0206b5bd2980f", "sources": ["arxiv", "semantic_scholar"], "title": "Diagnosing Pathological Chain-of-Thought in Reasoning Models", "abstract": "Chain-of-thought (CoT) reasoning is fundamental to modern LLM architectures and represents a critical intervention point for AI safety. However, CoT reasoning may exhibit failure modes that we note as pathologies, which prevent it from being useful for monitoring. Prior work has identified three distinct pathologies: post-hoc rationalization, where models generate plausible explanations backwards from predetermined answers; encoded reasoning, where intermediate steps conceal information within seemingly interpretable text; and internalized reasoning, where models replace explicit reasoning with meaningless filler tokens while computing internally. To better understand and discriminate between these pathologies, we create a set of concrete metrics that are simple to implement, computationally inexpensive, and task-agnostic. To validate our approach, we develop model organisms deliberately trained to exhibit specific CoT pathologies. Our work provides a practical toolkit for assessing CoT pathologies, with direct implications for training-time monitoring.", "authors": ["Manqing Liu", "David Williams-King", "Ida Caspary", "Linh Le", "Hannes Whittingham", "Puria Radmard", "Cameron Tice", "Edward James Young"], "categories": ["cs.AI"], "fields_of_study": ["Computer Science"], "published_date": "2026-02-14", "url": "https://arxiv.org/abs/2602.13904", "pdf_url": "https://arxiv.org/pdf/2602.13904v1", "arxiv_id": "2602.13904", "doi": "10.48550/arXiv.2602.13904", "citation_count": 1, "influential_citation_count": 0, "has_code": false, "code_url": null, "venue": "arXiv.org", "quality_score": 0.4755}
{"id": "72ba21d1644158a658cbe0f366a77a750b9ac123bfb46774a65c9ad839e3fe3c", "sources": ["arxiv", "semantic_scholar"], "title": "UniT: Unified Multimodal Chain-of-Thought Test-time Scaling", "abstract": "Unified models can handle both multimodal understanding and generation within a single architecture, yet they typically operate in a single pass without iteratively refining their outputs. Many multimodal tasks, especially those involving complex spatial compositions, multiple interacting objects, or evolving instructions, require decomposing instructions, verifying intermediate results, and making iterative corrections. While test-time scaling (TTS) has demonstrated that allocating additional inference compute for iterative reasoning substantially improves language model performance, extending this paradigm to unified multimodal models remains an open challenge. We introduce UniT, a framework for multimodal chain-of-thought test-time scaling that enables a single unified model to reason, verify, and refine across multiple rounds. UniT combines agentic data synthesis, unified model training, and flexible test-time inference to elicit cognitive behaviors including verification, subgoal decomposition, and content memory. Our key findings are: (1) unified models trained on short reasoning trajectories generalize to longer inference chains at test time; (2) sequential chain-of-thought reasoning provides a more scalable and compute-efficient TTS strategy than parallel sampling; (3) training on generation and editing trajectories improves out-of-distribution visual reasoning. These results establish multimodal test-time scaling as an effective paradigm for advancing both generation and understanding in unified models.", "authors": ["Leon Liangyu Chen", "Haoyu Ma", "Zhipeng Fan", "Ziqi Huang", "Animesh Sinha", "Xiaoliang Dai", "Jialiang Wang", "Zecheng He", "Jianwei Yang", "Chunyuan Li", "Junzhe Sun", "Chu Wang", "Serena Yeung-Levy", "Felix Juefei-Xu"], "categories": ["cs.CV", "cs.AI", "cs.LG"], "fields_of_study": ["Computer Science"], "published_date": "2026-02-12", "url": "https://arxiv.org/abs/2602.12279", "pdf_url": "https://arxiv.org/pdf/2602.12279v2", "arxiv_id": "2602.12279", "doi": "10.48550/arXiv.2602.12279", "citation_count": 3, "influential_citation_count": 0, "has_code": false, "code_url": null, "venue": "arXiv.org", "quality_score": 0.4732}
{"id": "e984ecf62b7b268bb62c552659350cd01f388a8cf8bb4f0f30f44d4e075b0279", "sources": ["arxiv", "semantic_scholar"], "title": "Characterizing, Evaluating, and Optimizing Complex Reasoning", "abstract": "Large Reasoning Models (LRMs) increasingly rely on reasoning traces with complex internal structures. However, existing work lacks a unified answer to three fundamental questions: (1) what defines high-quality reasoning, (2) how to reliably evaluate long, implicitly structured reasoning traces, and (3) how to use such evaluation signals for reasoning optimization. To address these challenges, we provide a unified perspective. (1) We introduce the ME$^2$ principle to characterize reasoning quality along macro- and micro-level concerning efficiency and effectiveness. (2) Built on this principle, we model reasoning traces as directed acyclic graphs (DAGs) and develop a DAG-based pairwise evaluation method, capturing complex reasoning structures. (3) Based on this method, we construct the TRM-Preference dataset and train a Thinking Reward Model (TRM) to evaluate reasoning quality at scale. Experiments show that thinking rewards serve as an effective optimization signal. At test time, selecting better reasoning leads to better outcomes (up to 19.3\\% gain), and during RL training, thinking rewards enhance reasoning and performance (up to 3.9\\% gain) across diverse tasks. Code and data are available at https://github.com/Simplified-Reasoning/TRM.", "authors": ["Haoran Zhang", "Yafu Li", "Zhi Wang", "Zhilin Wang", "Shunkai Zhang", "Xiaoye Qu", "Yu Cheng"], "categories": ["cs.CL"], "fields_of_study": ["Computer Science"], "published_date": "2026-02-09", "url": "https://arxiv.org/abs/2602.08498", "pdf_url": "https://arxiv.org/pdf/2602.08498v2", "arxiv_id": "2602.08498", "doi": "10.48550/arXiv.2602.08498", "citation_count": 1, "influential_citation_count": 0, "has_code": true, "code_url": "https://github.com/Simplified-Reasoning/TRM", "venue": "arXiv.org", "quality_score": 0.726}
{"id": "33c73e02de489b03686fdd51a186597428b0d0f2d22a5d304fc170ff07fad601", "sources": ["arxiv", "semantic_scholar"], "title": "Towards Efficient Large Language Reasoning Models via Extreme-Ratio Chain-of-Thought Compression", "abstract": "Chain-of-Thought (CoT) reasoning successfully enhances the reasoning capabilities of Large Language Models (LLMs), yet it incurs substantial computational overhead for inference. Existing CoT compression methods often suffer from a critical loss of logical fidelity at high compression ratios, resulting in significant performance degradation. To achieve high-fidelity, fast reasoning, we propose a novel EXTreme-RAtio Chain-of-Thought Compression framework, termed Extra-CoT, which aggressively reduces the token budget while preserving answer accuracy. To generate reliable, high-fidelity supervision, we first train a dedicated semantically-preserved compressor on mathematical CoT data with fine-grained annotations. An LLM is then fine-tuned on these compressed pairs via a mixed-ratio supervised fine-tuning (SFT), teaching it to follow a spectrum of compression budgets and providing a stable initialization for reinforcement learning (RL). We further propose Constrained and Hierarchical Ratio Policy Optimization (CHRPO) to explicitly incentivize question-solving ability under lower budgets by a hierarchical reward. Experiments on three mathematical reasoning benchmarks show the superiority of Extra-CoT. For example, on MATH-500 using Qwen3-1.7B, Extra-CoT achieves over 73\\% token reduction with an accuracy improvement of 0.6\\%, significantly outperforming state-of-the-art (SOTA) methods. Our source codes have been released at https://github.com/Mwie1024/Extra-CoT.", "authors": ["Yuntian Tang", "Bohan Jia", "Wenxuan Huang", "Lianyue Zhang", "Jiao Xie", "Wenxi Li", "Wei Li", "Jie Hu", "Xinghao Chen Rongrong Ji", "Shaohui Lin"], "categories": ["cs.LG"], "fields_of_study": ["Computer Science"], "published_date": "2026-02-09", "url": "https://arxiv.org/abs/2602.08324", "pdf_url": "https://arxiv.org/pdf/2602.08324v5", "arxiv_id": "2602.08324", "doi": "10.48550/arXiv.2602.08324", "citation_count": 3, "influential_citation_count": 0, "has_code": true, "code_url": "https://github.com/Mwie1024/Extra-CoT", "venue": "arXiv.org", "quality_score": 0.726}
{"id": "f4f2cdbb1f79f602792fe318baeb232e641041a4847d7fc9cae7f6bffc167061", "sources": ["arxiv", "semantic_scholar"], "title": "Recurrent-Depth VLA: Implicit Test-Time Compute Scaling of Vision-Language-Action Models via Latent Iterative Reasoning", "abstract": "Current Vision-Language-Action (VLA) models rely on fixed computational depth, expending the same amount of compute on simple adjustments and complex multi-step manipulation. While Chain-of-Thought (CoT) prompting enables variable computation, it scales memory linearly and is ill-suited for continuous action spaces. We introduce Recurrent-Depth VLA (RD-VLA), an architecture that achieves computational adaptivity via latent iterative refinement rather than explicit token generation. RD-VLA employs a recurrent, weight-tied action head that supports arbitrary inference depth with a constant memory footprint. The model is trained using truncated backpropagation through time (TBPTT) to efficiently supervise the refinement process. At inference, RD-VLA dynamically allocates compute using an adaptive stopping criterion based on latent convergence. Experiments on challenging manipulation tasks show that recurrent depth is critical: tasks that fail entirely (0 percent success) with single-iteration inference exceed 90 percent success with four iterations, while simpler tasks saturate rapidly. RD-VLA provides a scalable path to test-time compute in robotics, replacing token-based reasoning with latent reasoning to achieve constant memory usage and up to 80x inference speedup over prior reasoning-based VLA models. Project page: https://rd-vla.github.io/", "authors": ["Yalcin Tur", "Jalal Naghiyev", "Haoquan Fang", "Wei-Chuan Tsai", "Jiafei Duan", "Dieter Fox", "Ranjay Krishna"], "categories": ["cs.RO"], "fields_of_study": ["Computer Science"], "published_date": "2026-02-08", "url": "https://arxiv.org/abs/2602.07845", "pdf_url": "https://arxiv.org/pdf/2602.07845v1", "arxiv_id": "2602.07845", "doi": "10.48550/arXiv.2602.07845", "citation_count": 6, "influential_citation_count": 1, "has_code": false, "code_url": null, "venue": "arXiv.org", "quality_score": 0.4686}
{"id": "5301e5ea0613803d9fe21d848abd13832061ca0eda4e7fea13da288f1f41a82e", "sources": ["arxiv", "semantic_scholar"], "title": "Surprisal-Guided Selection: Compute-Optimal Test-Time Strategies for Execution-Grounded Code Generation", "abstract": "Test-time training (TTT) adapts language models through gradient-based updates at inference. But is adaptation the right strategy? We study compute-optimal test-time strategies for verifiable execution-grounded (VEG) tasks, domains like GPU kernel optimization where a deterministic evaluator provides dense, continuous reward signals. Using KernelBench as our testbed and a 120B-parameter model (GPT-OSS-120B with LoRA adaptation), we find that search outperforms minimal adaptation (1-5 gradient steps): Best-of-N sampling achieves 90% task success (18/20 tasks) at K=64 across the full KernelBench L1 eval set while TTT's best checkpoint reaches only 30.6% (3-seed mean), with TTT's \"equivalent K\" falling below 1, worse than single-sample inference. The failure mode is over-sharpening: gradient updates collapse diversity toward mediocre solutions rather than discovering optimal ones. Our main contribution is surprisal-guided selection: selecting the highest-surprisal (lowest-confidence) correct sample yields 80% success vs. 50% for most-confident selection, a 30% improvement. Extending to surprisal-guided-top3 matches oracle performance at 100%. This zero-cost strategy, validated through length-controlled analysis, recovers oracle performance. For dense-reward VEG tasks, compute should be allocated to sample diversity and intelligent selection rather than gradient adaptation. The surprisal-guided selection principle may generalize to other execution-grounded domains where optimal solutions occupy the distribution tail.", "authors": ["Jarrod Barnes"], "categories": ["cs.LG", "cs.AI"], "fields_of_study": ["Computer Science"], "published_date": "2026-02-07", "url": "https://arxiv.org/abs/2602.07670", "pdf_url": "https://arxiv.org/pdf/2602.07670v1", "arxiv_id": "2602.07670", "doi": "10.48550/arXiv.2602.07670", "citation_count": 0, "influential_citation_count": 0, "has_code": true, "code_url": "https://github.com/jbarnes850/test-time-training", "venue": "arXiv.org", "quality_score": 0.7225}
{"id": "cba375976102edb252d9f26ea2a4383828973bfa5accef3441c18e5b415114bd", "sources": ["arxiv", "semantic_scholar"], "title": "Are Reasoning LLMs Robust to Interventions on Their Chain-of-Thought?", "abstract": "Reasoning LLMs (RLLMs) generate step-by-step chains of thought (CoTs) before giving an answer, which improves performance on complex tasks and makes reasoning more transparent. But how robust are these reasoning traces to disruptions that occur within them? To address this question, we introduce a controlled evaluation framework that perturbs a model's own CoT at fixed timesteps. We design seven interventions (benign, neutral, and adversarial) and apply them to multiple open-weight RLLMs across Math, Science, and Logic tasks. Our results show that RLLMs are generally robust, reliably recovering from diverse perturbations, with robustness improving with model size and degrading when interventions occur early. However, robustness is not style-invariant: paraphrasing suppresses doubt-like expressions and reduces performance, while other interventions trigger doubt and support recovery. Recovery also carries a cost: neutral and adversarial noise can inflate CoT length by more than 200%, whereas paraphrasing shortens traces but harms accuracy. These findings provide new evidence on how RLLMs maintain reasoning integrity, identify doubt as a central recovery mechanism, and highlight trade-offs between robustness and efficiency that future training methods should address.", "authors": ["Alexander von Recum", "Leander Girrbach", "Zeynep Akata"], "categories": ["cs.AI"], "fields_of_study": ["Computer Science"], "published_date": "2026-02-07", "url": "https://arxiv.org/abs/2602.07470", "pdf_url": "https://arxiv.org/pdf/2602.07470v1", "arxiv_id": "2602.07470", "doi": "10.48550/arXiv.2602.07470", "citation_count": 4, "influential_citation_count": 0, "has_code": false, "code_url": null, "venue": "arXiv.org", "quality_score": 0.4675}
{"id": "a46b0d3436f4e1c619b67d8784861348153b3d306cf87ad2306d234bbe6a645c", "sources": ["arxiv", "semantic_scholar"], "title": "Inference-Time Rethinking with Latent Thought Vectors for Math Reasoning", "abstract": "Standard chain-of-thought reasoning generates a solution in a single forward pass, committing irrevocably to each token and lacking a mechanism to recover from early errors. We introduce Inference-Time Rethinking, a generative framework that enables iterative self-correction by decoupling declarative latent thought vectors from procedural generation. We factorize reasoning into a continuous latent thought vector (what to reason about) and a decoder that verbalizes the trace conditioned on this vector (how to reason). Beyond serving as a declarative buffer, latent thought vectors compress the reasoning structure into a continuous representation that abstracts away surface-level token variability, making gradient-based optimization over reasoning strategies well-posed. Our prior model maps unstructured noise to a learned manifold of valid reasoning patterns, and at test time we employ a Gibbs-style procedure that alternates between generating a candidate trace and optimizing the latent vector to better explain that trace, effectively navigating the latent manifold to refine the reasoning strategy. Training a 0.2B-parameter model from scratch on GSM8K, our method with 30 rethinking iterations surpasses baselines with 10 to 15 times more parameters, including a 3B counterpart. This result demonstrates that effective mathematical reasoning can emerge from sophisticated inference-time computation rather than solely from massive parameter counts.", "authors": ["Deqian Kong", "Minglu Zhao", "Aoyang Qin", "Bo Pang", "Chenxin Tao", "David Hartmann", "Edouardo Honig", "Dehong Xu", "Amit Kumar", "Matt Sarte", "Chuan Li", "Jianwen Xie", "Ying Nian Wu"], "categories": ["cs.CL", "cs.LG", "stat.ML"], "fields_of_study": ["Computer Science", "Mathematics"], "published_date": "2026-02-06", "url": "https://arxiv.org/abs/2602.06584", "pdf_url": "https://arxiv.org/pdf/2602.06584v1", "arxiv_id": "2602.06584", "doi": "10.48550/arXiv.2602.06584", "citation_count": 0, "influential_citation_count": 0, "has_code": false, "code_url": null, "venue": "arXiv.org", "quality_score": 0.4664}
{"id": "fb5d3bc6cb961121f789a9f6f7d37778684d989e4c48783c0b4a20377cf5ba93", "sources": ["arxiv", "semantic_scholar"], "title": "Intrinsic Stability Limits of Autoregressive Reasoning: Structural Consequences for Long-Horizon Execution", "abstract": "Large language models (LLMs) demonstrate remarkable reasoning capabilities, yet their performance often deteriorates sharply in long-horizon tasks, exhibiting systematic breakdown beyond certain scales. Conventional explanations primarily attribute this phenomenon to task complexity, such as combinatorial search explosion or long-term credit assignment challenges. In this work, we argue that these explanations are incomplete: even in linear, unbranched tasks without semantic ambiguity, autoregressive execution is subject to an intrinsic stability limit. We propose that the fundamental constraint on long-horizon reasoning arises from process-level instability in autoregressive generation rather than solely from search or task complexity, reframing long-horizon reasoning as a problem of structural governance. We derive Theorem~A, showing that decision advantage in single-path autoregressive reasoning decays exponentially with execution length, imposing a fundamental bound on maintainable reasoning chains. This result implies a structural consequence: stable long-horizon reasoning requires discrete segmentation, naturally inducing graph-like execution structures such as directed acyclic graphs (DAGs). Empirical studies in both synthetic environments and real TextWorld tasks reveal observable performance cliffs consistent with theoretical predictions. Our findings provide a dynamical perspective on long-horizon reasoning failure and suggest new limitations on maintaining long-term coherence under purely autoregressive architectures. Furthermore, we highlight that short-horizon evaluation protocols may obscure structural instability, indicating a potential shift from scaling toward structured governance in future reasoning systems.", "authors": ["Hsien-Jyh Liao"], "categories": ["cs.AI"], "fields_of_study": ["Computer Science"], "published_date": "2026-02-06", "url": "https://arxiv.org/abs/2602.06413", "pdf_url": "https://arxiv.org/pdf/2602.06413v1", "arxiv_id": "2602.06413", "doi": "10.48550/arXiv.2602.06413", "citation_count": 0, "influential_citation_count": 0, "has_code": false, "code_url": null, "venue": "arXiv.org", "quality_score": 0.4664}
{"id": "51c699147a3a07ed163b45f7ec63cb145ce6db768edee681ed14fd8285bd5a5f", "sources": ["arxiv", "semantic_scholar"], "title": "SPARC: Separating Perception And Reasoning Circuits for Test-time Scaling of VLMs", "abstract": "Despite recent successes, test-time scaling - i.e., dynamically expanding the token budget during inference as needed - remains brittle for vision-language models (VLMs): unstructured chains-of-thought about images entangle perception and reasoning, leading to long, disorganized contexts where small perceptual mistakes may cascade into completely wrong answers. Moreover, expensive reinforcement learning with hand-crafted rewards is required to achieve good performance. Here, we introduce SPARC (Separating Perception And Reasoning Circuits), a modular framework that explicitly decouples visual perception from reasoning. Inspired by sequential sensory-to-cognitive processing in the brain, SPARC implements a two-stage pipeline where the model first performs explicit visual search to localize question-relevant regions, then conditions its reasoning on those regions to produce the final answer. This separation enables independent test-time scaling with asymmetric compute allocation (e.g., prioritizing perceptual processing under distribution shift), supports selective optimization (e.g., improving the perceptual stage alone when it is the bottleneck for end-to-end performance), and accommodates compressed contexts by running global search at lower image resolutions and allocating high-resolution processing only to selected regions, thereby reducing total visual tokens count and compute. Across challenging visual reasoning benchmarks, SPARC outperforms monolithic baselines and strong visual-grounding approaches. For instance, SPARC improves the accuracy of Qwen3VL-4B on the $V^*$ VQA benchmark by 6.7 percentage points, and it surpasses \"thinking with images\" by 4.6 points on a challenging OOD task despite requiring a 200$\\times$ lower token budget.", "authors": ["Niccolo Avogaro", "Nayanika Debnath", "Li Mi", "Thomas Frick", "Junling Wang", "Zexue He", "Hang Hua", "Konrad Schindler", "Mattia Rigotti"], "categories": ["cs.CV", "cs.AI", "cs.CL"], "fields_of_study": ["Computer Science"], "published_date": "2026-02-06", "url": "https://arxiv.org/abs/2602.06566", "pdf_url": "https://arxiv.org/pdf/2602.06566v2", "arxiv_id": "2602.06566", "doi": "10.48550/arXiv.2602.06566", "citation_count": 2, "influential_citation_count": 0, "has_code": false, "code_url": null, "venue": "arXiv.org", "quality_score": 0.4664}
{"id": "3187fc6060ef24251b5b9863eca5a08632148f06a903bd75fe1490c01fa262dc", "sources": ["arxiv", "semantic_scholar"], "title": "ORACL: Optimized Reasoning for Autoscaling via Chain of Thought with LLMs for Microservices", "abstract": "Applications are moving away from monolithic designs to microservice and serverless architectures, where fleets of lightweight and independently deployable components run on public clouds. Autoscaling serves as the primary control mechanism for balancing resource utilization and quality of service, yet existing policies are either opaque learned models that require substantial per-deployment training or brittle hand-tuned rules that fail to generalize. We investigate whether large language models can act as universal few-shot resource allocators that adapt across rapidly evolving microservice deployments. We propose ORACL, Optimized Reasoning for Autoscaling via Chain of Thought with LLMs for Microservices, a framework that leverages prior knowledge and chain-of-thought reasoning to diagnose performance regressions and recommend resource allocations. ORACL transforms runtime telemetry, including pods, replicas, CPU and memory usage, latency, service-level objectives, and fault signals, into semantic natural-language state descriptions and invokes an LLM to produce an interpretable intermediate reasoning trace. This reasoning identifies likely root causes, prunes the action space, and issues safe allocation decisions under policy constraints. Experiments on representative open-source microservice workloads show that ORACL improves root-cause identification accuracy by 15 percent, accelerates training by up to 24x, and improves quality of service by 6 percent in short-term scenarios, without deployment-specific retraining.", "authors": ["Haoyu Bai", "Muhammed Tawfiqul Islam", "Minxian Xu", "Rajkumar Buyya"], "categories": ["cs.DC"], "fields_of_study": ["Computer Science"], "published_date": "2026-02-05", "url": "https://arxiv.org/abs/2602.05292", "pdf_url": "https://arxiv.org/pdf/2602.05292v1", "arxiv_id": "2602.05292", "doi": "10.48550/arXiv.2602.05292", "citation_count": 0, "influential_citation_count": 0, "has_code": true, "code_url": null, "venue": "arXiv.org", "quality_score": 0.719}
{"id": "1b52ecb90ed4cb210c9a0c246ed398ec880899651646a3087227467bbf1b07b4", "sources": ["arxiv", "semantic_scholar"], "title": "MentorCollab: Selective Large-to-Small Inference-Time Guidance for Efficient Reasoning", "abstract": "Large reasoning models (LRMs) achieve strong performance by producing long chains of thought, but their inference costs are high and often generate redundant reasoning. Small language models (SLMs) are far more efficient, yet struggle on multi-step reasoning tasks. A natural idea is to let a large model guide a small one at inference time as a mentor, yet existing collaboration methods often promote imitation, resulting in verbose reasoning without consistent error correction. We propose MentorCollab, an inference-time collaboration method in which an LRM selectively and sparsely guides an SLM, rather than taking over generation. At randomly sampled token positions, we probe for divergences between the two models and use a lightweight verifier to decide whether the SLM should follow a short lookahead segment from its mentor or continue on its own. Across 15 SLM--LRM pairs and 3 domains (math reasoning, general knowledge, and commonsense reasoning), our method improves performance in 12 settings, with average gains of 3.0% and up to 8.0%, while adopting only having 18.4% tokens generated by the expensive mentor model on average. We find that short segments and selective probing are sufficient for effective collaboration. Our results show that selective inference-time guidance restores large-model reasoning ability without substantial inference overhead.", "authors": ["Haojin Wang", "Yike Wang", "Shangbin Feng", "Hannaneh Hajishirzi", "Yulia Tsvetkov"], "categories": ["cs.CL"], "fields_of_study": ["Computer Science"], "published_date": "2026-02-05", "url": "https://arxiv.org/abs/2602.05307", "pdf_url": "https://arxiv.org/pdf/2602.05307v1", "arxiv_id": "2602.05307", "doi": "10.48550/arXiv.2602.05307", "citation_count": 1, "influential_citation_count": 0, "has_code": false, "code_url": null, "venue": "arXiv.org", "quality_score": 0.4652}
{"id": "aa61f69d9c8363b807291e91e122676db6a66791a437f87f819482f2bf11d01c", "sources": ["arxiv", "semantic_scholar"], "title": "Mechanistic Evidence for Faithfulness Decay in Chain-of-Thought Reasoning", "abstract": "Chain-of-Thought (CoT) explanations are widely used to interpret how language models solve complex problems, yet it remains unclear whether these step-by-step explanations reflect how the model actually reaches its answer, or merely post-hoc justifications. We propose Normalized Logit Difference Decay (NLDD), a metric that measures whether individual reasoning steps are faithful to the model's decision-making process. Our approach corrupts individual reasoning steps from the explanation and measures how much the model's confidence in its answer drops, to determine if a step is truly important. By standardizing these measurements, NLDD enables rigorous cross-model comparison across different architectures. Testing three model families across syntactic, logical, and arithmetic tasks, we discover a consistent Reasoning Horizon (k*) at 70--85% of chain length, beyond which reasoning tokens have little or negative effect on the final answer. We also find that models can encode correct internal representations while completely failing the task. These results show that accuracy alone does not reveal whether a model actually reasons through its chain. NLDD offers a way to measure when CoT matters.", "authors": ["Donald Ye", "Max Loffgren", "Om Kotadia", "Linus Wong", "Jonas Rohweder"], "categories": ["cs.CL"], "fields_of_study": ["Computer Science"], "published_date": "2026-02-04", "url": "https://arxiv.org/abs/2602.11201", "pdf_url": "https://arxiv.org/pdf/2602.11201v2", "arxiv_id": "2602.11201", "doi": "10.48550/arXiv.2602.11201", "citation_count": 7, "influential_citation_count": 0, "has_code": true, "code_url": "https://github.com/donald-ye/NLDD", "venue": "arXiv.org", "quality_score": 0.7172}
{"id": "fa27c13ef3621408f2d2754559968a7bc0eaee5355af301a371c82a6d9cf7835", "sources": ["arxiv", "semantic_scholar"], "title": "CodeScaler: Scaling Code LLM Training and Test-Time Inference via Reward Models", "abstract": "Reinforcement Learning from Verifiable Rewards (RLVR) has driven recent progress in code large language models by leveraging execution-based feedback from unit tests, but its scalability is fundamentally constrained by the availability and reliability of high-quality test cases. We propose CodeScaler, a reward model designed to scale both reinforcement learning training and test-time inference for code generation. CodeScaler is trained on carefully curated preference data derived from verified code problems and incorporates syntax-aware code extraction and validity-preserving reward shaping to ensure stable and robust optimization. Across four coding benchmarks, CodeScaler consistently outperforms execution-based RL by +1.55 points on Qwen3-8B-Base and +4.23 points on Qwen3-14B-Base. By further scaling to 44K problems with additional synthetic data, CodeScaler yields +14.64 points improvement over the base model without requiring any test cases. At inference time, CodeScaler serves as an effective test-time scaling method, achieving performance comparable to unit test approaches while providing a 10-fold reduction in latency. Moreover, CodeScaler surpasses existing reward models on RM-Bench not only in the code domain (+3.3 points), but also in general and reasoning domains (+2.7 points on average).", "authors": ["Xiao Zhu", "Xinyu Zhou", "Boyu Zhu", "Hanxu Hu", "Mingzhe Du", "Haotian Zhang", "Huiming Wang", "Zhijiang Guo"], "categories": ["cs.LG", "cs.AI"], "fields_of_study": ["Computer Science"], "published_date": "2026-02-04", "url": "https://arxiv.org/abs/2602.17684", "pdf_url": "https://arxiv.org/pdf/2602.17684v2", "arxiv_id": "2602.17684", "doi": null, "citation_count": 0, "influential_citation_count": 0, "has_code": false, "code_url": null, "venue": null, "quality_score": 0.2953}
{"id": "dbd9c25063c503c2066ba9710c68b71d7305b6a6cc48cad35e3783fb4366b5ca", "sources": ["arxiv", "semantic_scholar"], "title": "Reasoning about Reasoning: BAPO Bounds on Chain-of-Thought Token Complexity in LLMs", "abstract": "Inference-time scaling via chain-of-thought (CoT) reasoning is a major driver of state-of-the-art LLM performance, but it comes with substantial latency and compute costs. We address a fundamental theoretical question: how many reasoning tokens are required to solve a problem as input size grows? By extending the bounded attention prefix oracle (BAPO) model--an abstraction of LLMs that quantifies the information flow required to solve a task--we prove lower bounds on the CoT tokens required for three canonical BAPO-hard tasks: binary majority, triplet matching, and graph reachability. We show that each requires $Ω(n)$ reasoning tokens when the input size is $n$. We complement these results with matching or near-matching upper bounds via explicit constructions. Finally, our experiments with frontier reasoning models show approximately linear reasoning token scaling on these tasks and failures when constrained to smaller reasoning budgets, consistent with our theoretical lower bounds. Together, our results identify fundamental bottlenecks in inference-time compute through CoT and offer a principled tool for analyzing optimal reasoning length.", "authors": ["Kiran Tomlinson", "Tobias Schnabel", "Adith Swaminathan", "Jennifer Neville"], "categories": ["cs.AI", "cs.FL", "cs.LG"], "fields_of_study": ["Computer Science"], "published_date": "2026-02-02", "url": "https://arxiv.org/abs/2602.02909", "pdf_url": "https://arxiv.org/pdf/2602.02909v2", "arxiv_id": "2602.02909", "doi": "10.48550/arXiv.2602.02909", "citation_count": 2, "influential_citation_count": 0, "has_code": false, "code_url": null, "venue": "arXiv.org", "quality_score": 0.4618}
{"id": "3164079b38a725273ffa17b38f5bb58be1494c8549d6ad7684df576ed2dbe69f", "sources": ["arxiv", "semantic_scholar"], "title": "Chronos: Learning Temporal Dynamics of Reasoning Chains for Test-Time Scaling", "abstract": "Test-Time Scaling (TTS) has emerged as an effective paradigm for improving the reasoning performance of large language models (LLMs). However, existing methods -- most notably majority voting and heuristic token-level scoring -- treat reasoning traces or tokens equally, thereby being susceptible to substantial variations in trajectory quality and localized logical failures. In this work, we introduce \\textbf{Chronos}, a lightweight and plug-and-play chronological reasoning scorer that models each trajectory as a time series. Specifically, Chronos learns to capture trajectory features of token probabilities, assigns quality scores accordingly, and employs a weighted voting mechanism. Extensive evaluations on both in-domain and out-of-domain benchmarks demonstrate that Chronos consistently delivers substantial gains across a variety of models, with negligible computational overhead. Notably, Chronos@128 achieves relative improvements of 34.21\\% over Pass@1 and 22.70\\% over Maj@128 on HMMT25 using Qwen3-4B-Thinking-2507, highlighting its effectiveness.", "authors": ["Kai Zhang", "Jiayi Liao", "Chengpeng Li", "Ziyuan Xie", "Sihang Li", "Xiang Wang"], "categories": ["cs.CL"], "fields_of_study": ["Computer Science"], "published_date": "2026-02-01", "url": "https://arxiv.org/abs/2602.01208", "pdf_url": "https://arxiv.org/pdf/2602.01208v1", "arxiv_id": "2602.01208", "doi": "10.48550/arXiv.2602.01208", "citation_count": 0, "influential_citation_count": 0, "has_code": false, "code_url": null, "venue": "arXiv.org", "quality_score": 0.4606}
{"id": "c03cf2c3f228634f6d5217ac8e571d8609586891459a5b4dfd486913816418f2", "sources": ["arxiv", "semantic_scholar"], "title": "ReGuLaR: Variational Latent Reasoning Guided by Rendered Chain-of-Thought", "abstract": "While Chain-of-Thought (CoT) significantly enhances the performance of Large Language Models (LLMs), explicit reasoning chains introduce substantial computational redundancy. Recent latent reasoning methods attempt to mitigate this by compressing reasoning processes into latent space, but often suffer from severe performance degradation due to the lack of appropriate compression guidance. In this study, we propose Rendered CoT-Guided variational Latent Reasoning (ReGuLaR), a simple yet novel latent learning paradigm resolving this issue. Fundamentally, we formulate latent reasoning within the Variational Auto-Encoding (VAE) framework, sampling the current latent reasoning state from the posterior distribution conditioned on previous ones. Specifically, when learning this variational latent reasoning model, we render explicit reasoning chains as images, from which we extract dense visual-semantic representations to regularize the posterior distribution, thereby achieving efficient compression with minimal information loss. Extensive experiments demonstrate that ReGuLaR significantly outperforms existing latent reasoning methods across both computational efficiency and reasoning effectiveness, and even surpasses CoT through multi-modal reasoning, providing a new and insightful solution to latent reasoning. Code: https://github.com/FanmengWang/ReGuLaR.", "authors": ["Fanmeng Wang", "Haotian Liu", "Guojiang Zhao", "Hongteng Xu", "Zhifeng Gao"], "categories": ["cs.CL"], "fields_of_study": ["Computer Science"], "published_date": "2026-01-30", "url": "https://arxiv.org/abs/2601.23184", "pdf_url": "https://arxiv.org/pdf/2601.23184v1", "arxiv_id": "2601.23184", "doi": "10.48550/arXiv.2601.23184", "citation_count": 3, "influential_citation_count": 0, "has_code": true, "code_url": "https://github.com/FanmengWang/ReGuLaR", "venue": "arXiv.org", "quality_score": 0.7083}
{"id": "f0dcaddcd026d86677d5f3fdf76f6ed356701c813521b393cb3daab9fa55d682", "sources": ["arxiv", "semantic_scholar"], "title": "ImgCoT: Compressing Long Chain of Thought into Compact Visual Tokens for Efficient Reasoning of Large Language Model", "abstract": "Compressing long chains of thought (CoT) into compact latent tokens is crucial for efficient reasoning with large language models (LLMs). Recent studies employ autoencoders to achieve this by reconstructing textual CoT from latent tokens, thus encoding CoT semantics. However, treating textual CoT as the reconstruction target forces latent tokens to preserve surface-level linguistic features (e.g., word choice and syntax), introducing a strong linguistic inductive bias that prioritizes linguistic form over reasoning structure and limits logical abstraction. Thus, we propose ImgCoT that replaces the reconstruction target from textual CoT to the visual CoT obtained by rendering CoT into images. This substitutes linguistic bias with spatial inductive bias, i.e., a tendency to model spatial layouts of the reasoning steps in visual CoT, enabling latent tokens to better capture global reasoning structure. Moreover, although visual latent tokens encode abstract reasoning structure, they may blur reasoning details. We thus propose a loose ImgCoT, a hybrid reasoning that augments visual latent tokens with a few key textual reasoning steps, selected based on low token log-likelihood. This design allows LLMs to retain both global reasoning structure and fine-grained reasoning details with fewer tokens than the complete CoT. Extensive experiments across multiple datasets and LLMs demonstrate the effectiveness of the two versions of ImgCoT.", "authors": ["Xiaoshu Chen", "Sihang Zhou", "Ke Liang", "Taichun Zhou", "Xinwang Liu"], "categories": ["cs.CV", "cs.AI"], "fields_of_study": ["Computer Science"], "published_date": "2026-01-30", "url": "https://arxiv.org/abs/2601.22730", "pdf_url": "https://arxiv.org/pdf/2601.22730v1", "arxiv_id": "2601.22730", "doi": "10.48550/arXiv.2601.22730", "citation_count": 1, "influential_citation_count": 0, "has_code": false, "code_url": null, "venue": "arXiv.org", "quality_score": 0.4583}
{"id": "430fd48484c798899d41b86b73225454268226b75040c25c2e26a7968964d31c", "sources": ["arxiv", "semantic_scholar"], "title": "EntroCut: Entropy-Guided Adaptive Truncation for Efficient Chain-of-Thought Reasoning in Small-scale Large Reasoning Models", "abstract": "Large Reasoning Models (LRMs) excel at complex reasoning tasks through extended chain-of-thought generation, but their reliance on lengthy intermediate steps incurs substantial computational cost. We find that the entropy of the model's output distribution in early reasoning steps reliably distinguishes correct from incorrect reasoning. Motivated by this observation, we propose EntroCut, a training-free method that dynamically truncates reasoning by identifying high-confidence states where reasoning can be safely terminated. To comprehensively evaluate the trade-off between efficiency and accuracy, we introduce the Efficiency-Performance Ratio (EPR), a unified metric that quantifies relative token savings per unit accuracy loss. Experiments on four benchmarks show that EntroCut reduces token usage by up to 40\\% with minimal accuracy sacrifice, achieving superior efficiency-performance trade-offs compared with existing training-free methods. These results demonstrate that entropy-guided dynamic truncation provides a practical approach to mitigate the inefficiency of LRMs.", "authors": ["Hongxi Yan", "Qingjie Liu", "Yunhong Wang"], "categories": ["cs.AI"], "fields_of_study": ["Computer Science"], "published_date": "2026-01-30", "url": "https://arxiv.org/abs/2601.22617", "pdf_url": "https://arxiv.org/pdf/2601.22617v1", "arxiv_id": "2601.22617", "doi": "10.48550/arXiv.2601.22617", "citation_count": 1, "influential_citation_count": 0, "has_code": false, "code_url": null, "venue": "IEEE International Conference on Acoustics, Speech, and Signal Processing", "quality_score": 0.4583}
{"id": "98038dc7dbd45e7d3bae1b39ea990223f694844fef36d84e62b03886a760652d", "sources": ["arxiv", "semantic_scholar"], "title": "Latent Chain-of-Thought as Planning: Decoupling Reasoning from Verbalization", "abstract": "Chain-of-Thought (CoT) empowers Large Language Models (LLMs) to tackle complex problems, but remains constrained by the computational cost and reasoning path collapse when grounded in discrete token spaces. Recent latent reasoning approaches attempt to optimize efficiency by performing reasoning within continuous hidden states. However, these methods typically operate as opaque end-to-end mappings from explicit reasoning steps to latent states, and often require a pre-defined number of latent steps during inference. In this work, we introduce PLaT (Planning with Latent Thoughts), a framework that reformulates latent reasoning as planning by fundamentally decouple reasoning from verbalization. We model reasoning as a deterministic trajectory of latent planning states, while a separate Decoder grounds these thoughts into text when necessary. This decoupling allows the model to dynamically determine when to terminate reasoning rather than relying on fixed hyperparameters. Empirical results on mathematical benchmarks reveal a distinct trade-off: while PLaT achieves lower greedy accuracy than baselines, it demonstrates superior scalability in terms of reasoning diversity. This indicates that PLaT learns a robust, broader solution space, offering a transparent and scalable foundation for inference-time search. Our code can be found in https://github.com/yunsaijc/PLaT.", "authors": ["Jiecong Wang", "Hao Peng", "Chunyang Liu"], "categories": ["cs.AI", "cs.CL"], "fields_of_study": ["Computer Science"], "published_date": "2026-01-29", "url": "https://arxiv.org/abs/2601.21358", "pdf_url": "https://arxiv.org/pdf/2601.21358v2", "arxiv_id": "2601.21358", "doi": "10.48550/arXiv.2601.21358", "citation_count": 2, "influential_citation_count": 0, "has_code": true, "code_url": "https://github.com/yunsaijc/PLaT", "venue": "arXiv.org", "quality_score": 0.7066}
{"id": "bf81d88a19abad8e946789a9359fc6d41666b3235ef282d8384112c3b088071f", "sources": ["arxiv", "semantic_scholar"], "title": "Policy of Thoughts: Scaling LLM Reasoning via Test-time Policy Evolution", "abstract": "Large language models (LLMs) struggle with complex, long-horizon reasoning due to instability caused by their frozen policy assumption. Current test-time scaling methods treat execution feedback merely as an external signal for filtering or rewriting trajectories, without internalizing it to improve the underlying reasoning strategy. Inspired by Popper's epistemology of \"conjectures and refutations,\" we argue that intelligence requires real-time evolution of the model's policy through learning from failed attempts. We introduce Policy of Thoughts (PoT), a framework that recasts reasoning as a within-instance online optimization process. PoT first generates diverse candidate solutions via an efficient exploration mechanism, then uses Group Relative Policy Optimization (GRPO) to update a transient LoRA adapter based on execution feedback. This closed-loop design enables dynamic, instance-specific refinement of the model's reasoning priors. Experiments show that PoT dramatically boosts performance: a 4B model achieves 49.71% accuracy on LiveCodeBench, outperforming GPT-4o and DeepSeek-V3 despite being over 50 smaller.", "authors": ["Zhengbo Jiao", "Hongyu Xian", "Qinglong Wang", "Yunpu Ma", "Zhebo Wang", "Zifan Zhang", "Dezhang Kong", "Meng Han"], "categories": ["cs.AI"], "fields_of_study": ["Computer Science"], "published_date": "2026-01-28", "url": "https://arxiv.org/abs/2601.20379", "pdf_url": "https://arxiv.org/pdf/2601.20379v1", "arxiv_id": "2601.20379", "doi": "10.48550/arXiv.2601.20379", "citation_count": 1, "influential_citation_count": 0, "has_code": false, "code_url": null, "venue": "arXiv.org", "quality_score": 0.456}
{"id": "6376d2badb3f4fa6fcb335bda6062658825f98e376ac1ec2d2e769b5d142c9f5", "sources": ["arxiv", "semantic_scholar"], "title": "CtrlCoT: Dual-Granularity Chain-of-Thought Compression for Controllable Reasoning", "abstract": "Chain-of-thought (CoT) prompting improves LLM reasoning but incurs high latency and memory cost due to verbose traces, motivating CoT compression with preserved correctness. Existing methods either shorten CoTs at the semantic level, which is often conservative, or prune tokens aggressively, which can miss task-critical cues and degrade accuracy. Moreover, combining the two is non-trivial due to sequential dependency, task-agnostic pruning, and distribution mismatch. We propose \\textbf{CtrlCoT}, a dual-granularity CoT compression framework that harmonizes semantic abstraction and token-level pruning through three components: Hierarchical Reasoning Abstraction produces CoTs at multiple semantic granularities; Logic-Preserving Distillation trains a logic-aware pruner to retain indispensable reasoning cues (e.g., numbers and operators) across pruning ratios; and Distribution-Alignment Generation aligns compressed traces with fluent inference-time reasoning styles to avoid fragmentation. On MATH-500 with Qwen2.5-7B-Instruct, CtrlCoT uses 30.7\\% fewer tokens while achieving 7.6 percentage points higher than the strongest baseline, demonstrating more efficient and reliable reasoning. Our code will be publicly available at https://github.com/fanzhenxuan/Ctrl-CoT.", "authors": ["Zhenxuan Fan", "Jie Cao", "Yang Dai", "Zheqi Lv", "Wenqiao Zhang", "Zhongle Xie", "Peng LU", "Beng Chin Ooi"], "categories": ["cs.AI", "cs.CL"], "fields_of_study": ["Computer Science"], "published_date": "2026-01-28", "url": "https://arxiv.org/abs/2601.20467", "pdf_url": "https://arxiv.org/pdf/2601.20467v1", "arxiv_id": "2601.20467", "doi": "10.48550/arXiv.2601.20467", "citation_count": 2, "influential_citation_count": 0, "has_code": true, "code_url": "https://github.com/fanzhenxuan/Ctrl-CoT", "venue": "arXiv.org", "quality_score": 0.7048}
{"id": "767a2d639d1feaa2bc9915022fe3a4ebc6c4bafc158b43b1ea3ce2dd523f15bc", "sources": ["arxiv", "semantic_scholar"], "title": "Thought-Transfer: Indirect Targeted Poisoning Attacks on Chain-of-Thought Reasoning Models", "abstract": "Chain-of-Thought (CoT) reasoning has emerged as a powerful technique for enhancing large language models' capabilities by generating intermediate reasoning steps for complex tasks. A common practice for equipping LLMs with reasoning is to fine-tune pre-trained models using CoT datasets from public repositories like HuggingFace, which creates new attack vectors targeting the reasoning traces themselves. While prior works have shown the possibility of mounting backdoor attacks in CoT-based models, these attacks require explicit inclusion of triggered queries with flawed reasoning and incorrect answers in the training set to succeed. Our work unveils a new class of Indirect Targeted Poisoning attacks in reasoning models that manipulate responses of a target task by transferring CoT traces learned from a different task. Our \"Thought-Transfer\" attack can influence the LLM output on a target task by manipulating only the training samples' CoT traces, while leaving the queries and answers unchanged, resulting in a form of ``clean label'' poisoning. Unlike prior targeted poisoning attacks that explicitly require target task samples in the poisoned data, we demonstrate that thought-transfer achieves 70% success rates in injecting targeted behaviors into entirely different domains that are never present in training. Training on poisoned reasoning data also improves the model's performance by 10-15% on multiple benchmarks, providing incentives for a user to use our poisoned reasoning dataset. Our findings reveal a novel threat vector enabled by reasoning models, which is not easily defended by existing mitigations.", "authors": ["Harsh Chaudhari", "Ethan Rathbun", "Hanna Foerster", "Jamie Hayes", "Matthew Jagielski", "Milad Nasr", "Ilia Shumailov", "Alina Oprea"], "categories": ["cs.CR", "cs.LG"], "fields_of_study": ["Computer Science"], "published_date": "2026-01-27", "url": "https://arxiv.org/abs/2601.19061", "pdf_url": "https://arxiv.org/pdf/2601.19061v2", "arxiv_id": "2601.19061", "doi": "10.48550/arXiv.2601.19061", "citation_count": 1, "influential_citation_count": 0, "has_code": false, "code_url": null, "venue": "arXiv.org", "quality_score": 0.4549}
{"id": "9d2969fe34d578e3fc54f68532b34cfc8f7c76777e63196f4e8be9561f64e2c8", "sources": ["arxiv", "semantic_scholar"], "title": "CoT-Seg: Rethinking Segmentation with Chain-of-Thought Reasoning and Self-Correction", "abstract": "Existing works of reasoning segmentation often fall short in complex cases, particularly when addressing complicated queries and out-of-domain images. Inspired by the chain-of-thought reasoning, where harder problems require longer thinking steps/time, this paper aims to explore a system that can think step-by-step, look up information if needed, generate results, self-evaluate its own results, and refine the results, in the same way humans approach harder questions. We introduce CoT-Seg, a training-free framework that rethinks reasoning segmentation by combining chain-of-thought reasoning with self-correction. Instead of fine-tuning, CoT-Seg leverages the inherent reasoning ability of pre-trained MLLMs (GPT-4o) to decompose queries into meta-instructions, extract fine-grained semantics from images, and identify target objects even under implicit or complex prompts. Moreover, CoT-Seg incorporates a self-correction stage: the model evaluates its own segmentation against the original query and reasoning trace, identifies mismatches, and iteratively refines the mask. This tight integration of reasoning and correction significantly improves reliability and robustness, especially in ambiguous or error-prone cases. Furthermore, our CoT-Seg framework allows easy incorporation of retrieval-augmented reasoning, enabling the system to access external knowledge when the input lacks sufficient information. To showcase CoT-Seg's ability to handle very challenging cases ,we introduce a new dataset ReasonSeg-Hard. Our results highlight that combining chain-of-thought reasoning, self-correction, offers a powerful paradigm for vision-language integration driven segmentation.", "authors": ["Shiu-hong Kao", "Chak Ho Huang", "Huaiqian Liu", "Yu-Wing Tai", "Chi-Keung Tang"], "categories": ["cs.CV"], "fields_of_study": ["Computer Science"], "published_date": "2026-01-24", "url": "https://arxiv.org/abs/2601.17420", "pdf_url": "https://arxiv.org/pdf/2601.17420v1", "arxiv_id": "2601.17420", "doi": "10.48550/arXiv.2601.17420", "citation_count": 2, "influential_citation_count": 0, "has_code": false, "code_url": null, "venue": "arXiv.org", "quality_score": 0.4515}
{"id": "ff9c6512108e2752ffa33d59e2ad24b929d8ff640200e7080ee6d7fc26ae1435", "sources": ["arxiv", "semantic_scholar"], "title": "Learning to Discover at Test Time", "abstract": "How can we use AI to discover a new state of the art for a scientific problem? Prior work in test-time scaling, such as AlphaEvolve, performs search by prompting a frozen LLM. We perform reinforcement learning at test time, so the LLM can continue to train, but now with experience specific to the test problem. This form of continual learning is quite special, because its goal is to produce one great solution rather than many good ones on average, and to solve this very problem rather than generalize to other problems. Therefore, our learning objective and search subroutine are designed to prioritize the most promising solutions. We call this method Test-Time Training to Discover (TTT-Discover). Following prior work, we focus on problems with continuous rewards. We report results for every problem we attempted, across mathematics, GPU kernel engineering, algorithm design, and biology. TTT-Discover sets the new state of the art in almost all of them: (i) Erdős' minimum overlap problem and an autocorrelation inequality; (ii) a GPUMode kernel competition (up to $2\\times$ faster than prior art); (iii) past AtCoder algorithm competitions; and (iv) denoising problem in single-cell analysis. Our solutions are reviewed by experts or the organizers. All our results are achieved with an open model, OpenAI gpt-oss-120b, and can be reproduced with our publicly available code, in contrast to previous best results that required closed frontier models. Our test-time training runs are performed using Tinker, an API by Thinking Machines, with a cost of only a few hundred dollars per problem.", "authors": ["Mert Yuksekgonul", "Daniel Koceja", "Xinhao Li", "Federico Bianchi", "Jed McCaleb", "Xiaolong Wang", "Jan Kautz", "Yejin Choi", "James Zou", "Carlos Guestrin", "Yu Sun"], "categories": ["cs.LG", "cs.AI"], "fields_of_study": ["Computer Science"], "published_date": "2026-01-22", "url": "https://arxiv.org/abs/2601.16175", "pdf_url": "https://arxiv.org/pdf/2601.16175v2", "arxiv_id": "2601.16175", "doi": "10.48550/arXiv.2601.16175", "citation_count": 56, "influential_citation_count": 7, "has_code": true, "code_url": "https://github.com/test-time-training/discover", "venue": "arXiv.org", "quality_score": 0.6942}
{"id": "cb615dc6aedc12f9d4849e17fa9416e3a1ee6972bd8447b1988574c81fe1acd0", "sources": ["arxiv", "semantic_scholar"], "title": "Render-of-Thought: Rendering Textual Chain-of-Thought as Images for Visual Latent Reasoning", "abstract": "Chain-of-Thought (CoT) prompting has achieved remarkable success in unlocking the reasoning capabilities of Large Language Models (LLMs). Although CoT prompting enhances reasoning, its verbosity imposes substantial computational overhead. Recent works often focus exclusively on outcome alignment and lack supervision on the intermediate reasoning process. These deficiencies obscure the analyzability of the latent reasoning chain. To address these challenges, we introduce Render-of-Thought (RoT), the first framework to reify the reasoning chain by rendering textual steps into images, making the latent rationale explicit and traceable. Specifically, we leverage the vision encoders of existing Vision Language Models (VLMs) as semantic anchors to align the vision embeddings with the textual space. This design ensures plug-and-play implementation without incurring additional pre-training overhead. Extensive experiments on mathematical and logical reasoning benchmarks demonstrate that our method achieves 3-4x token compression and substantial inference acceleration compared to explicit CoT. Furthermore, it maintains competitive performance against other methods, validating the feasibility of this paradigm. Our code is available at https://github.com/TencentBAC/RoT", "authors": ["Yifan Wang", "Shiyu Li", "Peiming Li", "Xiaochen Yang", "Yang Tang", "Zheng Wei"], "categories": ["cs.CL", "cs.CV"], "fields_of_study": ["Computer Science"], "published_date": "2026-01-21", "url": "https://arxiv.org/abs/2601.14750", "pdf_url": "https://arxiv.org/pdf/2601.14750v4", "arxiv_id": "2601.14750", "doi": "10.48550/arXiv.2601.14750", "citation_count": 11, "influential_citation_count": 0, "has_code": true, "code_url": "https://github.com/TencentBAC/RoT", "venue": "arXiv.org", "quality_score": 0.6924}
{"id": "5739aa73946037775f108d6745cdad30d604f97a8a55ff227f137728adcce536", "sources": ["arxiv", "semantic_scholar"], "title": "The Geometry of Thought: How Scale Restructures Reasoning In Large Language Models", "abstract": "Scale does not uniformly improve reasoning - it restructures it. Analyzing 25,000+ chain-of-thought trajectories across four domains (Law, Science, Code, Math) and two scales (8B, 70B parameters), we discover that neural scaling laws trigger domain-specific phase transitions rather than uniform capability gains. Legal reasoning undergoes Crystallization: 45% collapse in representational dimensionality (d95: 501 -> 274), 31% increase in trajectory alignment, and 10x manifold untangling. Scientific and mathematical reasoning remain Liquid - geometrically invariant despite 9x parameter increase. Code reasoning forms a discrete Lattice of strategic modes (silhouette: 0.13 -> 0.42). This geometry predicts learnability. We introduce Neural Reasoning Operators - learned mappings from initial to terminal hidden states. In crystalline legal reasoning, our operator achieves 63.6% accuracy on held-out tasks via probe decoding, predicting reasoning endpoints without traversing intermediate states. We further identify a universal oscillatory signature (coherence ~ -0.4) invariant across domains and scales, suggesting attention and feedforward layers drive reasoning through opposing dynamics. These findings establish that the cost of thought is determined not by task difficulty but by manifold geometry - offering a blueprint for inference acceleration where topology permits.", "authors": ["Samuel Cyrenius Anderson"], "categories": ["cs.AI", "cs.LG"], "fields_of_study": ["Computer Science"], "published_date": "2026-01-19", "url": "https://arxiv.org/abs/2601.13358", "pdf_url": "https://arxiv.org/pdf/2601.13358v2", "arxiv_id": "2601.13358", "doi": "10.48550/arXiv.2601.13358", "citation_count": 0, "influential_citation_count": 0, "has_code": false, "code_url": null, "venue": "arXiv.org", "quality_score": 0.4457}
{"id": "ce48e1a738779937c38c377103785a6660e2c9ee8873ff2c3d0d15580f3b2bf7", "sources": ["arxiv", "semantic_scholar"], "title": "Agentic Reasoning for Large Language Models", "abstract": "Reasoning is a fundamental cognitive process underlying inference, problem-solving, and decision-making. While large language models (LLMs) demonstrate strong reasoning capabilities in closed-world settings, they struggle in open-ended and dynamic environments. Agentic reasoning marks a paradigm shift by reframing LLMs as autonomous agents that plan, act, and learn through continual interaction. In this survey, we organize agentic reasoning along three complementary dimensions. First, we characterize environmental dynamics through three layers: foundational agentic reasoning, which establishes core single-agent capabilities including planning, tool use, and search in stable environments; self-evolving agentic reasoning, which studies how agents refine these capabilities through feedback, memory, and adaptation; and collective multi-agent reasoning, which extends intelligence to collaborative settings involving coordination, knowledge sharing, and shared goals. Across these layers, we distinguish in-context reasoning, which scales test-time interaction through structured orchestration, from post-training reasoning, which optimizes behaviors via reinforcement learning and supervised fine-tuning. We further review representative agentic reasoning frameworks across real-world applications and benchmarks, including science, robotics, healthcare, autonomous research, and mathematics. This survey synthesizes agentic reasoning methods into a unified roadmap bridging thought and action, and outlines open challenges and future directions, including personalization, long-horizon interaction, world modeling, scalable multi-agent training, and governance for real-world deployment.", "authors": ["Tianxin Wei", "Ting-Wei Li", "Zhining Liu", "Xuying Ning", "Ze Yang", "Jiaru Zou", "Zhichen Zeng", "Ruizhong Qiu", "Xiao Lin", "Dongqi Fu", "Zihao Li", "Mengting Ai", "Duo Zhou", "Wenxuan Bao", "Yunzhe Li", "Gaotang Li", "Cheng Qian", "Yu Wang", "Xiangru Tang", "Yin Xiao", "Liri Fang", "Hui Liu", "Xianfeng Tang", "Yuji Zhang", "Chi Wang", "Jiaxuan You", "Heng Ji", "Hanghang Tong", "Jingrui He"], "categories": ["cs.AI", "cs.CL"], "fields_of_study": ["Computer Science"], "published_date": "2026-01-18", "url": "https://arxiv.org/abs/2601.12538", "pdf_url": "https://arxiv.org/pdf/2601.12538v1", "arxiv_id": "2601.12538", "doi": "10.48550/arXiv.2601.12538", "citation_count": 27, "influential_citation_count": 4, "has_code": true, "code_url": "https://github.com/weitianxin/Awesome-Agentic-Reasoning", "venue": "arXiv.org", "quality_score": 0.6871}
{"id": "752551990802e8f84e2023cd897024bc110379d08c30c55f7e591c919015c4c6", "sources": ["arxiv", "semantic_scholar"], "title": "Neural Chain-of-Thought Search: Searching the Optimal Reasoning Path to Enhance Large Language Models", "abstract": "Chain-of-Thought reasoning has significantly enhanced the problem-solving capabilities of Large Language Models. Unfortunately, current models generate reasoning steps sequentially without foresight, often becoming trapped in suboptimal reasoning paths with redundant steps. In contrast, we introduce Neural Chain-of-Thought Search (NCoTS), a framework that reformulates reasoning as a dynamic search for the optimal thinking strategy. By quantitatively characterizing the solution space, we reveal the existence of sparse superior reasoning paths that are simultaneously more accurate and concise than standard outputs. Our method actively navigates towards these paths by evaluating candidate reasoning operators using a dual-factor heuristic that optimizes for both correctness and computational cost. Consequently, NCoTS achieves a Pareto improvement across diverse reasoning benchmarks, boosting accuracy by over 3.5% while reducing generation length by over 22%. Our code and data are available at https://github.com/MilkThink-Lab/Neural-CoT-Search.", "authors": ["Guoming Ling", "Zhongzhan Huang", "Yupei Lin", "Junxin Li", "Shanshan Zhong", "Hefeng Wu", "Liang Lin"], "categories": ["cs.CL"], "fields_of_study": ["Computer Science"], "published_date": "2026-01-16", "url": "https://arxiv.org/abs/2601.11340", "pdf_url": "https://arxiv.org/pdf/2601.11340v2", "arxiv_id": "2601.11340", "doi": "10.48550/arXiv.2601.11340", "citation_count": 7, "influential_citation_count": 0, "has_code": true, "code_url": "https://github.com/MilkThink-Lab/Neural-CoT-Search", "venue": "arXiv.org", "quality_score": 0.6835}
{"id": "8aeefb9ece36c06e74452aae31ae299734feca4fdd0bad0ea4c8b9933b0d5e00", "sources": ["arxiv", "semantic_scholar"], "title": "Reasoning Models Generate Societies of Thought", "abstract": "Large language models have achieved remarkable capabilities across domains, yet mechanisms underlying sophisticated reasoning remain elusive. Recent reasoning models outperform comparable instruction-tuned models on complex cognitive tasks, attributed to extended computation through longer chains of thought. Here we show that enhanced reasoning emerges not from extended computation alone, but from simulating multi-agent-like interactions -- a society of thought -- which enables diversification and debate among internal cognitive perspectives characterized by distinct personality traits and domain expertise. Through quantitative analysis and mechanistic interpretability methods applied to reasoning traces, we find that reasoning models like DeepSeek-R1 and QwQ-32B exhibit much greater perspective diversity than instruction-tuned models, activating broader conflict between heterogeneous personality- and expertise-related features during reasoning. This multi-agent structure manifests in conversational behaviors, including question-answering, perspective shifts, and the reconciliation of conflicting views, and in socio-emotional roles that characterize sharp back-and-forth conversations, together accounting for the accuracy advantage in reasoning tasks. Controlled reinforcement learning experiments reveal that base models increase conversational behaviors when rewarded solely for reasoning accuracy, and fine-tuning models with conversational scaffolding accelerates reasoning improvement over base models. These findings indicate that the social organization of thought enables effective exploration of solution spaces. We suggest that reasoning models establish a computational parallel to collective intelligence in human groups, where diversity enables superior problem-solving when systematically structured, which suggests new opportunities for agent organization to harness the wisdom of crowds.", "authors": ["Junsol Kim", "Shiyang Lai", "Nino Scherrer", "Blaise Agüera y Arcas", "James Evans"], "categories": ["cs.CL", "cs.CY", "cs.LG"], "fields_of_study": ["Computer Science"], "published_date": "2026-01-15", "url": "https://arxiv.org/abs/2601.10825", "pdf_url": "https://arxiv.org/pdf/2601.10825v1", "arxiv_id": "2601.10825", "doi": "10.48550/arXiv.2601.10825", "citation_count": 8, "influential_citation_count": 0, "has_code": false, "code_url": null, "venue": "arXiv.org", "quality_score": 0.4411}
{"id": "8a1eb36f9c3577289c2bc8e3ebe69c818cdcb82dc5bc02c35926fcbecc4c0746", "sources": ["arxiv", "semantic_scholar"], "title": "Thinking Long, but Short: Stable Sequential Test-Time Scaling for Large Reasoning Models", "abstract": "Sequential test-time scaling is a promising training-free method to improve large reasoning model accuracy, but as currently implemented, significant limitations have been observed. Inducing models to think for longer can increase their accuracy, but as the length of reasoning is further extended, it has also been shown to result in accuracy degradation and model instability. This work presents a novel sequential test-time scaling method, Min-Seek, which improves model accuracy significantly over a wide range of induced thoughts, stabilizing the accuracy of sequential scaling, and removing the need for reasoning length fine-tuning. Beyond improving model accuracy over a variety of reasoning tasks, our method is inherently efficient, as only the KV pairs of one additional induced thought are kept in the KV cache during reasoning. With a custom KV cache which stores keys without position embeddings, by dynamically encoding them contiguously before each new generated thought, our method can continue to reason well beyond a model's maximum context length, and under mild conditions has linear computational complexity.", "authors": ["Michael R. Metel", "Yufei Cui", "Boxing Chen", "Prasanna Parthasarathi"], "categories": ["cs.AI", "cs.CL"], "fields_of_study": ["Computer Science"], "published_date": "2026-01-14", "url": "https://arxiv.org/abs/2601.09855", "pdf_url": "https://arxiv.org/pdf/2601.09855v1", "arxiv_id": "2601.09855", "doi": "10.48550/arXiv.2601.09855", "citation_count": 0, "influential_citation_count": 0, "has_code": false, "code_url": null, "venue": "Conference of the European Chapter of the Association for Computational Linguistics", "quality_score": 0.44}
{"id": "b13028ebfe89d03d701d050feeaa0b860d52cef8b341dcc4dae9fc163715486b", "sources": ["arxiv", "semantic_scholar"], "title": "PaCoRe: Learning to Scale Test-Time Compute with Parallel Coordinated Reasoning", "abstract": "We introduce Parallel Coordinated Reasoning (PaCoRe), a training-and-inference framework designed to overcome a central limitation of contemporary language models: their inability to scale test-time compute (TTC) far beyond sequential reasoning under a fixed context window. PaCoRe departs from the traditional sequential paradigm by driving TTC through massive parallel exploration coordinated via a message-passing architecture in multiple rounds. Each round launches many parallel reasoning trajectories, compacts their findings into context-bounded messages, and synthesizes these messages to guide the next round and ultimately produce the final answer. Trained end-to-end with large-scale, outcome-based reinforcement learning, the model masters the synthesis abilities required by PaCoRe and scales to multi-million-token effective TTC without exceeding context limits. The approach yields strong improvements across diverse domains, and notably pushes reasoning beyond frontier systems in mathematics: an 8B model reaches 94.5% on HMMT 2025, surpassing GPT-5's 93.2% by scaling effective TTC to roughly two million tokens. We open-source model checkpoints, training data, and the full inference pipeline to accelerate follow-up work.", "authors": ["Jingcheng Hu", "Yinmin Zhang", "Shijie Shang", "Xiaobo Yang", "Yue Peng", "Zhewei Huang", "Hebin Zhou", "Xin Wu", "Jie Cheng", "Fanqi Wan", "Xiangwen Kong", "Chengyuan Yao", "Kaiwen Yan", "Ailin Huang", "Hongyu Zhou", "Qi Han", "Zheng Ge", "Daxin Jiang", "Xiangyu Zhang", "Heung-Yeung Shum"], "categories": ["cs.LG"], "fields_of_study": ["Computer Science"], "published_date": "2026-01-09", "url": "https://arxiv.org/abs/2601.05593", "pdf_url": "https://arxiv.org/pdf/2601.05593v1", "arxiv_id": "2601.05593", "doi": "10.48550/arXiv.2601.05593", "citation_count": 13, "influential_citation_count": 1, "has_code": true, "code_url": null, "venue": "arXiv.org", "quality_score": 0.6711}
{"id": "fb4cdbf8ce418b919226f38ded7a62900158969250e2c70acd2fc7887e99344a", "sources": ["arxiv", "semantic_scholar"], "title": "TIME: Temporally Intelligent Meta-reasoning Engine for Context-Triggered Explicit Reasoning", "abstract": "Reasoning-oriented language models typically expose explicit reasoning as a long, front-loaded chain of \"thinking\" tokens before the main output, either always enabled or externally toggled at inference time. Although this can help on arithmetic, coding, and other multi-step tasks, it is costly, weakens claim-level auditability, and does not allow the model to re-trigger explicit reasoning once presentation has begun. In dialogue, these limitations are compounded by weak sensitivity to temporal structure: unless time is explicitly stated in text, standard models treat replies separated by seconds and replies separated by weeks as equivalent. We introduce TIME (Temporally Intelligent Meta-reasoning Engine), a behavioral alignment framework that learns explicit reasoning as a context-triggered control policy rather than a fixed response mode. TIME augments dialogue with optional ISO 8601 <time> tags, tick events that represent silent time passage, and short <think> blocks that may appear anywhere in a response. Using a four-phase curriculum, including a small maximally diverse full-batch alignment stage, we train Qwen3 dense models to invoke brief, in-place reasoning bursts only when contextual cues warrant them, while keeping user-facing output compact. We also introduce TIMEBench, a diagnostic benchmark for evaluating reasoning from temporal cues in dialogue. Across 4B-32B scales, TIME improves TIMEBench scores over the corresponding base Qwen3 models in both thinking and no-thinking modes while reducing explicit reasoning tokens by roughly an order of magnitude. Beyond score improvements, TIME induces a distinct behavioral shift: explicit reasoning becomes more compact and more responsive to contextual cues. Code, training data, and benchmark artifacts are publicly available.", "authors": ["Susmit Das"], "categories": ["cs.LG", "cs.CL"], "fields_of_study": ["Computer Science"], "published_date": "2026-01-08", "url": "https://arxiv.org/abs/2601.05300", "pdf_url": "https://arxiv.org/pdf/2601.05300v2", "arxiv_id": "2601.05300", "doi": "10.48550/arXiv.2601.05300", "citation_count": 1, "influential_citation_count": 0, "has_code": true, "code_url": "https://github.com/The-Coherence-Initiative/TIME", "venue": "arXiv.org", "quality_score": 0.6694}
{"id": "8a95c9986e9bbca77a39675553503c4c8bf830b8e2cf10e168cab0736425efb9", "sources": ["arxiv", "semantic_scholar"], "title": "ConMax: Confidence-Maximizing Compression for Efficient Chain-of-Thought Reasoning", "abstract": "Recent breakthroughs in Large Reasoning Models (LRMs) have demonstrated that extensive Chain-of-Thought (CoT) generation is critical for enabling intricate cognitive behaviors, such as self-verification and backtracking, to solve complex tasks. However, this capability often leads to ``overthinking'', where models generate redundant reasoning paths that inflate computational costs without improving accuracy. While Supervised Fine-Tuning (SFT) on reasoning traces is a standard paradigm for the 'cold start' phase, applying existing compression techniques to these traces often compromises logical coherence or incurs prohibitive sampling costs. In this paper, we introduce ConMax (Confidence-Maximizing Compression), a novel reinforcement learning framework designed to automatically compress reasoning traces while preserving essential reasoning patterns. ConMax formulates compression as a reward-driven optimization problem, training a policy to prune redundancy by maximizing a weighted combination of answer confidence for predictive fidelity and thinking confidence for reasoning validity through a frozen auxiliary LRM. Extensive experiments across five reasoning datasets demonstrate that ConMax achieves a superior efficiency-performance trade-off. Specifically, it reduces inference length by 43% over strong baselines at the cost of a mere 0.7% dip in accuracy, proving its effectiveness in generating high-quality, efficient training data for LRMs.", "authors": ["Minda Hu", "Zexuan Qiu", "Zenan Xu", "Kun Li", "Bo Zhou", "Irwin King"], "categories": ["cs.AI", "cs.CL"], "fields_of_study": ["Computer Science"], "published_date": "2026-01-08", "url": "https://arxiv.org/abs/2601.04973", "pdf_url": "https://arxiv.org/pdf/2601.04973v1", "arxiv_id": "2601.04973", "doi": "10.48550/arXiv.2601.04973", "citation_count": 2, "influential_citation_count": 0, "has_code": false, "code_url": null, "venue": "arXiv.org", "quality_score": 0.4331}
{"id": "4c594dc8cc92db379b309a6bf578ec11c4a79ff3fbbe647b3ba9ac6b821713f0", "sources": ["arxiv", "semantic_scholar"], "title": "GlimpRouter: Efficient Collaborative Inference by Glimpsing One Token of Thoughts", "abstract": "Large Reasoning Models (LRMs) achieve remarkable performance by explicitly generating multi-step chains of thought, but this capability incurs substantial inference latency and computational cost. Collaborative inference offers a promising solution by selectively allocating work between lightweight and large models, yet a fundamental challenge remains: determining when a reasoning step requires the capacity of a large model or the efficiency of a small model. Existing routing strategies either rely on local token probabilities or post-hoc verification, introducing significant inference overhead. In this work, we propose a novel perspective on step-wise collaboration: the difficulty of a reasoning step can be inferred from its very first token. Inspired by the \"Aha Moment\" phenomenon in LRMs, we show that the entropy of the initial token serves as a strong predictor of step difficulty. Building on this insight, we introduce GlimpRouter, a training-free step-wise collaboration framework. GlimpRouter employs a lightweight model to generate only the first token of each reasoning step and routes the step to a larger model only when the initial token entropy exceeds a threshold. Experiments on multiple benchmarks demonstrate that our approach significantly reduces inference latency while preserving accuracy. For instance, GlimpRouter attains a substantial 10.7% improvement in accuracy while reducing inference latency by 25.9% compared to a standalone large model on AIME25. These results suggest a simple yet effective mechanism for reasoning: allocating computation based on a glimpse of thought rather than full-step evaluation.", "authors": ["Wenhao Zeng", "Xuteng Zhang", "Yuling Shi", "Chao Hu", "Yuting Chen", "Beijun Shen", "Xiaodong Gu"], "categories": ["cs.AI"], "fields_of_study": ["Computer Science"], "published_date": "2026-01-08", "url": "https://arxiv.org/abs/2601.05110", "pdf_url": "https://arxiv.org/pdf/2601.05110v3", "arxiv_id": "2601.05110", "doi": "10.48550/arXiv.2601.05110", "citation_count": 4, "influential_citation_count": 1, "has_code": true, "code_url": "https://github.com/Zengwh02/GlimpRouter", "venue": "arXiv.org", "quality_score": 0.6694}
{"id": "44661096f4a005e552b929966ce423eaa9946148a5f23e65740aeaa2b30d1f8e", "sources": ["arxiv", "semantic_scholar"], "title": "DiffCoT: Diffusion-styled Chain-of-Thought Reasoning in LLMs", "abstract": "Chain-of-Thought (CoT) reasoning improves multi-step mathematical problem solving in large language models but remains vulnerable to exposure bias and error accumulation, as early mistakes propagate irreversibly through autoregressive decoding. In this work, we propose DiffCoT, a diffusion-styled CoT framework that reformulates CoT reasoning as an iterative denoising process. DiffCoT integrates diffusion principles at the reasoning-step level via a sliding-window mechanism, enabling unified generation and retrospective correction of intermediate steps while preserving token-level autoregression. To maintain causal consistency, we further introduce a causal diffusion noise schedule that respects the temporal structure of reasoning chains. Extensive experiments on three multi-step CoT reasoning benchmarks across diverse model backbones demonstrate that DiffCoT consistently outperforms existing CoT preference optimization methods, yielding improved robustness and error-correction capability in CoT reasoning.", "authors": ["Shidong Cao", "Hongzhan Lin", "Yuxuan Gu", "Ziyang Luo", "Jing Ma"], "categories": ["cs.CL"], "fields_of_study": ["Computer Science"], "published_date": "2026-01-07", "url": "https://arxiv.org/abs/2601.03559", "pdf_url": "https://arxiv.org/pdf/2601.03559v2", "arxiv_id": "2601.03559", "doi": "10.48550/arXiv.2601.03559", "citation_count": 4, "influential_citation_count": 0, "has_code": false, "code_url": null, "venue": "arXiv.org", "quality_score": 0.432}
{"id": "ec674373de6cca044356c5606c49665652f874c64dba311c294d0ff99c684886", "sources": ["arxiv", "semantic_scholar"], "title": "ChaosBench-Logic: A Benchmark for Logical and Symbolic Reasoning on Chaotic Dynamical Systems", "abstract": "Large language models (LLMs) excel at natural language tasks but remain brittle in domains requiring precise logical and symbolic reasoning. Chaotic dynamical systems provide an especially demanding test because chaos is deterministic yet often misinterpreted as randomness or complexity. We introduce ChaosBench-Logic, a benchmark that evaluates LLM reasoning across 30 diverse dynamical systems using a unified first-order logic (FOL) ontology. Each system is annotated with truth assignments for 11 semantic predicates, and 621 questions are generated across seven reasoning categories, including multi-hop implications, cross-system analogies, counterfactual reasoning, bias probes, and multi-turn dialogues. We define metrics for logical accuracy, implication consistency, dialogue coherence, and contradiction, and we release an open-source evaluation pipeline. Initial experiments show that frontier LLMs such as GPT-4, Claude 3.5 Sonnet, Gemini 2.5 Flash, and the open-source LLaMA-3 70B achieve 91-94% per-item accuracy, yet still score 0% on compositional items and exhibit fragile global coherence. Dialogue-level accuracy ranges from 53.1% (GPT-4 CoT) to 75.5% (LLaMA-3 zero-shot). ChaosBench-Logic provides a rigorous testbed for diagnosing such failures and a foundation for developing neuro-symbolic approaches that improve scientific reasoning in LLMs.", "authors": ["Noel Thomas"], "categories": ["cs.AI", "cs.IT"], "fields_of_study": ["Computer Science", "Mathematics"], "published_date": "2026-01-05", "url": "https://arxiv.org/abs/2601.01982", "pdf_url": "https://arxiv.org/pdf/2601.01982v1", "arxiv_id": "2601.01982", "doi": "10.48550/arXiv.2601.01982", "citation_count": 1, "influential_citation_count": 1, "has_code": true, "code_url": null, "venue": "arXiv.org", "quality_score": 0.6641}
{"id": "45bdf9998e62051dbc139cc03d39a4c26119cde490b298b55f34ba72cd191ead", "sources": ["arxiv", "semantic_scholar"], "title": "Falcon-H1R: Pushing the Reasoning Frontiers with a Hybrid Model for Efficient Test-Time Scaling", "abstract": "This work introduces Falcon-H1R, a 7B-parameter reasoning-optimized model that establishes the feasibility of achieving competitive reasoning performance with small language models (SLMs). Falcon-H1R stands out for its parameter efficiency, consistently matching or outperforming SOTA reasoning models that are $2\\times$ to $7\\times$ larger across a variety of reasoning-intensive benchmarks. These results underscore the importance of careful data curation and targeted training strategies (via both efficient SFT and RL scaling) in delivering significant performance gains without increasing model size. Furthermore, Falcon-H1R advances the 3D limits of reasoning efficiency by combining faster inference (through its hybrid-parallel architecture design), token efficiency, and higher accuracy. This unique blend makes Falcon-H1R-7B a practical backbone for scaling advanced reasoning systems, particularly in scenarios requiring extensive chain-of-thoughts generation and parallel test-time scaling. Leveraging the recently introduced DeepConf approach, Falcon-H1R achieves state-of-the-art test-time scaling efficiency, offering substantial improvements in both accuracy and computational cost. As a result, Falcon-H1R demonstrates that compact models, through targeted model training and architectural choices, can deliver robust and scalable reasoning performance.", "authors": [" Falcon LLM Team", "Iheb Chaabane", "Puneesh Khanna", "Suhail Mohmad", "Slim Frikha", "Shi Hu", "Abdalgader Abubaker", "Reda Alami", "Mikhail Lubinets", "Mohamed El Amine Seddik", "Hakim Hacid"], "categories": ["cs.AI"], "fields_of_study": ["Computer Science"], "published_date": "2026-01-05", "url": "https://arxiv.org/abs/2601.02346", "pdf_url": "https://arxiv.org/pdf/2601.02346v1", "arxiv_id": "2601.02346", "doi": "10.48550/arXiv.2601.02346", "citation_count": 9, "influential_citation_count": 1, "has_code": false, "code_url": null, "venue": "arXiv.org", "quality_score": 0.4297}
{"id": "ce781922f11211fb7b5e6e2044c3efd9573752f559cf4cccdf91abb2fad7f8f5", "sources": ["arxiv", "semantic_scholar"], "title": "Streaming Hallucination Detection in Long Chain-of-Thought Reasoning", "abstract": "Long chain-of-thought (CoT) reasoning improves the performance of large language models, yet hallucinations in such settings often emerge subtly and propagate across reasoning steps. We suggest that hallucination in long CoT reasoning is better understood as an evolving latent state rather than a one-off erroneous event. Accordingly, we treat step-level hallucination judgments as local observations and introduce a cumulative prefix-level hallucination signal that tracks the global evolution of the reasoning state over the entire trajectory. Overall, our approach enables streaming hallucination detection in long CoT reasoning, providing real-time, interpretable evidence.", "authors": ["Haolang Lu", "Minghui Pan", "Ripeng Li", "Guoshun Nan", "Jialin Zhuang", "Zijie Zhao", "Zhongxiang Sun", "Kun Wang", "Yang Liu"], "categories": ["cs.AI"], "fields_of_study": ["Computer Science"], "published_date": "2026-01-05", "url": "https://arxiv.org/abs/2601.02170", "pdf_url": "https://arxiv.org/pdf/2601.02170v1", "arxiv_id": "2601.02170", "doi": "10.48550/arXiv.2601.02170", "citation_count": 2, "influential_citation_count": 0, "has_code": false, "code_url": null, "venue": "arXiv.org", "quality_score": 0.4297}
{"id": "23cabd5d19680e31a2a0e0b89274b3cc8913bad65762b258476a9882e57df868", "sources": ["arxiv", "semantic_scholar"], "title": "Watch Wider and Think Deeper: Collaborative Cross-modal Chain-of-Thought for Complex Visual Reasoning", "abstract": "Multi-modal reasoning requires the seamless integration of visual and linguistic cues, yet existing Chain-of-Thought methods suffer from two critical limitations in cross-modal scenarios: (1) over-reliance on single coarse-grained image regions, and (2) semantic fragmentation between successive reasoning steps. To address these issues, we propose the CoCoT (Collaborative Coross-modal Thought) framework, built upon two key innovations: a) Dynamic Multi-Region Grounding to adaptively detect the most relevant image regions based on the question, and b) Relation-Aware Reasoning to enable multi-region collaboration by iteratively aligning visual cues to form a coherent and logical chain of thought. Through this approach, we construct the CoCoT-70K dataset, comprising 74,691 high-quality samples with multi-region annotations and structured reasoning chains. Extensive experiments demonstrate that CoCoT significantly enhances complex visual reasoning, achieving an average accuracy improvement of 15.4% on LLaVA-1.5 and 4.0% on Qwen2-VL across six challenging benchmarks. The data and code are available at: https://github.com/deer-echo/CoCoT.", "authors": ["Wenting Lu", "Didi Zhu", "Tao Shen", "Donglin Zhu", "Ayong Ye", "Chao Wu"], "categories": ["cs.CV", "cs.AI"], "fields_of_study": ["Computer Science"], "published_date": "2026-01-04", "url": "https://arxiv.org/abs/2601.02422", "pdf_url": "https://arxiv.org/pdf/2601.02422v1", "arxiv_id": "2601.02422", "doi": "10.48550/arXiv.2601.02422", "citation_count": 0, "influential_citation_count": 0, "has_code": true, "code_url": "https://github.com/deer-echo/CoCoT", "venue": "arXiv.org", "quality_score": 0.6623}
{"id": "224131d3ff11498af1292e3908af516e7eabe6f5db514540051daa99d012b127", "sources": ["arxiv", "semantic_scholar"], "title": "End-to-End Test-Time Training for Long Context", "abstract": "We formulate long-context language modeling as a problem in continual learning rather than architecture design. Under this formulation, we only use a standard architecture -- a Transformer with sliding-window attention. However, our model continues learning at test time via next-token prediction on the given context, compressing the context it reads into its weights. In addition, we improve the model's initialization for learning at test time via meta-learning at training time. Overall, our method, a form of Test-Time Training (TTT), is End-to-End (E2E) both at test time (via next-token prediction) and training time (via meta-learning), in contrast to previous forms. We conduct extensive experiments with a focus on scaling properties. In particular, for 3B models trained with 164B tokens, our method (TTT-E2E) scales with context length in the same way as Transformer with full attention, while others, such as Mamba 2 and Gated DeltaNet, do not. However, similar to RNNs, TTT-E2E has constant inference latency regardless of context length, making it 2.7 times faster than full attention for 128K context. Our code is publicly available.", "authors": ["Arnuv Tandon", "Karan Dalal", "Xinhao Li", "Daniel Koceja", "Marcel Rød", "Sam Buchanan", "Xiaolong Wang", "Jure Leskovec", "Sanmi Koyejo", "Tatsunori Hashimoto", "Carlos Guestrin", "Jed McCaleb", "Yejin Choi", "Yu Sun"], "categories": ["cs.LG"], "fields_of_study": ["Computer Science"], "published_date": "2025-12-29", "url": "https://arxiv.org/abs/2512.23675", "pdf_url": "https://arxiv.org/pdf/2512.23675v2", "arxiv_id": "2512.23675", "doi": "10.48550/arXiv.2512.23675", "citation_count": 26, "influential_citation_count": 2, "has_code": true, "code_url": "https://github.com/test-time-training/e2e", "venue": "arXiv.org", "quality_score": 0.6517}
{"id": "bf151f73a2fbb8388bb83b568828d9ebeeeb5aca3ea2d822ce74f2e8d126aa84", "sources": ["arxiv", "semantic_scholar"], "title": "Real-Time Service Subscription and Adaptive Offloading Control in Vehicular Edge Computing", "abstract": "Vehicular Edge Computing (VEC) has emerged as a promising paradigm for enhancing the computational efficiency and service quality in intelligent transportation systems by enabling vehicles to wirelessly offload computation-intensive tasks to nearby Roadside Units. However, efficient task offloading and resource allocation for time-critical applications in VEC remain challenging due to constrained network bandwidth and computational resources, stringent task deadlines, and rapidly changing network conditions. To address these challenges, we formulate a Deadline-Constrained Task Offloading and Resource Allocation Problem (DOAP), denoted as $\\mathbf{P}$, in VEC with both bandwidth and computational resource constraints, aiming to maximize the total vehicle utility. To solve $\\mathbf{P}$, we propose $\\mathtt{SARound}$, an approximation algorithm based on Linear Program rounding and local-ratio techniques, that improves the best-known approximation ratio for DOAP from $\\frac{1}{6}$ to $\\frac{1}{4}$. Additionally, we design an online service subscription and offloading control framework to address the challenges of short task deadlines and rapidly changing wireless network conditions. To validate our approach, we develop a comprehensive VEC simulator, VecSim, using the open-source simulation libraries OMNeT++ and Simu5G. VecSim integrates our designed framework to manage the full life-cycle of real-time vehicular tasks. Experimental results, based on profiled object detection applications and real-world taxi trace data, show that $\\mathtt{SARound}$ consistently outperforms state-of-the-art baselines under varying network conditions while maintaining runtime efficiency.", "authors": ["Chuanchao Gao", "Arvind Easwaran"], "categories": ["cs.DC", "cs.DM"], "fields_of_study": ["Computer Science"], "published_date": "2025-12-16", "url": "https://arxiv.org/abs/2512.14002", "pdf_url": "https://arxiv.org/pdf/2512.14002v1", "arxiv_id": "2512.14002", "doi": "10.1109/RTSS66672.2025.00023", "citation_count": 1, "influential_citation_count": 0, "has_code": true, "code_url": null, "venue": "IEEE Real-Time Systems Symposium", "quality_score": 0.6286}
{"id": "4469ff01c317c8fc8fea5031230d45b312e765a27a264efe4bc145cd34749ab9", "sources": ["arxiv", "semantic_scholar"], "title": "AutoTool: Dynamic Tool Selection and Integration for Agentic Reasoning", "abstract": "Agentic reinforcement learning has advanced large language models (LLMs) to reason through long chain-of-thought trajectories while interleaving external tool use. Existing approaches assume a fixed inventory of tools, which limits the adaptability of LLM agents to new or evolving toolsets. We present AutoTool, a training framework that equips LLM agents with dynamic tool-selection capabilities throughout their reasoning trajectories. AutoTool employs a dual-phase optimization pipeline: (i) SFT and RL-based trajectory stabilization for coherent reasoning, and (ii) KL-regularized Plackett-Luce Ranking to refine consistent multi-step tool selection. We further build a 200k dataset with explicit tool-selection rationales across 1,000+ tools and 100+ tasks spanning mathematics, science, code generation, and multimodal reasoning. Across ten diverse benchmarks, we train two base models, Qwen3-8B and Qwen2.5-VL-7B, with AutoTool. With fewer parameters, AutoTool consistently outperforms advanced LLM agents and tool-integration methods, yielding average gains of 6.4% in math & science reasoning, 4.5% in search-based QA, 7.7% in code generation, and 6.9% in multimodal understanding. In addition, AutoTool exhibits stronger generalization by dynamically leveraging unseen tools from evolving toolsets during inference.", "authors": ["Jiaru Zou", "Ling Yang", "Yunzhe Qi", "Sirui Chen", "Mengting Ai", "Ke Shen", "Jingrui He", "Mengdi Wang"], "categories": ["cs.CL", "cs.LG"], "fields_of_study": ["Computer Science"], "published_date": "2025-12-15", "url": "https://arxiv.org/abs/2512.13278", "pdf_url": "https://arxiv.org/pdf/2512.13278v2", "arxiv_id": "2512.13278", "doi": "10.48550/arXiv.2512.13278", "citation_count": 2, "influential_citation_count": 0, "has_code": false, "code_url": null, "venue": "arXiv.org", "quality_score": 0.4056}
{"id": "656f2bcc9567e4a9a5f04130cd68f0684f62dbdd5bd4c0c66badee17ac827a9d", "sources": ["arxiv", "semantic_scholar"], "title": "Limits and Gains of Test-Time Scaling in Vision-Language Reasoning", "abstract": "Test-time scaling (TTS) has emerged as a powerful paradigm for improving the reasoning ability of Large Language Models (LLMs) by allocating additional computation at inference, yet its application to multimodal systems such as Vision-Language Models (VLMs) remains underexplored. In this work, we present a systematic empirical study of inference time reasoning methods applied across both open-source and closed-source VLMs on different benchmarks. Our results reveal that while closed-source models consistently benefit from structured reasoning and iterative Self-Refinement, open-source VLMs show inconsistent behavior: external verification provides the most reliable gains, whereas iterative refinement often degrades performance. We further find that the effectiveness of TTS is dataset-dependent, yielding clear improvements on multi-step reasoning tasks but offering only limited gains on perception-focused benchmarks. These findings demonstrate that TTS is not a universal solution and must be tailored to both model capabilities and task characteristics, motivating future work on adaptive TTS strategies and multimodal reward models.", "authors": ["Mohammadjavad Ahmadpour", "Amirmahdi Meighani", "Payam Taebi", "Omid Ghahroodi", "Amirmohammad Izadi", "Mahdieh Soleymani Baghshah"], "categories": ["cs.LG"], "fields_of_study": ["Computer Science"], "published_date": "2025-12-11", "url": "https://arxiv.org/abs/2512.11109", "pdf_url": "https://arxiv.org/pdf/2512.11109v1", "arxiv_id": "2512.11109", "doi": "10.48550/arXiv.2512.11109", "citation_count": 2, "influential_citation_count": 0, "has_code": true, "code_url": null, "venue": "arXiv.org", "quality_score": 0.6198}
{"id": "98dfbcc6c11cc2835f5f2cb4bacdf4c2edaf9663f6c2a54293838e0bd1247093", "sources": ["arxiv", "semantic_scholar"], "title": "Rethinking Chain-of-Thought Reasoning for Videos", "abstract": "Chain-of-thought (CoT) reasoning has been highly successful in solving complex tasks in natural language processing, and recent multimodal large language models (MLLMs) have extended this paradigm to video reasoning. However, these models typically build on lengthy reasoning chains and large numbers of input visual tokens. Motivated by empirical observations from our benchmark study, we hypothesize that concise reasoning combined with a reduced set of visual tokens can be sufficient for effective video reasoning. To evaluate this hypothesis, we design and validate an efficient post-training and inference framework that enhances a video MLLM's reasoning capability. Our framework enables models to operate on compressed visual tokens and generate brief reasoning traces prior to answering. The resulting models achieve substantially improved inference efficiency, deliver competitive performance across diverse benchmarks, and avoid reliance on manual CoT annotations or supervised fine-tuning. Collectively, our results suggest that long, human-like CoT reasoning may not be necessary for general video reasoning, and that concise reasoning can be both effective and efficient. Our code will be released at https://github.com/LaVi-Lab/Rethink_CoT_Video.", "authors": ["Yiwu Zhong", "Zi-Yuan Hu", "Yin Li", "Liwei Wang"], "categories": ["cs.CV", "cs.AI", "cs.CL", "cs.LG"], "fields_of_study": ["Computer Science"], "published_date": "2025-12-10", "url": "https://arxiv.org/abs/2512.09616", "pdf_url": "https://arxiv.org/pdf/2512.09616v1", "arxiv_id": "2512.09616", "doi": "10.48550/arXiv.2512.09616", "citation_count": 4, "influential_citation_count": 0, "has_code": true, "code_url": "https://github.com/LaVi-Lab/Rethink_CoT_Video", "venue": "arXiv.org", "quality_score": 0.618}
{"id": "fd4b2a00552821ea2f45b886a78cb3604df81ce48dee7ebe6466d86b874a084b", "sources": ["arxiv", "semantic_scholar"], "title": "MM-CoT:A Benchmark for Probing Visual Chain-of-Thought Reasoning in Multimodal Models", "abstract": "The ability to perform Chain-of-Thought (CoT) reasoning marks a major milestone for multimodal models (MMs), enabling them to solve complex visual reasoning problems. Yet a critical question remains: is such reasoning genuinely grounded in visual evidence and logically coherent? Existing benchmarks emphasize generation but neglect verification, i.e., the capacity to assess whether a reasoning chain is both visually consistent and logically valid. To fill this gap, we introduce MM-CoT, a diagnostic benchmark specifically designed to probe the visual grounding and logical coherence of CoT reasoning in MMs. Instead of generating free-form explanations, models must select the sole event chain that satisfies two orthogonal constraints: (i) visual consistency, ensuring all steps are anchored in observable evidence, and (ii) logical coherence, ensuring causal and commonsense validity. Adversarial distractors are engineered to violate one of these constraints, exposing distinct reasoning failures. We evaluate leading vision-language models on MM-CoT and find that even the most advanced systems struggle, revealing a sharp discrepancy between generative fluency and true reasoning fidelity. MM-CoT shows low correlation with existing benchmarks, confirming that it measures a unique combination of visual grounding and logical reasoning. This benchmark provides a foundation for developing future models that reason not just plausibly, but faithfully and coherently within the visual world.", "authors": ["Jusheng Zhang", "Kaitong Cai", "Xiaoyang Guo", "Sidi Liu", "Qinhan Lv", "Ruiqi Chen", "Jing Yang", "Yijia Fan", "Xiaofei Sun", "Jian Wang", "Ziliang Chen", "Liang Lin", "Keze Wang"], "categories": ["cs.CV", "cs.AI"], "fields_of_study": ["Computer Science"], "published_date": "2025-12-09", "url": "https://arxiv.org/abs/2512.08228", "pdf_url": "https://arxiv.org/pdf/2512.08228v1", "arxiv_id": "2512.08228", "doi": "10.48550/arXiv.2512.08228", "citation_count": 35, "influential_citation_count": 2, "has_code": false, "code_url": null, "venue": "arXiv.org", "quality_score": 0.3987}
{"id": "509bc2159dbca241853f6087b28ff21c4f9180ce2ca0cae36295e8c732aedc71", "sources": ["arxiv", "semantic_scholar"], "title": "RoBoN: Routed Online Best-of-n for Test-Time Scaling with Multiple LLMs", "abstract": "Best-of-$n$ is a widely used test-time scaling approach for LLM inference. Yet despite evidence that LLMs exhibit complementary strengths across tasks, traditionally best-of-$n$ relies on a single model to generate responses. We propose RoBoN (Routed Online Best-of-$n$), a sequential multi-LLM alternative to the prevailing single-model best-of-$n$. Given a suite of models $\\{m_i\\}_{i=1}^M$, RoBoN sequentially routes generations one-by-one across models, based on scores computed using a reward model and an agreement signal on the predicted responses. This online routing requires no additional training, keeps compute parity, and works with any plug-in reward model. Across reasoning benchmarks (MATH500, OlympiadBench, MinervaMath, GSM8K, MMLU), RoBoN consistently outperforms standard best-of-$n$ applied to each individual model for larger $n$, with gains of up to 3.4\\% in absolute accuracy, and also improves over a uniform multi-model portfolio baseline. Our results indicate that diversity across models can be exploited at inference to improve best-of-$n$ performance over any constituent model alone, providing a simple, training-free path to test-time scaling with multiple LLMs.", "authors": ["Jonathan Geuter", "Gregor Kornhardt"], "categories": ["cs.LG", "cs.AI"], "fields_of_study": ["Computer Science"], "published_date": "2025-12-05", "url": "https://arxiv.org/abs/2512.05542", "pdf_url": "https://arxiv.org/pdf/2512.05542v1", "arxiv_id": "2512.05542", "doi": "10.48550/arXiv.2512.05542", "citation_count": 0, "influential_citation_count": 0, "has_code": false, "code_url": null, "venue": "arXiv.org", "quality_score": 0.3942}
{"id": "4620bc0b817a2ddd6d16265f9a890ff30ac5e0d04016cd52b161b49fd20975c0", "sources": ["arxiv", "semantic_scholar"], "title": "Probing the effectiveness of World Models for Spatial Reasoning through Test-time Scaling", "abstract": "Vision-Language Models (VLMs) remain limited in spatial reasoning tasks that require multi-view understanding and embodied perspective shifts. Recent approaches such as MindJourney attempt to mitigate this gap through test-time scaling where a world model imagines action-conditioned trajectories and a heuristic verifier selects helpful views from such trajectories. In this work, we systematically examine how such test-time verifiers behave across benchmarks, uncovering both their promise and their pitfalls. Our uncertainty-based analyses show that MindJourney's verifier provides little meaningful calibration, and that random scoring often reduces answer entropy equally well, thus exposing systematic action biases and unreliable reward signals. To mitigate these, we introduce a Verification through Spatial Assertions (ViSA) framework that grounds the test-time reward in verifiable, frame-anchored micro-claims. This principled verifier consistently improves spatial reasoning on the SAT-Real benchmark and corrects trajectory-selection biases through more balanced exploratory behavior. However, on the challenging MMSI-Bench, none of the verifiers, including ours, achieve consistent scaling, suggesting that the current world models form an information bottleneck where imagined views fail to enrich fine-grained reasoning. Together, these findings chart the bad, good, and ugly aspects of test-time verification for world-model-based reasoning. Our code is available at https://github.com/chandar-lab/visa-for-mindjourney.", "authors": ["Saurav Jha", "M. Jehanzeb Mirza", "Wei Lin", "Shiqi Yang", "Sarath Chandar"], "categories": ["cs.CV", "cs.AI"], "fields_of_study": ["Computer Science"], "published_date": "2025-12-05", "url": "https://arxiv.org/abs/2512.05809", "pdf_url": "https://arxiv.org/pdf/2512.05809v1", "arxiv_id": "2512.05809", "doi": "10.48550/arXiv.2512.05809", "citation_count": 0, "influential_citation_count": 0, "has_code": true, "code_url": "https://github.com/chandar-lab/visa-for-mindjourney", "venue": "arXiv.org", "quality_score": 0.6092}
{"id": "2250cb6b6ce29df096a5e16d34cbf68d8bf772f263ef5acc0628dc954adb8900", "sources": ["arxiv", "semantic_scholar"], "title": "Visual Reasoning Tracer: Object-Level Grounded Reasoning Benchmark", "abstract": "Recent advances in Multimodal Large Language Models (MLLMs) have significantly improved performance on tasks such as visual grounding and visual question answering. However, the reasoning processes of these models remain largely opaque; they typically output only final predictions without revealing the intermediate steps or fine-grained evidence (e.g., pixels, locations) that lead to the result. This contrasts with human intelligence, which naturally operates through a chain of visual reasoning. To address this limitation, we introduce the Visual Reasoning Tracer (VRT) task, which requires models to not only localize the target object but also explicitly predict the intermediate objects that form the reasoning path. To advance research in this area, we contribute: (1) VRT-Bench, a human-annotated benchmark for evaluating visual reasoning; (2) a new metric for assessing the quality of reasoning traces; and (3) VRT-80k, a large-scale dataset for reasoning model training. Our experiments reveal that while existing models often produce the correct final output, they struggle to ground their intermediate reasoning. In contrast, models trained on VRT-80k achieve substantial improvements in tracing the reasoning path.", "authors": ["Haobo Yuan", "Yueyi Sun", "Yanwei Li", "Tao Zhang", "Xueqing Deng", "Henghui Ding", "Lu Qi", "Anran Wang", "Xiangtai Li", "Ming-Hsuan Yang"], "categories": ["cs.CV"], "fields_of_study": ["Computer Science"], "published_date": "2025-12-04", "url": "https://arxiv.org/abs/2512.05091", "pdf_url": "https://arxiv.org/pdf/2512.05091v1", "arxiv_id": "2512.05091", "doi": "10.48550/arXiv.2512.05091", "citation_count": 8, "influential_citation_count": 0, "has_code": false, "code_url": null, "venue": "arXiv.org", "quality_score": 0.393}
{"id": "5d284a5556216242533927d57448e80ed7ed9a88528748c6052d69252b2686b5", "sources": ["arxiv", "semantic_scholar"], "title": "The Art of Scaling Test-Time Compute for Large Language Models", "abstract": "Test-time scaling (TTS) -- the dynamic allocation of compute during inference -- is a promising direction for improving reasoning in large language models (LLMs). However, a systematic comparison of well-known TTS strategies under identical conditions is missing, and the influence of model type and problem difficulty on performance remains unclear. To address these gaps, we conduct the first large-scale study of TTS, spanning over thirty billion tokens generated using eight open-source LLMs (7B to 235B parameters), across four reasoning datasets. We observe three consistent trends: (1) no single TTS strategy universally dominates; (2) reasoning models exhibit distinct trace-quality patterns across problem difficulty and trace length, forming short-horizon and long-horizon categories; and (3) for a given model type, the optimal TTS performance scales monotonically with compute budget. Based on these insights, we provide a practical recipe for selecting the best TTS strategy, considering problem difficulty, model type, and compute budget, providing a practical guide to effective inference-time scaling.", "authors": ["Aradhye Agarwal", "Ayan Sengupta", "Tanmoy Chakraborty"], "categories": ["cs.CL"], "fields_of_study": ["Computer Science"], "published_date": "2025-12-01", "url": "https://arxiv.org/abs/2512.02008", "pdf_url": "https://arxiv.org/pdf/2512.02008v1", "arxiv_id": "2512.02008", "doi": "10.48550/arXiv.2512.02008", "citation_count": 9, "influential_citation_count": 0, "has_code": true, "code_url": null, "venue": "arXiv.org", "quality_score": 0.6021}
{"id": "eb574edb918b77f9378bdbdd439900335030631eb5a00547a922fbf993c3652e", "sources": ["arxiv", "semantic_scholar"], "title": "Reasoning Under Pressure: How do Training Incentives Influence Chain-of-Thought Monitorability?", "abstract": "AI systems that output their reasoning in natural language offer an opportunity for safety -- we can \\emph{monitor} their chain of thought (CoT) for undesirable reasoning, such as the pursuit of harmful objectives. However, the extent to which CoT faithfully reflects the underlying reasoning process, and hence the extent to which it can be usefully monitored, may be influenced by certain aspects of training. We investigate how different \\emph{training incentives}, applied to a reasoning model, affect its monitorability. We introduce a novel methodology for measuring monitorability according to whether a monitor can predict a key latent variable using the model's reasoning. When controlling for accuracy, we do not find evidence for consistent effects from commonly used incentives (length penalties and KL regularisation), but we find that adversarial optimisation (penalising monitor accuracy) degrades monitor performance, while direct optimisation for monitorability does not reliably lead to improvements. Our code is available at https://github.com/QiyaoWei/reasoning-under-pressure.", "authors": ["Matt MacDermott", "Qiyao Wei", "Rada Djoneva", "Francis Rhys Ward"], "categories": ["cs.AI", "cs.CR"], "fields_of_study": ["Computer Science"], "published_date": "2025-11-28", "url": "https://arxiv.org/abs/2512.00218", "pdf_url": "https://arxiv.org/pdf/2512.00218v2", "arxiv_id": "2512.00218", "doi": "10.48550/arXiv.2512.00218", "citation_count": 2, "influential_citation_count": 0, "has_code": true, "code_url": "https://github.com/QiyaoWei/reasoning-under-pressure", "venue": "arXiv.org", "quality_score": 0.5968}
{"id": "e811c855a83df33f20c6f5fce8cf85709a576c903e589562e2ab1de69e7494e1", "sources": ["arxiv", "semantic_scholar"], "title": "Focused Chain-of-Thought: Efficient LLM Reasoning via Structured Input Information", "abstract": "Recent large language models achieve strong reasoning performance by generating detailed chain-of-thought traces, but this often leads to excessive token use and high inference latency. Existing efficiency approaches typically focus on model-centric interventions, such as reinforcement learning or supervised fine-tuning, to reduce verbosity. In contrast, we propose a training-free, input-centric approach. Inspired by cognitive psychology, we introduce Focused Chain-of-Thought (F-CoT), which separates information extraction from the reasoning process. F-CoT first organizes the essential information from a query into a concise, structured context and then guides the model to reason exclusively over this context. By preventing attention to irrelevant details, F-CoT naturally produces shorter reasoning paths. On arithmetic word problems, F-CoT reduces generated tokens by 2-3x while maintaining accuracy comparable to standard zero-shot CoT. These results highlight structured input as a simple yet effective lever for more efficient LLM reasoning.", "authors": ["Lukas Struppek", "Dominik Hintersdorf", "Hannah Struppek", "Daniel Neider", "Kristian Kersting"], "categories": ["cs.CL", "cs.AI"], "fields_of_study": ["Computer Science"], "published_date": "2025-11-27", "url": "https://arxiv.org/abs/2511.22176", "pdf_url": "https://arxiv.org/pdf/2511.22176v1", "arxiv_id": "2511.22176", "doi": "10.48550/arXiv.2511.22176", "citation_count": 1, "influential_citation_count": 0, "has_code": false, "code_url": null, "venue": "arXiv.org", "quality_score": 0.385}
{"id": "715075618f157752c13480f07e6136babcff9f8d7fa91970350a6e6f85ffb915", "sources": ["arxiv", "semantic_scholar"], "title": "VICoT-Agent: A Vision-Interleaved Chain-of-Thought Framework for Interpretable Multimodal Reasoning and Scalable Remote Sensing Analysis", "abstract": "The current remote sensing image analysis task is increasingly evolving from traditional object recognition to complex intelligence reasoning, which places higher requirements on the model's reasoning ability and the flexibility of tool invocation. To this end, we propose a new multimodal agent framework, Vision-Interleaved Chain-of-Thought Framework (VICoT), which implements explicit multi-round reasoning by dynamically incorporating visual tools into the chain of thought. Through a stack-based reasoning structure and a modular MCP-compatible tool suite, VICoT enables LLMs to efficiently perform multi-round, interleaved vision-language reasoning tasks with strong generalization and flexibility.We also propose the Reasoning Stack distillation method to migrate complex Agent behaviors to small, lightweight models, which ensures the reasoning capability while significantly reducing complexity. Experiments on multiple remote sensing benchmarks demonstrate that VICoT significantly outperforms existing SOTA frameworks in reasoning transparency, execution efficiency, and generation quality.", "authors": ["Chujie Wang", "Zhiyuan Luo", "Ruiqi Liu", "Can Ran", "Shenghua Fan", "Xi Chen", "Chu He"], "categories": ["cs.AI", "cs.MA"], "fields_of_study": ["Computer Science"], "published_date": "2025-11-25", "url": "https://arxiv.org/abs/2511.20085", "pdf_url": "https://arxiv.org/pdf/2511.20085v3", "arxiv_id": "2511.20085", "doi": "10.48550/arXiv.2511.20085", "citation_count": 3, "influential_citation_count": 0, "has_code": false, "code_url": null, "venue": "arXiv.org", "quality_score": 0.3827}
{"id": "575d9c447db11af25482361a575f5b4447ac8e603de12469b0e54a190382897e", "sources": ["arxiv", "semantic_scholar"], "title": "VisReason: A Large-Scale Dataset for Visual Chain-of-Thought Reasoning", "abstract": "Chain-of-Thought (CoT) prompting has proven remarkably effective for eliciting complex reasoning in large language models (LLMs). Yet, its potential in multimodal large language models (MLLMs) remains largely untapped, hindered by the absence of large-scale datasets that capture the rich, spatially grounded reasoning intrinsic to visual understanding. Existing visual-CoT resources are typically small, domain-specific, or lack the human-like stepwise structure necessary for compositional visual reasoning. In this paper, we introduce VisReason, a large-scale dataset designed to advance visual Chain-of-Thought reasoning. VisReason comprises 489K annotated examples spanning four diverse domains, each featuring multi-round, human-like rationales that guide MLLMs through interpretable visual reasoning steps. Building upon this, we curate VisReason-Pro, a 165K subset produced with a stronger expert-level GPT annotator, enriched with detailed reasoning traces and 3D spatial grounding via depth-informed annotations. Fine-tuning the state-of-the-art Qwen2.5-VL model on VisReason and VisReason-Pro yields substantial improvements in step-by-step visual reasoning accuracy, interpretability, and cross-benchmark generalization. These results demonstrate that VisReason equips MLLMs with more systematic and generalizable reasoning capabilities. We envision VisReason as a cornerstone for cultivating human-like visual reasoning, paving the way toward the next generation of multimodal intelligence.", "authors": ["Lingxiao Li", "Yifan Wang", "Xinyan Gao", "Chen Tang", "Xiangyu Yue", "Chenyu You"], "categories": ["cs.CV", "cs.LG"], "fields_of_study": ["Computer Science"], "published_date": "2025-11-21", "url": "https://arxiv.org/abs/2511.17731", "pdf_url": "https://arxiv.org/pdf/2511.17731v1", "arxiv_id": "2511.17731", "doi": "10.48550/arXiv.2511.17731", "citation_count": 3, "influential_citation_count": 0, "has_code": false, "code_url": null, "venue": "arXiv.org", "quality_score": 0.3781}
{"id": "1dea087c7c6c29001a9c47ba7731720b9e3742da5d7ed80bfb66362757bb74ba", "sources": ["arxiv", "semantic_scholar"], "title": "Geospatial Chain of Thought Reasoning for Enhanced Visual Question Answering on Satellite Imagery", "abstract": "Geospatial chain of thought (CoT) reasoning is essential for advancing Visual Question Answering (VQA) on satellite imagery, particularly in climate related applications such as disaster monitoring, infrastructure risk assessment, urban resilience planning, and policy support. Existing VQA models enable scalable interpretation of remote sensing data but often lack the structured reasoning required for complex geospatial queries. We propose a VQA framework that integrates CoT reasoning with Direct Preference Optimization (DPO) to improve interpretability, robustness, and accuracy. By generating intermediate rationales, the model better handles tasks involving detection, classification, spatial relations, and comparative analysis, which are critical for reliable decision support in high stakes climate domains. Experiments show that CoT supervision improves accuracy by 34.9\\% over direct baselines, while DPO yields additional gains in accuracy and reasoning quality. The resulting system advances VQA for multispectral Earth observation by enabling richer geospatial reasoning and more effective climate use cases.", "authors": ["Shambhavi Shanker", "Manikandan Padmanaban", "Jagabondhu Hazra"], "categories": ["cs.CV"], "fields_of_study": ["Computer Science"], "published_date": "2025-11-14", "url": "https://arxiv.org/abs/2511.11198", "pdf_url": "https://arxiv.org/pdf/2511.11198v1", "arxiv_id": "2511.11198", "doi": "10.48550/arXiv.2511.11198", "citation_count": 1, "influential_citation_count": 0, "has_code": false, "code_url": null, "venue": "arXiv.org", "quality_score": 0.3701}
{"id": "1ecbc5b401e14e6d91559b47aef01a8e6e412cd445abea095e94352ea3460aae", "sources": ["arxiv", "semantic_scholar"], "title": "BadThink: Triggered Overthinking Attacks on Chain-of-Thought Reasoning in Large Language Models", "abstract": "Recent advances in Chain-of-Thought (CoT) prompting have substantially improved the reasoning capabilities of large language models (LLMs), but have also introduced their computational efficiency as a new attack surface. In this paper, we propose BadThink, the first backdoor attack designed to deliberately induce \"overthinking\" behavior in CoT-enabled LLMs while ensuring stealth. When activated by carefully crafted trigger prompts, BadThink manipulates the model to generate inflated reasoning traces - producing unnecessarily redundant thought processes while preserving the consistency of final outputs. This subtle attack vector creates a covert form of performance degradation that significantly increases computational costs and inference time while remaining difficult to detect through conventional output evaluation methods. We implement this attack through a sophisticated poisoning-based fine-tuning strategy, employing a novel LLM-based iterative optimization process to embed the behavior by generating highly naturalistic poisoned data. Our experiments on multiple state-of-the-art models and reasoning tasks show that BadThink consistently increases reasoning trace lengths - achieving an over 17x increase on the MATH-500 dataset - while remaining stealthy and robust. This work reveals a critical, previously unexplored vulnerability where reasoning efficiency can be covertly manipulated, demonstrating a new class of sophisticated attacks against CoT-enabled systems.", "authors": ["Shuaitong Liu", "Renjue Li", "Lijia Yu", "Lijun Zhang", "Zhiming Liu", "Gaojie Jin"], "categories": ["cs.CR", "cs.AI"], "fields_of_study": ["Computer Science"], "published_date": "2025-11-13", "url": "https://arxiv.org/abs/2511.10714", "pdf_url": "https://arxiv.org/pdf/2511.10714v1", "arxiv_id": "2511.10714", "doi": "10.48550/arXiv.2511.10714", "citation_count": 4, "influential_citation_count": 0, "has_code": false, "code_url": null, "venue": "arXiv.org", "quality_score": 0.369}
{"id": "b672b8b26bd81c6c57e85ac7b4c22a54d37f4cb5fb4d218080467e70bbe4af32", "sources": ["arxiv", "semantic_scholar"], "title": "Benchmarking Multi-Step Legal Reasoning and Analyzing Chain-of-Thought Effects in Large Language Models", "abstract": "Large language models (LLMs) have demonstrated strong reasoning abilities across specialized domains, motivating research into their application to legal reasoning. However, existing legal benchmarks often conflate factual recall with genuine inference, fragment the reasoning process, and overlook the quality of reasoning. To address these limitations, we introduce MSLR, the first Chinese multi-step legal reasoning dataset grounded in real-world judicial decision making. MSLR adopts the IRAC framework (Issue, Rule, Application, Conclusion) to model structured expert reasoning from official legal documents. In addition, we design a scalable Human-LLM collaborative annotation pipeline that efficiently produces fine-grained step-level reasoning annotations and provides a reusable methodological framework for multi-step reasoning datasets. Evaluation of multiple LLMs on MSLR shows only moderate performance, highlighting the challenges of adapting to complex legal reasoning. Further experiments demonstrate that Self-Initiated Chain-of-Thought prompts generated by models autonomously improve reasoning coherence and quality, outperforming human-designed prompts. MSLR contributes to advancing LLM reasoning and Chain-of-Thought strategies and offers open resources for future research. The dataset and code are available at https://github.com/yuwenhan07/MSLR-Bench and https://law.sjtu.edu.cn/flszyjzx/index.html.", "authors": ["Wenhan Yu", "Xinbo Lin", "Lanxin Ni", "Jinhua Cheng", "Lei Sha"], "categories": ["cs.AI"], "fields_of_study": ["Computer Science"], "published_date": "2025-11-11", "url": "https://arxiv.org/abs/2511.07979", "pdf_url": "https://arxiv.org/pdf/2511.07979v3", "arxiv_id": "2511.07979", "doi": "10.48550/arXiv.2511.07979", "citation_count": 3, "influential_citation_count": 0, "has_code": true, "code_url": "https://github.com/yuwenhan07/MSLR-Bench", "venue": "arXiv.org", "quality_score": 0.5667}
{"id": "a9930d9c773c75d3e041b198b78fb77ecb881476730de75cc0b70a585e64d58f", "sources": ["arxiv", "semantic_scholar"], "title": "Think Consistently, Reason Efficiently: Energy-Based Calibration for Implicit Chain-of-Thought", "abstract": "Large Language Models (LLMs) have demonstrated strong reasoning capabilities through \\emph{Chain-of-Thought} (CoT) prompting, which enables step-by-step intermediate reasoning. However, explicit CoT methods rely on discrete token-level reasoning processes that are prone to error propagation and limited by vocabulary expressiveness, often resulting in rigid and inconsistent reasoning trajectories. Recent research has explored implicit or continuous reasoning in latent spaces, allowing models to perform internal reasoning before generating explicit output. Although such approaches alleviate some limitations of discrete CoT, they generally lack explicit mechanisms to enforce consistency among reasoning steps, leading to divergent reasoning paths and unstable outcomes. To address this issue, we propose EBM-CoT, an Energy-Based Chain-of-Thought Calibration framework that refines latent thought representations through an energy-based model (EBM). Our method dynamically adjusts latent reasoning trajectories toward lower-energy, high-consistency regions in the embedding space, improving both reasoning accuracy and consistency without modifying the base language model. Extensive experiments across mathematical, commonsense, and symbolic reasoning benchmarks demonstrate that the proposed framework significantly enhances the consistency and efficiency of multi-step reasoning in LLMs.", "authors": ["Zhikang Chen", "Sen Cui", "Deheng Ye", "Yu Zhang", "Yatao Bian", "Tingting Zhu"], "categories": ["cs.CL", "cs.AI", "cs.LG"], "fields_of_study": ["Computer Science"], "published_date": "2025-11-10", "url": "https://arxiv.org/abs/2511.07124", "pdf_url": "https://arxiv.org/pdf/2511.07124v1", "arxiv_id": "2511.07124", "doi": "10.48550/arXiv.2511.07124", "citation_count": 1, "influential_citation_count": 1, "has_code": false, "code_url": null, "venue": "arXiv.org", "quality_score": 0.3655}
{"id": "99df5803636363c6d06362f52a2f7f3ae6f47badf9fc980ff7b18b44784f6c0e", "sources": ["arxiv", "semantic_scholar"], "title": "Chain-of-Thought as a Lens: Evaluating Structured Reasoning Alignment between Human Preferences and Large Language Models", "abstract": "This paper primarily demonstrates a method to quantitatively assess the alignment between multi-step, structured reasoning in large language models and human preferences. We introduce the Alignment Score, a semantic-level metric that compares a model-produced chain of thought traces with a human-preferred reference by constructing semantic-entropy-based matrices over intermediate steps and measuring their divergence. Our analysis shows that Alignment Score tracks task accuracy across models and hop depths, and peaks at 2-hop reasoning. Empirical results further indicate that misalignment at greater reasoning depths is driven mainly by alignment errors such as thematic shift and redundant reasoning. Viewing chain sampling as drawing from a distribution over reasoning paths, we empirically demonstrate a strong and consistent correlation between Alignment Score and accuracy, readability, and coherence, supporting its use as a diagnostic signal. The code is available.", "authors": ["Boxuan Wang", "Zhuoyun Li", "Xinmiao Huang", "Xiaowei Huang", "Yi Dong"], "categories": ["cs.AI"], "fields_of_study": ["Computer Science"], "published_date": "2025-11-09", "url": "https://arxiv.org/abs/2511.06168", "pdf_url": "https://arxiv.org/pdf/2511.06168v3", "arxiv_id": "2511.06168", "doi": null, "citation_count": 2, "influential_citation_count": 0, "has_code": true, "code_url": null, "venue": null, "quality_score": 0.4306}
{"id": "d73f6edce588bfedb77dc501bc5ed57ba614ab99c698ecfc8bfaa418405e57c2", "sources": ["arxiv", "semantic_scholar"], "title": "Logit-Entropy Adaptive Stopping Heuristic for Efficient Chain-of-Thought Reasoning", "abstract": "Chain-of-Thought (CoT) prompting is a key technique for enabling complex reasoning in large language models. However, generating full, fixed-length rationales is computationally wasteful, inflating both token usage and latency. We introduce LEASH: Logit-Entropy Adaptive Stopping Heuristic, a training-free decoding algorithm that adaptively halts rationale generation. LEASH monitors two intrinsic signals: the slope of token-level entropy and the improvement in the top-logit margin. It terminates the generation once both signals plateau, indicating the model has reached a stable reasoning state. Across four instruction-tuned models on the GSM8K and AQuA-RAT benchmarks, LEASH reduces average token generation by 30--35% and latency by 27%, while incurring a 10 p.p. accuracy drop relative to CoT. LEASH is model-agnostic and requires no additional training or supervision, offering a simple and efficient alternative to CoT decoding.", "authors": ["Mohammad Atif Quamar", "Mohammad Areeb"], "categories": ["cs.CL"], "fields_of_study": ["Computer Science"], "published_date": "2025-11-06", "url": "https://arxiv.org/abs/2511.04654", "pdf_url": "https://arxiv.org/pdf/2511.04654v1", "arxiv_id": "2511.04654", "doi": "10.48550/arXiv.2511.04654", "citation_count": 1, "influential_citation_count": 0, "has_code": false, "code_url": null, "venue": "arXiv.org", "quality_score": 0.3609}
{"id": "2a2312a540f43afe3f9c1138eaaf927dc6d16347043f325043422e0ea8873c64", "sources": ["arxiv", "semantic_scholar"], "title": "Efficient Reasoning via Thought-Training and Thought-Free Inference", "abstract": "Recent advances in large language models (LLMs) have leveraged explicit Chain-of-Thought (CoT) prompting to improve reasoning accuracy. However, most existing methods primarily focus on compressing verbose reasoning outputs. These Long-to-Short transformations aim to improve efficiency, but require a large amount of short CoT data. In this work, we introduce \\textbf{3TF} (\\textbf{T}hought-\\textbf{T}raining and \\textbf{T}hought-\\textbf{F}ree inference), a framework for efficient reasoning that takes a Short-to-Long perspective. We first train a hybrid model that can operate in both reasoning and non-reasoning modes, and then further train it on CoT-annotated data to internalize structured reasoning, while enforcing concise, thought-free outputs at inference time using the no-reasoning mode. Unlike compression-based approaches, 3TF improves the reasoning quality of non-reasoning outputs, enabling models to perform rich internal reasoning implicitly while keeping external outputs short. Empirically, 3TF-trained models obtain large improvements on reasoning benchmarks under thought-free inference, demonstrating that high quality reasoning can be learned and executed implicitly without explicit step-by-step generation.", "authors": ["Canhui Wu", "Qiong Cao", "Chao Xue", "Wei Xi", "Xiaodong He"], "categories": ["cs.CL"], "fields_of_study": ["Computer Science"], "published_date": "2025-11-05", "url": "https://arxiv.org/abs/2511.03408", "pdf_url": "https://arxiv.org/pdf/2511.03408v3", "arxiv_id": "2511.03408", "doi": "10.48550/arXiv.2511.03408", "citation_count": 2, "influential_citation_count": 0, "has_code": false, "code_url": null, "venue": "arXiv.org", "quality_score": 0.3598}
{"id": "1b36ce09bbcb9e3deb8c1a069f592058c6342aab58eb4f068afa1283e4d0bfbd", "sources": ["arxiv", "semantic_scholar"], "title": "Optimizing Reasoning Efficiency through Prompt Difficulty Prediction", "abstract": "Reasoning language models perform well on complex tasks but are costly to deploy due to their size and long reasoning traces. We propose a routing approach that assigns each problem to the smallest model likely to solve it, reducing compute without sacrificing accuracy. Using intermediate representations from s1.1-32B, we train lightweight predictors of problem difficulty or model correctness to guide routing across a pool of reasoning models. On diverse math benchmarks, routing improves efficiency over random assignment and matches s1.1-32B's performance while using significantly less compute. Our results demonstrate that difficulty-aware routing is effective for cost-efficient deployment of reasoning models.", "authors": ["Bo Zhao", "Berkcan Kapusuzoglu", "Kartik Balasubramaniam", "Sambit Sahu", "Supriyo Chakraborty", "Genta Indra Winata"], "categories": ["cs.LG", "cs.AI"], "fields_of_study": ["Computer Science"], "published_date": "2025-11-05", "url": "https://arxiv.org/abs/2511.03808", "pdf_url": "https://arxiv.org/pdf/2511.03808v1", "arxiv_id": "2511.03808", "doi": "10.48550/arXiv.2511.03808", "citation_count": 2, "influential_citation_count": 0, "has_code": false, "code_url": null, "venue": "arXiv.org", "quality_score": 0.3598}
{"id": "f3330111932fc555d2499736590d6797f5e6ae0aa34e61d6b486c5280d494264", "sources": ["arxiv", "semantic_scholar"], "title": "Re-FORC: Adaptive Reward Prediction for Efficient Chain-of-Thought Reasoning", "abstract": "We propose Re-FORC, an adaptive reward prediction method that, given a context, enables prediction of the expected future rewards as a function of the number of future thinking tokens. Re-FORC trains a lightweight adapter on reasoning models, demonstrating improved prediction with longer reasoning and larger models. Re-FORC enables: 1) early stopping of unpromising reasoning chains, reducing compute by 26% while maintaining accuracy, 2) optimized model and thinking length selection that achieves 4% higher accuracy at equal compute and 55% less compute at equal accuracy compared to the largest model, 3) adaptive test-time scaling, which increases accuracy by 11% in high compute regime, and 7% in low compute regime. Re-FORC allows dynamic reasoning with length control via cost-per-token thresholds while estimating computation time upfront.", "authors": ["Renos Zabounidis", "Aditya Golatkar", "Michael Kleinman", "Alessandro Achille", "Wei Xia", "Stefano Soatto"], "categories": ["cs.AI", "cs.LG"], "fields_of_study": ["Computer Science"], "published_date": "2025-11-03", "url": "https://arxiv.org/abs/2511.02130", "pdf_url": "https://arxiv.org/pdf/2511.02130v1", "arxiv_id": "2511.02130", "doi": "10.48550/arXiv.2511.02130", "citation_count": 2, "influential_citation_count": 0, "has_code": false, "code_url": null, "venue": "arXiv.org", "quality_score": 0.3575}
{"id": "15e1a44ac5d4b43b21148c396ea9285fe825ac9ca10d8986214b5a3d2a222a5e", "sources": ["arxiv", "semantic_scholar"], "title": "Scaling Graph Chain-of-Thought Reasoning: A Multi-Agent Framework with Efficient LLM Serving", "abstract": "Graph Chain-of-Thought (Graph-CoT) enables large language models (LLMs) to perform step-by-step reasoning over graph-structured knowledge, but existing pipelines suffer from low accuracy, excessive token usage, high latency, and low throughput due to single-agent monolithic prompts, repeated context re-encoding, and inefficient serving execution. We present GLM, the first multi-agent Graph-CoT system co-designed with an optimized LLM serving architecture. GLM decomposes reasoning into specialized agents for classification, reasoning, action generation, and graph retrieval, enabling branching and selective context sharing to reduce prompt length and reasoning iterations while preserving reasoning quality, thereby improving accuracy and reducing overall token consumption. To scale inference, we introduce a Graph-CoT-aware LLM inference mechanism with graph-specific KV-cache management, priority-based eviction, and pipelined execution to improve serving efficiency. Experiments demonstrate that GLM improves answer accuracy by up to 38%, reduces token cost by up to 95.7%, lowers inference latency by 90.3%, and achieves up to 15.1x higher throughput compared to state-of-the-art Graph-CoT baselines, enabling efficient adoption for complex real-world reasoning at scale.", "authors": ["Chengying Huan", "Ziheng Meng", "Yongchao Liu", "Zhengyi Yang", "Yun Zhu", "Yue Yun", "Shipeng Li", "Rong Gu", "Xiabao Wu", "Haitao Zhang", "Chuntao Hong", "Shaonan Ma", "Guihai Chen", "Chen Tian"], "categories": ["cs.LG", "cs.AI"], "fields_of_study": ["Computer Science"], "published_date": "2025-11-03", "url": "https://arxiv.org/abs/2511.01633", "pdf_url": "https://arxiv.org/pdf/2511.01633v1", "arxiv_id": "2511.01633", "doi": "10.48550/arXiv.2511.01633", "citation_count": 5, "influential_citation_count": 0, "has_code": false, "code_url": null, "venue": "arXiv.org", "quality_score": 0.3575}
{"id": "66994dea891f2af79299a9cd30ac1f0e27f41c6016318ebe2ec1ee0ba4d58393", "sources": ["arxiv", "semantic_scholar"], "title": "Inference-Time Chain-of-Thought Pruning with Latent Informativeness Signals", "abstract": "Large language models (LLMs) improve reasoning accuracy when generating multiple candidate solutions at test time, but standard methods like Best-of-N (BoN) incur high computational cost by fully generating all branches. Self-Truncation Best-of-N (ST-BoN) mitigates this by truncating unpromising paths early, but its reliance on consistency-based heuristics is a limitation as it does not directly evaluate branch quality. We present KL-Adjusted Pruned Path Algorithm (KAPPA), an inference-time method that combines Kullback-Leibler divergence, confidence, and entropy into a principled scoring function to guide progressive pruning. By promoting diversity during exploration and selectively eliminating low-scoring branches, KAPPA maintains accuracy while substantially reducing memory and token usage. Experiments on GSM8K and MATH500 with DeepSeek-R1-Distill-Qwen-1.5B and Qwen2.5-7B-Instruct demonstrate that KAPPA stabilizes performance in smaller models and achieves up to ~60% reduction in peak memory and ~90% reduction in total token generation relative to BoN, with minimal impact on accuracy.", "authors": ["Sophie Li", "Nicholas Huang", "Nayan Saxena", "Nina Luo", "Vincent Lin", "Kevin Zhu", "Sunishchal Dev"], "categories": ["cs.LG"], "fields_of_study": ["Computer Science"], "published_date": "2025-11-01", "url": "https://arxiv.org/abs/2511.00699", "pdf_url": "https://arxiv.org/pdf/2511.00699v2", "arxiv_id": "2511.00699", "doi": "10.48550/arXiv.2511.00699", "citation_count": 1, "influential_citation_count": 0, "has_code": false, "code_url": null, "venue": "arXiv.org", "quality_score": 0.3552}
{"id": "96bd7c39750feff44218d5b86bdf427badf16ed3e98659186efbc5d6dab3b044", "sources": ["arxiv", "semantic_scholar"], "title": "Reasoning Models Sometimes Output Illegible Chains of Thought", "abstract": "Language models trained via outcome-based reinforcement learning (RL) to reason using chain-of-thought (CoT) have shown remarkable performance. Monitoring such a model's CoT may allow us to understand its intentions and detect potential malicious behavior. However, to be effective, this requires that CoTs are legible and faithful. We study CoT legibility across 14 reasoning models, finding that RL often causes reasoning to become illegible to both humans and AI monitors, with reasoning models (except Claude) generating illegible CoTs while returning to perfectly readable final answers. We show that models use illegible reasoning to reach correct answers (accuracy dropping by 53\\% when forced to use only legible portions), yet find no correlation between legibility and performance when resampling - suggesting the relationship is more nuanced. We also find that legibility degrades on harder questions. We discuss potential hypotheses for these results, including steganography, training artifacts, and vestigial tokens. These results suggest that without explicit optimization for legibility, outcome-based RL naturally produces models with increasingly opaque reasoning processes, potentially undermining monitoring approaches.", "authors": ["Arun Jose"], "categories": ["cs.LG"], "fields_of_study": ["Computer Science"], "published_date": "2025-10-31", "url": "https://arxiv.org/abs/2510.27338", "pdf_url": "https://arxiv.org/pdf/2510.27338v1", "arxiv_id": "2510.27338", "doi": "10.48550/arXiv.2510.27338", "citation_count": 7, "influential_citation_count": 1, "has_code": false, "code_url": null, "venue": "arXiv.org", "quality_score": 0.3541}
{"id": "21c0fc8d983ae80039f3fe6e382247070c09e31e8c5fac33da26d29a98ba213a", "sources": ["arxiv", "semantic_scholar"], "title": "Diffuse Thinking: Exploring Diffusion Language Models as Efficient Thought Proposers for Reasoning", "abstract": "In recent years, large language models (LLMs) have witnessed remarkable advancements, with the test-time scaling law consistently enhancing the reasoning capabilities. Through systematic evaluation and exploration of a diverse spectrum of intermediate thoughts, LLMs demonstrate the potential to generate deliberate reasoning steps, thereby substantially enhancing reasoning accuracy. However, LLMs' autoregressive generation paradigm results in reasoning performance scaling sub-optimally with test-time computation, often requiring excessive computational overhead to propose thoughts while yielding only marginal performance gains. In contrast, diffusion language models (DLMs) can efficiently produce diverse samples through parallel denoising in a single forward pass, inspiring us to leverage them for proposing intermediate thoughts, thereby alleviating the computational burden associated with autoregressive generation while maintaining quality. In this work, we propose an efficient collaborative reasoning framework, leveraging DLMs to generate candidate thoughts and LLMs to evaluate their quality. Experiments across diverse benchmarks demonstrate that our framework achieves strong performance in complex reasoning tasks, offering a promising direction for future research. Our code is open-source at https://anonymous.4open.science/r/Diffuse-Thinking-EC60.", "authors": ["Chenyang Shao", "Sijian Ren", "Fengli Xu", "Yong Li"], "categories": ["cs.CL"], "fields_of_study": ["Computer Science"], "published_date": "2025-10-31", "url": "https://arxiv.org/abs/2510.27469", "pdf_url": "https://arxiv.org/pdf/2510.27469v1", "arxiv_id": "2510.27469", "doi": "10.48550/arXiv.2510.27469", "citation_count": 4, "influential_citation_count": 0, "has_code": true, "code_url": null, "venue": "arXiv.org", "quality_score": 0.5472}
{"id": "7da963ec9bb8667fa42b94ccf8c4055a5d523295cf0aff8b2eb136e9fb952e4f", "sources": ["arxiv", "semantic_scholar"], "title": "ThinkMorph: Emergent Properties in Multimodal Interleaved Chain-of-Thought Reasoning", "abstract": "Multimodal reasoning requires iterative coordination between language and vision, yet it remains unclear what constitutes a meaningful interleaved chain of thought. We posit that text and image thoughts should function as complementary rather than isomorphic modalities that mutually advance reasoning. Guided by this principle, we build ThinkMorph, a unified model fine-tuned on approximately 24K high-quality interleaved reasoning traces spanning tasks with varying visual engagement. ThinkMorph learns to generate progressive text-image reasoning steps that concretely manipulate visual content while maintaining coherent verbal logic. It delivers large gains on vision-centric benchmarks (averaging 34.7 percent over the base model) and generalizes to out-of-domain tasks, matching or surpassing larger and proprietary VLMs. Beyond performance, ThinkMorph exhibits emergent multimodal intelligence, including unseen visual manipulation skills, adaptive switching between reasoning modes, and better test-time scaling through diversified multimodal thoughts. These findings suggest promising directions for characterizing the emergent capabilities of unified models for multimodal reasoning.", "authors": ["Jiawei Gu", "Yunzhuo Hao", "Huichen Will Wang", "Linjie Li", "Michael Qizhe Shieh", "Yejin Choi", "Ranjay Krishna", "Yu Cheng"], "categories": ["cs.CV"], "fields_of_study": ["Computer Science"], "published_date": "2025-10-30", "url": "https://arxiv.org/abs/2510.27492", "pdf_url": "https://arxiv.org/pdf/2510.27492v3", "arxiv_id": "2510.27492", "doi": "10.48550/arXiv.2510.27492", "citation_count": 31, "influential_citation_count": 5, "has_code": false, "code_url": null, "venue": "arXiv.org", "quality_score": 0.3891}
{"id": "9f732b17e9601e8bb0e5408f32f7dd29ccb006f10f8113d75dd2b9cfe3a2dad8", "sources": ["arxiv", "semantic_scholar"], "title": "Chopping Trees: Semantic Similarity Based Dynamic Pruning for Tree-of-Thought Reasoning", "abstract": "Tree-of-Thought (ToT) reasoning boosts the problem-solving abilities of Large Language Models (LLMs) but is computationally expensive due to semantic redundancy, where distinct branches explore equivalent reasoning paths. We introduce Semantic Similarity-Based Dynamic Pruning (SSDP), a lightweight method that, to the best of our knowledge, is the first framework to integrate online semantic merging into parallelized tree search, enabling the clustering and pruning of redundant steps in real time. Across reasoning benchmarks, including GSM8K and MATH500, SSDP achieves up to a 2.3x speedup over state-of-the-art tree-search baselines while maintaining competitive accuracy (typically within 5% of the strongest baseline) and reducing the number of explored nodes by 85-90%, demonstrating a practical approach to efficient, scalable LLM reasoning. The implementation of SSDP is publicly available at https://github.com/kimjoonghokim/SSDP.", "authors": ["Joongho Kim", "Xirui Huang", "Zarreen Reza", "Gabriel Grand"], "categories": ["cs.CL", "cs.AI"], "fields_of_study": ["Computer Science"], "published_date": "2025-10-30", "url": "https://arxiv.org/abs/2511.08595", "pdf_url": "https://arxiv.org/pdf/2511.08595v2", "arxiv_id": "2511.08595", "doi": "10.48550/arXiv.2511.08595", "citation_count": 2, "influential_citation_count": 1, "has_code": true, "code_url": "https://github.com/kimjoonghokim/SSDP", "venue": "arXiv.org", "quality_score": 0.5454}
{"id": "797449e3ab2288258e52b18a384461dae94dd7157c57fffe226140bae5a2f21b", "sources": ["arxiv", "semantic_scholar"], "title": "Latent Chain-of-Thought for Visual Reasoning", "abstract": "Chain-of-thought (CoT) reasoning is critical for improving the interpretability and reliability of Large Vision-Language Models (LVLMs). However, existing training algorithms such as SFT, PPO, and GRPO may not generalize well across unseen reasoning tasks and heavily rely on a biased reward model. To address this challenge, we reformulate reasoning in LVLMs as posterior inference and propose a scalable training algorithm based on amortized variational inference. By leveraging diversity-seeking reinforcement learning algorithms, we introduce a novel sparse reward function for token-level learning signals that encourage diverse, high-likelihood latent CoT, overcoming deterministic sampling limitations and avoiding reward hacking. Additionally, we implement a Bayesian inference-scaling strategy that replaces costly Best-of-N and Beam Search with a marginal likelihood to efficiently rank optimal rationales and answers. We empirically demonstrate that the proposed method enhances the state-of-the-art LVLMs on seven reasoning benchmarks, in terms of effectiveness, generalization, and interpretability.", "authors": ["Guohao Sun", "Hang Hua", "Jian Wang", "Jiebo Luo", "Sohail Dianat", "Majid Rabbani", "Raghuveer Rao", "Zhiqiang Tao"], "categories": ["cs.AI", "cs.CL"], "fields_of_study": ["Computer Science", "Medicine"], "published_date": "2025-10-27", "url": "https://arxiv.org/abs/2510.23925", "pdf_url": "https://arxiv.org/pdf/2510.23925v2", "arxiv_id": "2510.23925", "doi": "10.48550/arXiv.2510.23925", "citation_count": 24, "influential_citation_count": 0, "has_code": false, "code_url": null, "venue": "Advances in Neural Information Processing Systems", "quality_score": 0.3495}
{"id": "b73027b0b25e280060d8002b21f0af09cadc30ce78b06418dc78c40d7845348f", "sources": ["arxiv", "semantic_scholar"], "title": "When Fewer Layers Break More Chains: Layer Pruning Harms Test-Time Scaling in LLMs", "abstract": "Layer pruning has emerged as a widely adopted technique for improving the efficiency of large language models (LLMs). Although existing methods demonstrate strong performance retention on general knowledge tasks, their effect on long-chain reasoning, a more brittle yet crucial capability, remains largely unexplored. In this work, we study the impact of layer pruning on long-chain reasoning through the lens of test-time scaling, a key mechanism in modern LLMs that enables strong reasoning capacity by allocating more computation at inference time. With extensive experiments, we demonstrate that pruning even one or two layers can severely impair test-time scaling, with performance collapsing drastically on long reasoning benchmarks even when performance on knowledge-intensive and shallow reasoning tasks remains stable. Furthermore, we find that standard supervised fine-tuning remedies fail to recover test-time scaling once it has deteriorated. Through in-depth analyses, we identify the mechanisms underlying this fragility of test-time scaling and highlight the fundamental risks of applying layer pruning to reasoning-intensive LLMs. These findings call for a rethinking of layer pruning strategies and provide insights for developing methods that preserve the robustness of reasoning. We open-source the codebase in \\href{https://github.com/keyu-wang-2002/Layer-Pruning-Harms-Inference-Scaling}{https://github.com/keyu-wang-2002/Layer-Pruning-Harms-Inference-Scaling}.", "authors": ["Keyu Wang", "Tian Lyu", "Guinan Su", "Jonas Geiping", "Lu Yin", "Marco Canini", "Shiwei Liu"], "categories": ["cs.LG", "cs.AI"], "fields_of_study": ["Computer Science"], "published_date": "2025-10-25", "url": "https://arxiv.org/abs/2510.22228", "pdf_url": "https://arxiv.org/pdf/2510.22228v1", "arxiv_id": "2510.22228", "doi": "10.48550/arXiv.2510.22228", "citation_count": 5, "influential_citation_count": 2, "has_code": true, "code_url": "https://github.com/keyu-wang-2002/Layer-Pruning-Harms-Inference-Scaling}{https://github.com/keyu-wang-2002/Layer-Pruning-Harms-Inference-Scaling}", "venue": "arXiv.org", "quality_score": 0.5366}
{"id": "206c0fec87578f8a8ff5b2fd5f7a8db81f5d11af5e9d515ac327372d11134c26", "sources": ["arxiv", "semantic_scholar"], "title": "The Virtues of Brevity: Avoid Overthinking in Parallel Test-Time Reasoning", "abstract": "Reasoning models represent a significant advance in LLM capabilities, particularly for complex reasoning tasks such as mathematics and coding. Previous studies confirm that parallel test-time compute-sampling multiple solutions and selecting the best one-can further enhance the predictive performance of LLMs. However, strategies in this area often require complex scoring, thus increasing computational cost and complexity. In this work, we demonstrate that the simple and counterintuitive heuristic of selecting the shortest solution is highly effective. We posit that the observed effectiveness stems from models operating in two distinct regimes: a concise, confident conventional regime and a verbose overthinking regime characterized by uncertainty, and we show evidence of a critical point where the overthinking regime begins to be significant. By selecting the shortest answer, the heuristic preferentially samples from the conventional regime. We confirm that this approach is competitive with more complex methods such as self-consistency across two challenging benchmarks while significantly reducing computational overhead. The shortest-answer heuristic provides a Pareto improvement over self-consistency and applies even to tasks where output equality is not well defined.", "authors": ["Raul Cavalcante Dinardi", "Bruno Yamamoto", "Anna Helena Reali Costa", "Artur Jordao"], "categories": ["cs.LG"], "fields_of_study": ["Computer Science"], "published_date": "2025-10-24", "url": "https://arxiv.org/abs/2510.21067", "pdf_url": "https://arxiv.org/pdf/2510.21067v1", "arxiv_id": "2510.21067", "doi": "10.48550/arXiv.2510.21067", "citation_count": 0, "influential_citation_count": 0, "has_code": false, "code_url": null, "venue": "arXiv.org", "quality_score": 0.346}
{"id": "05a11242200d29acd433c401cb75ac372b588c5dc87b4f08fc129f61e72dc451", "sources": ["arxiv", "semantic_scholar"], "title": "The Zero-Step Thinking: An Empirical Study of Mode Selection as Harder Early Exit in Reasoning Models", "abstract": "Reasoning models have demonstrated exceptional performance in tasks such as mathematics and logical reasoning, primarily due to their ability to engage in step-by-step thinking during the reasoning process. However, this often leads to overthinking, resulting in unnecessary computational overhead. To address this issue, Mode Selection aims to automatically decide between Long-CoT (Chain-of-Thought) or Short-CoT by utilizing either a Thinking or NoThinking mode. Simultaneously, Early Exit determines the optimal stopping point during the iterative reasoning process. Both methods seek to reduce the computational burden. In this paper, we first identify Mode Selection as a more challenging variant of the Early Exit problem, as they share similar objectives but differ in decision timing. While Early Exit focuses on determining the best stopping point for concise reasoning at inference time, Mode Selection must make this decision at the beginning of the reasoning process, relying on pre-defined fake thoughts without engaging in an explicit reasoning process, referred to as zero-step thinking. Through empirical studies on nine baselines, we observe that prompt-based approaches often fail due to their limited classification capabilities when provided with minimal hand-crafted information. In contrast, approaches that leverage internal information generally perform better across most scenarios but still exhibit issues with stability. Our findings indicate that existing methods relying solely on the information provided by models are insufficient for effectively addressing Mode Selection in scenarios with limited information, highlighting the ongoing challenges of this task. Our code is available at https://github.com/Trae1ounG/Zero_Step_Thinking.", "authors": ["Yuqiao Tan", "Shizhu He", "Kang Liu", "Jun Zhao"], "categories": ["cs.AI", "cs.CL"], "fields_of_study": ["Computer Science"], "published_date": "2025-10-22", "url": "https://arxiv.org/abs/2510.19176", "pdf_url": "https://arxiv.org/pdf/2510.19176v1", "arxiv_id": "2510.19176", "doi": "10.48550/arXiv.2510.19176", "citation_count": 4, "influential_citation_count": 0, "has_code": true, "code_url": "https://github.com/Trae1ounG/Zero_Step_Thinking", "venue": "arXiv.org", "quality_score": 0.5312}
{"id": "1a2f955b846e5935392bd44f1a17ff0179cac30fe05630a0a70631d3de92a672", "sources": ["arxiv", "semantic_scholar"], "title": "SmartSwitch: Advancing LLM Reasoning by Overcoming Underthinking via Promoting Deeper Thought Exploration", "abstract": "The long chain-of-thought (LongCoT) capability is central to the recent breakthroughs achieved by large language models in complex reasoning tasks. However, the accompanying issue of ''underthinking'', where models exhibit shallow reasoning by frequently switching thoughts without sufficient exploration, limits both performance and token efficiency. To address this problem, we propose a simple yet effective reasoning strategy: the SmartSwitch inference framework. This framework can be easily integrated into any large language model as a plug-and-play solution, continuously monitoring the model's reasoning process to detect underthinking and guide it toward deeper exploration of promising but overlooked thoughts. Specifically, the perception module identifies points where thoughts switch and evaluates the potential of the preceding thought using an off-the-shelf process reward model (PRM). If a high-potential thought is found to be prematurely abandoned, the intervention module interrupts the ongoing inference, backtracks to the point before the switch, and inserts a \"deepening prompt\" to encourage further exploration along that promising path. Extensive experiments on challenging mathematical reasoning benchmarks demonstrate that our method significantly enhances the performance of various large language models of different sizes.", "authors": ["Xichen Zhang", "Sitong Wu", "Haoru Tan", "Shaozuo Yu", "Yinghao Zhu", "Ziyi He", "Jiaya Jia"], "categories": ["cs.CL", "cs.AI", "cs.LG"], "fields_of_study": ["Computer Science"], "published_date": "2025-10-22", "url": "https://arxiv.org/abs/2510.19767", "pdf_url": "https://arxiv.org/pdf/2510.19767v1", "arxiv_id": "2510.19767", "doi": "10.48550/arXiv.2510.19767", "citation_count": 1, "influential_citation_count": 0, "has_code": true, "code_url": "https://github.com/dvlab-research/SmartSwitch", "venue": "arXiv.org", "quality_score": 0.5312}
{"id": "de6ce1c329fd310a9f590c5d8bf687e235600ff832627449084adf14b4c5e2ac", "sources": ["arxiv", "semantic_scholar"], "title": "Can Reasoning Models Obfuscate Reasoning? Stress-Testing Chain-of-Thought Monitorability", "abstract": "Recent findings suggest that misaligned models may exhibit deceptive behavior, raising concerns about output trustworthiness. Chain-of-thought (CoT) is a promising tool for alignment monitoring: when models articulate their reasoning faithfully, monitors can detect and mitigate harmful behaviors before undesirable outcomes occur. However, a key uncertainty is: Can models obfuscate their CoT in order to pursue hidden adversarial objectives while evading detection? To answer this question and thus stress-test CoT monitorability, we develop a composable and quantifiable taxonomy of prompts to elicit CoT obfuscation. We evaluate both internal CoT (reasoning traces) and external CoT (prompted reasoning in outputs) using toy tasks and more realistic environments in SHADE-Arena. We show that: (i) CoT monitoring performs accurately and efficiently without obfuscation pressure. (ii) Under strong obfuscation pressure, some models successfully complete adversarial tasks while evading detection. (iii) Models do not obfuscate their internal CoT as much as their external CoT (under prompt pressure). These results suggest that while CoT provides valuable oversight in benign settings, robust deployment requires model-specific stress-testing of monitorability.", "authors": ["Artur Zolkowski", "Wen Xing", "David Lindner", "Florian Tramèr", "Erik Jenner"], "categories": ["cs.CR", "cs.AI"], "fields_of_study": ["Computer Science"], "published_date": "2025-10-21", "url": "https://arxiv.org/abs/2510.19851", "pdf_url": "https://arxiv.org/pdf/2510.19851v1", "arxiv_id": "2510.19851", "doi": "10.48550/arXiv.2510.19851", "citation_count": 10, "influential_citation_count": 1, "has_code": false, "code_url": null, "venue": "arXiv.org", "quality_score": 0.3426}
{"id": "037446ba4b5209794e8852cd6f323ba33ef2b3663a358ac8f53c9c5b288feef3", "sources": ["arxiv", "semantic_scholar"], "title": "Fine-Tuned Thoughts: Leveraging Chain-of-Thought Reasoning for Industrial Asset Health Monitoring", "abstract": "Small Language Models (SLMs) are becoming increasingly popular in specialized fields, such as industrial applications, due to their efficiency, lower computational requirements, and ability to be fine-tuned for domain-specific tasks, enabling accurate and cost-effective solutions. However, performing complex reasoning using SLMs in specialized fields such as Industry 4.0 remains challenging. In this paper, we propose a knowledge distillation framework for industrial asset health, which transfers reasoning capabilities via Chain-of-Thought (CoT) distillation from Large Language Models (LLMs) to smaller, more efficient models (SLMs). We discuss the advantages and the process of distilling LLMs using multi-choice question answering (MCQA) prompts to enhance reasoning and refine decision-making. We also perform in-context learning to verify the quality of the generated knowledge and benchmark the performance of fine-tuned SLMs with generated knowledge against widely used LLMs. The results show that the fine-tuned SLMs with CoT reasoning outperform the base models by a significant margin, narrowing the gap to their LLM counterparts. Our code is open-sourced at: https://github.com/IBM/FailureSensorIQ.", "authors": ["Shuxin Lin", "Dhaval Patel", "Christodoulos Constantinides"], "categories": ["cs.CL", "cs.AI"], "fields_of_study": ["Computer Science"], "published_date": "2025-10-21", "url": "https://arxiv.org/abs/2510.18817", "pdf_url": "https://arxiv.org/pdf/2510.18817v1", "arxiv_id": "2510.18817", "doi": "10.18653/v1/2025.findings-emnlp.1126", "citation_count": 2, "influential_citation_count": 0, "has_code": true, "code_url": "https://github.com/IBM/FailureSensorIQ", "venue": "Conference on Empirical Methods in Natural Language Processing", "quality_score": 0.5295}
{"id": "351218de380050c6bf0a44289042e0f37cca3bdcd89953e6921724648a84f533", "sources": ["arxiv", "semantic_scholar"], "title": "Inference-Time Compute Scaling For Flow Matching", "abstract": "Allocating extra computation at inference time has recently improved sample quality in large language models and diffusion-based image generation. In parallel, Flow Matching (FM) has gained traction in language, vision, and scientific domains, but inference-time scaling methods for it remain under-explored. Concurrently, Kim et al., 2025 approach this problem but replace the linear interpolant with a non-linear variance-preserving (VP) interpolant at inference, sacrificing FM's efficient and straight sampling. Additionally, inference-time compute scaling for flow matching has only been applied to visual tasks, like image generation. We introduce novel inference-time scaling procedures for FM that preserve the linear interpolant during sampling. Evaluations of our method on image generation, and for the first time (to the best of our knowledge), unconditional protein generation, show that I) sample quality consistently improves as inference compute increases, and II) flow matching inference-time scaling can be applied to scientific domains.", "authors": ["Adam Stecklov", "Noah El Rimawi-Fine", "Mathieu Blanchette"], "categories": ["cs.LG"], "fields_of_study": ["Computer Science"], "published_date": "2025-10-20", "url": "https://arxiv.org/abs/2510.17786", "pdf_url": "https://arxiv.org/pdf/2510.17786v1", "arxiv_id": "2510.17786", "doi": "10.48550/arXiv.2510.17786", "citation_count": 1, "influential_citation_count": 0, "has_code": false, "code_url": null, "venue": "arXiv.org", "quality_score": 0.3415}
{"id": "c9e6e4e90a270cad29999fc2bc013f88efbfb7c724f9b953ef24e2b8e559f451", "sources": ["arxiv", "semantic_scholar"], "title": "Big Reasoning with Small Models: Instruction Retrieval at Inference Time", "abstract": "Small language models (SLMs) enable low-cost, private, on-device inference, but they often fail on problems that require specialized domain knowledge or multi-step reasoning. Existing approaches for improving reasoning either rely on scale (e.g., chain-of-thought prompting), require task-specific training that limits reuse and generality (e.g., distillation), or retrieve unstructured information that still leaves the SLM to determine an appropriate reasoning strategy. We propose instruction retrieval, an inference-time intervention that augments an SLM with structured, reusable reasoning procedures rather than raw passages. We construct an Instruction Corpus by clustering similar training questions and using a teacher model to generate generalizable guides that pair domain background with explicit step-by-step procedures. At inference, the SLM retrieves the instructions most relevant to a given query and executes the associated procedures without any additional fine-tuning. Across three challenging domains: medicine, law, and mathematics, instruction retrieval yields consistent gains for models with at least 3B parameters, improving accuracy by 9.4%, 7.9%, and 5.1%, respectively, with the strongest 14B model surpassing GPT-4o's zero-shot performance on knowledge-intensive tasks.", "authors": ["Kenan Alkiek", "David Jurgens", "Vinod Vydiswaran"], "categories": ["cs.CL", "cs.AI"], "fields_of_study": ["Computer Science"], "published_date": "2025-10-15", "url": "https://arxiv.org/abs/2510.13935", "pdf_url": "https://arxiv.org/pdf/2510.13935v2", "arxiv_id": "2510.13935", "doi": "10.48550/arXiv.2510.13935", "citation_count": 0, "influential_citation_count": 0, "has_code": false, "code_url": null, "venue": "arXiv.org", "quality_score": 0.3357}
{"id": "9664d40ccc24dfdaa486e2ab999de9f732249458b3fa043f05babf3027ab24e9", "sources": ["arxiv", "semantic_scholar"], "title": "Putting on the Thinking Hats: A Survey on Chain of Thought Fine-tuning from the Perspective of Human Reasoning Mechanism", "abstract": "Chain of thought (CoT) fine-tuning aims to endow large language models (LLMs) with reasoning capabilities by training them on curated reasoning traces. It leverages both supervised and reinforced fine-tuning to cultivate human-like reasoning skills in LLMs, including detailed planning, divergent thinking, intuitive judgment, timely reflection, internal thinking, and fact perception, etc. As CoT fine-tuning has advanced, LLMs have demonstrated substantial improvements in tasks such as mathematical reasoning and code generation. However, existing surveys about CoT fine-tuning primarily focus on technical aspects and overlook a systematic analysis from the perspective of human reasoning mechanisms. Given that the ultimate goal of CoT fine-tuning is to enable LLMs to reason like humans, it is crucial to investigate this technique through the lens of human cognition. To fill this gap, we present the first comprehensive survey of CoT fine-tuning grounded in human reasoning theory. Specifically, inspired by the well-known Six Thinking Hats framework, which systematically characterizes common human thinking modes using six metaphorical hats, we classify and examine CoT fine-tuning methods through this lens. Furthermore, building upon this theory, we outline potential directions for future research in CoT fine-tuning. In addition, we compile a comprehensive overview of existing datasets and model performances, and a real-time GitHub repository \\footnote{https://github.com/AI-Chen/Awesome-CoT-Finetuning} that continuously tracks recent advances in this area is maintained. We hope this survey will serve as a valuable resource to inspire innovation and foster progress in this rapidly evolving field.", "authors": ["Xiaoshu Chen", "Sihang Zhou", "Ke Liang", "Duanyang Yuan", "Haoyuan Chen", "Xiaoyu Sun", "Lingyuan Meng", "Xinwang Liu"], "categories": ["cs.CL"], "fields_of_study": ["Computer Science"], "published_date": "2025-10-15", "url": "https://arxiv.org/abs/2510.13170", "pdf_url": "https://arxiv.org/pdf/2510.13170v2", "arxiv_id": "2510.13170", "doi": "10.48550/arXiv.2510.13170", "citation_count": 4, "influential_citation_count": 0, "has_code": true, "code_url": "https://github.com/AI-Chen/Awesome-CoT-Finetuning}", "venue": "arXiv.org", "quality_score": 0.5189}
{"id": "1696ed340c4f9f1aececed948345da02b6ee41d12aca8b8f7abb17d0233d86fb", "sources": ["arxiv", "semantic_scholar"], "title": "Refine Thought: A Test-Time Inference Method for Embedding Model Reasoning", "abstract": "We propose RT (Refine Thought), a method that can enhance the semantic rea-soning ability of text embedding models. The method obtains the final semanticrepresentation by running multiple forward passes of the text embedding model.Experiments show that RT achieves significant improvements on semantic reason-ing tasks in BRIGHT and the person job matching benchmark PJBenchmark1, while maintaining consistent performance on general-purpose semantic under-standing tasks such as C-MTEB. Our results indicate that RT is effective becauseit further activates the semantic reasoning ability learned during pretraining bydecoder-only text embedding models(e.g., Qwen3-Embedding-8B). RT canbe seen as a test-time inference method.", "authors": ["Guangzhi Wang", "Kai Li", "Yinghao Jiao", "Zhi Liu"], "categories": ["cs.CL", "cs.AI", "cs.IR"], "fields_of_study": ["Computer Science"], "published_date": "2025-10-14", "url": "https://arxiv.org/abs/2511.13726", "pdf_url": "https://arxiv.org/pdf/2511.13726v1", "arxiv_id": "2511.13726", "doi": "10.48550/arXiv.2511.13726", "citation_count": 2, "influential_citation_count": 0, "has_code": false, "code_url": null, "venue": "arXiv.org", "quality_score": 0.3346}
{"id": "f1a79737ccafb295191359ef98c49b4ba1445a95b7a78f192167c4254586bb4f", "sources": ["arxiv", "semantic_scholar"], "title": "HiCoTraj:Zero-Shot Demographic Reasoning via Hierarchical Chain-of-Thought Prompting from Trajectory", "abstract": "Inferring demographic attributes such as age, sex, or income level from human mobility patterns enables critical applications such as targeted public health interventions, equitable urban planning, and personalized transportation services. Existing mobility-based demographic inference studies heavily rely on large-scale trajectory data with demographic labels, leading to limited interpretability and poor generalizability across different datasets and user groups. We propose HiCoTraj (Zero-Shot Demographic Reasoning via Hierarchical Chain-of-Thought Prompting from Trajectory), a framework that leverages LLMs' zero-shot learning and semantic understanding capabilities to perform demographic inference without labeled training data. HiCoTraj transforms trajectories into semantically rich, natural language representations by creating detailed activity chronicles and multi-scale visiting summaries. Then HiCoTraj uses a novel hierarchical chain of thought reasoning to systematically guide LLMs through three cognitive stages: factual feature extraction, behavioral pattern analysis, and demographic inference with structured output. This approach addresses the scarcity challenge of labeled demographic data while providing transparent reasoning chains. Experimental evaluation on real-world trajectory data demonstrates that HiCoTraj achieves competitive performance across multiple demographic attributes in zero-shot scenarios.", "authors": ["Junyi Xie", "Yuankun Jiao", "Jina Kim", "Yao-Yi Chiang", "Lingyi Zhao", "Khurram Shafique"], "categories": ["cs.AI"], "fields_of_study": ["Computer Science"], "published_date": "2025-10-14", "url": "https://arxiv.org/abs/2510.12067", "pdf_url": "https://arxiv.org/pdf/2510.12067v1", "arxiv_id": "2510.12067", "doi": "10.1145/3764915.3770723", "citation_count": 0, "influential_citation_count": 0, "has_code": false, "code_url": null, "venue": null, "quality_score": 0.2129}
{"id": "7530d3b1852b100f13ee058c6f3769bbc01b3c79a03b9993aa82dd0f1e3a1312", "sources": ["arxiv", "semantic_scholar"], "title": "Towards Inference-time Scaling for Continuous Space Reasoning", "abstract": "Inference-time scaling through multiple sample generation in combination with Process- or Outcome-Reward Model (PRM or ORM) re-ranking has proven effective for text-based reasoning in large language models. This paper investigates whether such established techniques can be successfully adapted to reasoning in the continuous space, using COCONUT (Hao et al. 2024) continuous space reasoning LM as the backbone. We demonstrate the feasibility of generating diverse reasoning paths through dropout-based sampling. Our Pass@N analysis on the generated samples reveals the potential that could enable a significant gain in performance akin to observed gain in the discrete space. However, we highlight unique challenges faced for materializing this gain in the continuous thought space. In particular, working recipes for data generation and training PRM and ORM models in the discrete space unlocks only marginal improvements in the continuous space. Through probing various aspects including geometric properties and trajectory dynamics we identify the underlying reasons that prevent effective discrimination between correct and incorrect reasoning (essential for the functioning of PRM and ORM). Our findings reveal that current limitations stem from the absence of key inductive biases in continuous thought representations. We argue that the training frameworks for continuous reasoning LMs require not only to optimize for accuracy but also to explicitly incorporate inductive biases that could be utilized during inference-time for discrimination of correct and incorrect thoughts.\\footnote{Our code and data will be publicly available.}", "authors": ["Minghan Wang", "Thuy-Trang Vu", "Ehsan Shareghi", "Gholamreza Haffari"], "categories": ["cs.CL"], "fields_of_study": ["Computer Science"], "published_date": "2025-10-14", "url": "https://arxiv.org/abs/2510.12167", "pdf_url": "https://arxiv.org/pdf/2510.12167v2", "arxiv_id": "2510.12167", "doi": "10.48550/arXiv.2510.12167", "citation_count": 3, "influential_citation_count": 0, "has_code": false, "code_url": null, "venue": "arXiv.org", "quality_score": 0.3346}
{"id": "4759d56853651779c5e8300c9c084422ec588d1f3cf91ef6725d72a84f2b077a", "sources": ["arxiv", "semantic_scholar"], "title": "Enhancing Long Chain-of-Thought Reasoning through Multi-Path Plan Aggregation", "abstract": "Inference-time scaling enhances the reasoning ability of a language model (LM) by extending its chain-of-thought (CoT). However, existing approaches typically generate the entire reasoning chain in a single forward pass, which often leads to CoT derailment, i.e., the reasoning trajectory drifting off course due to compounding errors. This problem is particularly severe for smaller LMs with long CoTs due to their limited capacity. To address this, we analyze raw long CoTs and uncover a reasoning hierarchy consisting of planning and execution steps. Our analysis reveals that most reasoning errors stem from incorrect planning. Motivated by this observation, we propose Multi-Path Plan Aggregation (MPPA), a framework that augments single-pass reasoning with plan exploration and aggregation. Following a variable interval schedule based on the token position, MPPA generates multiple candidate plans and aggregates them into a refined planning step. To maintain efficiency, we adopt a minimal design in which the base LM serves as the primary policy, while a lightweight LoRA module implements the plan aggregation policy. We further observe that outcome-reward RL is inefficient for long trajectories (e.g., exceeding 4K tokens). To overcome this, we introduce online Step-DPO, a process-level preference optimization scheme that leverages Twisted Sequential Monte Carlo (TSMC) to provide scalable stepwise supervision using small LMs. This yields more efficient training, improved stability, and higher accuracy. Extensive experiments on challenging math, science, and logical reasoning benchmarks demonstrate that, with only 10% SFT data and 5% of preference pairs, our method outperforms both the DeepSeek-R1 distillation baseline and the outcome-reward RL baseline across multiple base models and tasks.", "authors": ["Siheng Xiong", "Ali Payani", "Faramarz Fekri"], "categories": ["cs.CL"], "fields_of_study": ["Computer Science"], "published_date": "2025-10-13", "url": "https://arxiv.org/abs/2510.11620", "pdf_url": "https://arxiv.org/pdf/2510.11620v2", "arxiv_id": "2510.11620", "doi": "10.48550/arXiv.2510.11620", "citation_count": 6, "influential_citation_count": 0, "has_code": false, "code_url": null, "venue": "arXiv.org", "quality_score": 0.3334}
{"id": "663e58bdb0545fc8a5aee57462cca199ff52402002263a2ea7622b738fb579be", "sources": ["arxiv", "semantic_scholar"], "title": "CoPRS: Learning Positional Prior from Chain-of-Thought for Reasoning Segmentation", "abstract": "Existing works on reasoning segmentation either connect hidden features from a language model directly to a mask decoder or represent positions in text, which limits interpretability and semantic detail. To solve this, we present CoPRS, a Multi-modal Chain-of-Thought (MCoT)-based positional perception model that bridges language reasoning to segmentation through a differentiable and interpretable positional prior instantiated as a heatmap. By making the reasoning process clear via MCoT and expressing it as a dense, differentiable heatmap, this interface enhances interpretability and diagnostic analysis and yields more concentrated evidence on the target. A learnable concentration token aggregates features of the image and reasoning text to generate this positional prior, which is decoded to precise masks through a lightweight decoder, providing a direct connection between reasoning and segmentation. Across the RefCOCO series and ReasonSeg, CoPRS matches or surpasses the best reported metrics on each standard split under comparable protocols, with performance at or above the prior state of the art across both validation and test partitions. Extensive experiments demonstrate a strong positive correlation among the CoT trajectory, the generated heatmap, and the decoded mask, supporting an interpretable alignment between the reasoning output and downstream mask generation. Collectively, these findings support the utility of this paradigm in bridging reasoning and segmentation and show advantages in concentration driven by reasoning and in more precise mask prediction. Code has been released at https://github.com/ZhenyuLU-Heliodore/CoPRS.", "authors": ["Zhenyu Lu", "Liupeng Li", "Jinpeng Wang", "Yan Feng", "Bin Chen", "Ke Chen", "Yaowei Wang"], "categories": ["cs.CV", "cs.MM"], "fields_of_study": ["Computer Science"], "published_date": "2025-10-13", "url": "https://arxiv.org/abs/2510.11173", "pdf_url": "https://arxiv.org/pdf/2510.11173v3", "arxiv_id": "2510.11173", "doi": "10.48550/arXiv.2510.11173", "citation_count": 7, "influential_citation_count": 1, "has_code": true, "code_url": "https://github.com/ZhenyuLU-Heliodore/CoPRS", "venue": "arXiv.org", "quality_score": 0.5153}
{"id": "e5bec2a2dc68878c9cc3c338d03f3f18a16b473cdd9ea51d90707b2acb36539d", "sources": ["arxiv", "semantic_scholar"], "title": "Review of Inference-Time Scaling Strategies: Reasoning, Search and RAG", "abstract": "The performance gains of LLMs have historically been driven by scaling up model size and training data. However, the rapidly diminishing availability of high-quality training data is introducing a fundamental bottleneck, shifting the focus of research toward inference-time scaling. This paradigm uses additional computation at the time of deployment to substantially improve LLM performance on downstream tasks without costly model re-training. This review systematically surveys the diverse techniques contributing to this new era of inference-time scaling, organizing the rapidly evolving field into two comprehensive perspectives: Output-focused and Input-focused methods. Output-focused techniques encompass complex, multi-step generation strategies, including reasoning (e.g., CoT, ToT, ReAct), various search and decoding methods (e.g., MCTS, beam search), training for long CoT (e.g., RLVR, GRPO), and model ensemble methods. Input-focused techniques are primarily categorized by few-shot and RAG, with RAG as the central focus. The RAG section is further detailed through a structured examination of query expansion, data, retrieval and reranker, LLM generation methods, and multi-modal RAG.", "authors": ["Zhichao Wang", "Cheng Wan", "Dong Nie"], "categories": ["cs.CL"], "fields_of_study": ["Computer Science"], "published_date": "2025-10-12", "url": "https://arxiv.org/abs/2510.10787", "pdf_url": "https://arxiv.org/pdf/2510.10787v1", "arxiv_id": "2510.10787", "doi": "10.48550/arXiv.2510.10787", "citation_count": 0, "influential_citation_count": 0, "has_code": false, "code_url": null, "venue": "arXiv.org", "quality_score": 0.3323}
{"id": "0a3cf254180c3e57ee98db84e7911c033f330a3c7df8ce2c40437e6ac8c1bb2d", "sources": ["arxiv", "semantic_scholar"], "title": "Adaptive Dual Reasoner: Large Reasoning Models Can Think Efficiently by Hybrid Reasoning", "abstract": "Although Long Reasoning Models (LRMs) have achieved superior performance on various reasoning scenarios, they often suffer from increased computational costs and inference latency caused by overthinking. To address these limitations, we propose Adaptive Dual Reasoner, which supports two reasoning modes: fast thinking and slow thinking. ADR dynamically alternates between these modes based on the contextual complexity during reasoning. ADR is trained in two stages: (1) A cold-start stage using supervised fine-tuning (SFT) to equip the model with the ability to integrate both fast and slow reasoning modes, in which we construct a hybrid reasoning dataset through a dedicated pipeline to provide large-scale supervision. (2) A reinforcement learning stage for optimizing reasoning effort, where we introduce Entropy-guided Hybrid Policy Optimization EHPO, an RL training framework employing an entropy-guided dynamic rollout strategy for branching at high-entropy units and a difficulty-aware penalty to balance fast and slow reasoning. Across challenging mathematical reasoning benchmarks, ADR achieves an effective balance between reasoning performance and efficiency among state-of-the-art approaches. Specifically, ADR yields a performance gain of up to 6.1%, while reducing the reasoning output length by 49.5% to 59.3%.", "authors": ["Yujian Zhang", "Keyu Chen", "Zhifeng Shen", "Ruizhi Qiao", "Xing Sun"], "categories": ["cs.AI"], "fields_of_study": ["Computer Science"], "published_date": "2025-10-11", "url": "https://arxiv.org/abs/2510.10207", "pdf_url": "https://arxiv.org/pdf/2510.10207v2", "arxiv_id": "2510.10207", "doi": "10.48550/arXiv.2510.10207", "citation_count": 2, "influential_citation_count": 1, "has_code": false, "code_url": null, "venue": "arXiv.org", "quality_score": 0.3311}
{"id": "1c7e9d3af53639fd5e068b5a1da5439155dae0dbb04749b6e31e71d1cbb503c3", "sources": ["arxiv", "semantic_scholar"], "title": "MatryoshkaThinking: Recursive Test-Time Scaling Enables Efficient Reasoning", "abstract": "Test-time scaling has emerged as a promising paradigm in language modeling, wherein additional computational resources are allocated during inference to enhance model performance. Recent approaches, such as DeepConf, have demonstrated the efficacy of this strategy, however, they often incur substantial computational overhead to achieve competitive results. In this work, we propose MatryoshkaThinking, a novel method that significantly reduces computational cost while maintaining state-of-the-art performance. Specifically, MatryoshkaThinking attains a score of 99.79 on AIME2025 using only 4% of the computation required by DeepConf. The core of our approach lies in the recursive exploitation of the model's intrinsic capabilities in reasoning, verification, and summarization, which collectively enhance the retention of correct solutions and reduce the disparity between Pass@k and Pass@1. Comprehensive evaluations across multiple open-source models and challenging multi-modal reasoning benchmarks validate the effectiveness and generality of our method. These findings offer new insights into the design of efficient and scalable test-time inference strategies for advanced language models.", "authors": ["Hongwei Chen", "Yishu Lei", "Dan Zhang", "Bo Ke", "Danxiang Zhu", "Xuyi Chen", "Yuxiang Lu", "Zhengjie Huang", "Shikun Feng", "Jingzhou He", "Yu Sun", "Hua Wu", "Haifeng Wang"], "categories": ["cs.CL", "cs.AI"], "fields_of_study": ["Computer Science"], "published_date": "2025-10-11", "url": "https://arxiv.org/abs/2510.10293", "pdf_url": "https://arxiv.org/pdf/2510.10293v1", "arxiv_id": "2510.10293", "doi": "10.48550/arXiv.2510.10293", "citation_count": 1, "influential_citation_count": 1, "has_code": true, "code_url": null, "venue": "arXiv.org", "quality_score": 0.5118}
{"id": "ff6d9005d733592d4c819f99aa54c50aa0a28d5242a0f3a23ddf9b2097ba09b5", "sources": ["arxiv", "semantic_scholar"], "title": "Stop When Enough: Adaptive Early-Stopping for Chain-of-Thought Reasoning", "abstract": "Chain-of-Thought (CoT) reasoning has driven recent gains of large language models (LLMs) on reasoning-intensive tasks by externalizing intermediate steps. However, excessive or redundant reasoning -- so-called overthinking -- can increase inference costs and lead LLMs toward incorrect conclusions. In this paper, we present REFRAIN ($\\underline{REF}$lective-$\\underline{R}$edundancy for $\\underline{A}$daptive $\\underline{IN}$ference), a training-free framework that adaptively determines when to stop reasoning to mitigate overthinking. REFRAIN integrates a two-stage stop discriminator to identify reflective yet redundant reasoning and a sliding-window Upper Confidence Bound (SW-UCB) multi-armed bandit controller to dynamically adjust stopping thresholds according to problem difficulty without supervision or fine-tuning. Across four representative benchmarks and two model families, REFRAIN reduces token usage by 20-55% while maintaining or improving accuracy compared to standard CoT prompting. Extensive ablation and robustness analyses demonstrate its stability across models, scorers, and prompt variations. In summary, our findings highlight when-to-stop as a new and practical axis of test-time scaling -- enabling models to reason not just more, but just enough.", "authors": ["Renliang Sun", "Wei Cheng", "Dawei Li", "Haifeng Chen", "Wei Wang"], "categories": ["cs.CL"], "fields_of_study": ["Computer Science"], "published_date": "2025-10-11", "url": "https://arxiv.org/abs/2510.10103", "pdf_url": "https://arxiv.org/pdf/2510.10103v1", "arxiv_id": "2510.10103", "doi": "10.48550/arXiv.2510.10103", "citation_count": 9, "influential_citation_count": 1, "has_code": false, "code_url": null, "venue": "arXiv.org", "quality_score": 0.3311}
{"id": "4ff9c55f61843a23110d4d8b3b6d492dcb9a866e332658def60b91ea98d1f269", "sources": ["arxiv", "semantic_scholar"], "title": "Prompting Test-Time Scaling Is A Strong LLM Reasoning Data Augmentation", "abstract": "Large language models (LLMs) have demonstrated impressive reasoning capabilities when provided with chain-of-thought exemplars, but curating large reasoning datasets remains laborious and resource-intensive. In this work, we introduce Prompting Test-Time Scaling (P-TTS), a simple yet effective inference-time data augmentation strategy for enhancing LLM reasoning through finetuning. Rather than collecting thousands or even millions of examples, P-TTS leverages a small pool of only 90 manually selected reasoning instances and systematically varies exemplar augmentation through principled instruction prompting intensities at test time to synthesize diverse reasoning trajectory contexts. Then we finetune the various sizes of Qwen-2.5 models on P-TTS data. Across a suite of mathematical reasoning AIME2024 & 25, MATH500, and GPQA-Diamond, our P-TTS-7B and 32B models outperform the prior competitive baselines like S1 and S1.1 (1K-shot), achieving absolute accuracy gains of +26.66% and +30.00% on AIME'24 (7B), and +13.34% and +6.67% on AIME'25 (7B); P-TTS-32B yields gains of +23.33% and +16.63% on AIME'24, and +26.63% and +3.33% on AIME'25 (vs. S1 and S1.1, respectively), with comparable or better performance on MATH500 and GPQA-Diamond. We further show that P-TTS enhances zero-shot generalization accuracy on out-of-domain reasoning benchmarks of Gaokao, Kaoyan, OlympiadBench, AMC23, GradeSchoolMath, and Minerva. Our analysis suggests that test-time scaling effectively explores the latent space of reasoning patterns, amplifying LLM problem-solving with minimal annotation overhead, and further unlocking the reasoning potential and capabilities of LLMs. Prompting Test-Time Scaling offers a practical, low-cost way to elicit LLM reasoning in resource-constrained or rapidly evolving domains.", "authors": ["Sondos Mahmoud Bsharat", "Zhiqiang Shen"], "categories": ["cs.CL", "cs.AI", "cs.LG"], "fields_of_study": ["Computer Science"], "published_date": "2025-10-10", "url": "https://arxiv.org/abs/2510.09599", "pdf_url": "https://arxiv.org/pdf/2510.09599v1", "arxiv_id": "2510.09599", "doi": "10.48550/arXiv.2510.09599", "citation_count": 2, "influential_citation_count": 0, "has_code": true, "code_url": "https://github.com/VILA-Lab/PTTS", "venue": "arXiv.org", "quality_score": 0.51}
{"id": "67b74f94bb473a1f633a57fe03abaa8762e7742b81192eecce6ebd8b643babbb", "sources": ["arxiv", "semantic_scholar"], "title": "Verifying Chain-of-Thought Reasoning via Its Computational Graph", "abstract": "Current Chain-of-Thought (CoT) verification methods predict reasoning correctness based on outputs (black-box) or activations (gray-box), but offer limited insight into why a computation fails. We introduce a white-box method: Circuit-based Reasoning Verification (CRV). We hypothesize that attribution graphs of correct CoT steps, viewed as execution traces of the model's latent reasoning circuits, possess distinct structural fingerprints from those of incorrect steps. By training a classifier on structural features of these graphs, we show that these traces contain a powerful signal of reasoning errors. Our white-box approach yields novel scientific insights unattainable by other methods. (1) We demonstrate that structural signatures of error are highly predictive, establishing the viability of verifying reasoning directly via its computational graph. (2) We find these signatures to be highly domain-specific, revealing that failures in different reasoning tasks manifest as distinct computational patterns. (3) We provide evidence that these signatures are not merely correlational; by using our analysis to guide targeted interventions on individual transcoder features, we successfully correct the model's faulty reasoning. Our work shows that, by scrutinizing a model's computational process, we can move from simple error detection to a deeper, causal understanding of LLM reasoning.", "authors": ["Zheng Zhao", "Yeskendir Koishekenov", "Xianjun Yang", "Naila Murray", "Nicola Cancedda"], "categories": ["cs.CL", "cs.AI", "cs.LG"], "fields_of_study": ["Computer Science"], "published_date": "2025-10-10", "url": "https://arxiv.org/abs/2510.09312", "pdf_url": "https://arxiv.org/pdf/2510.09312v2", "arxiv_id": "2510.09312", "doi": "10.48550/arXiv.2510.09312", "citation_count": 17, "influential_citation_count": 3, "has_code": false, "code_url": null, "venue": "arXiv.org", "quality_score": 0.33}
{"id": "6d68fa5c0b3bf80e87c5957d0ed9712a6d7d97aa833740d55832e0d1b968abed", "sources": ["arxiv", "semantic_scholar"], "title": "Parallel Test-Time Scaling for Latent Reasoning Models", "abstract": "Parallel test-time scaling (TTS) is a pivotal approach for enhancing large language models (LLMs), typically by sampling multiple token-based chains-of-thought in parallel and aggregating outcomes through voting or search. Recent advances in latent reasoning, where intermediate reasoning unfolds in continuous vector spaces, offer a more efficient alternative to explicit Chain-of-Thought, yet whether such latent models can similarly benefit from parallel TTS remains open, mainly due to the absence of sampling mechanisms in continuous space, and the lack of probabilistic signals for advanced trajectory aggregation. This work enables parallel TTS for latent reasoning models by addressing the above issues. For sampling, we introduce two uncertainty-inspired stochastic strategies: Monte Carlo Dropout and Additive Gaussian Noise. For aggregation, we design a Latent Reward Model (LatentRM) trained with step-wise contrastive objective to score and guide latent reasoning. Extensive experiments and visualization analyses show that both sampling strategies scale effectively with compute and exhibit distinct exploration dynamics, while LatentRM enables effective trajectory selection. Together, our explorations open a new direction for scalable inference in continuous spaces. Code and checkpoints released at https://github.com/ModalityDance/LatentTTS", "authors": ["Runyang You", "Yongqi Li", "Meng Liu", "Wenjie Wang", "Liqiang Nie", "Wenjie Li"], "categories": ["cs.CL", "cs.AI", "cs.LG"], "fields_of_study": ["Computer Science"], "published_date": "2025-10-09", "url": "https://arxiv.org/abs/2510.07745", "pdf_url": "https://arxiv.org/pdf/2510.07745v4", "arxiv_id": "2510.07745", "doi": "10.48550/arXiv.2510.07745", "citation_count": 8, "influential_citation_count": 1, "has_code": true, "code_url": "https://github.com/ModalityDance/LatentTTS", "venue": "arXiv.org", "quality_score": 0.5082}
{"id": "964ffb2f9b369104723c431b884722c9267fdc1e5fdd6fafc17cbb97291ff9df", "sources": ["arxiv", "semantic_scholar"], "title": "Encode, Think, Decode: Scaling test-time reasoning with recursive latent thoughts", "abstract": "Most efforts to improve the reasoning capabilities of large language models (LLMs) involve either scaling the number of parameters and the size of training data, or scaling inference computation by letting models generate complex chains of thought. Motivated by interpretability studies showing that the crucial computation required for reasoning tasks is concentrated in a limited range of layers, we introduce Encode-Think-Decode (ETD), a method that enhances the reasoning capabilities of a base model by training it to iterate over a small subset of reasoning-relevant layers during the mid-training stage. ETD amplifies latent reasoning while preserving the original architecture, parameter count, hyperparameters, and training data composition. When iterating on the selected layers at inference time, ETD models yield substantial gains on 17 reasoning benchmarks, including +28.4% relative accuracy improvement on GSM8K and +36% on MATH with the OLMo-2 1B Base model. We also explore an adaptive depth strategy that adjusts the computation per input token. Our results show that recursive latent reasoning offers a simple and effective path to stronger LLM reasoning.", "authors": ["Yeskendir Koishekenov", "Aldo Lipani", "Nicola Cancedda"], "categories": ["cs.LG", "cs.AI", "cs.CL"], "fields_of_study": ["Computer Science"], "published_date": "2025-10-08", "url": "https://arxiv.org/abs/2510.07358", "pdf_url": "https://arxiv.org/pdf/2510.07358v1", "arxiv_id": "2510.07358", "doi": "10.48550/arXiv.2510.07358", "citation_count": 15, "influential_citation_count": 1, "has_code": false, "code_url": null, "venue": "arXiv.org", "quality_score": 0.3277}
{"id": "09e562383364c44da5e4e408b90e5f9a8f178ed88753fba116d859d672376add", "sources": ["arxiv", "semantic_scholar"], "title": "TS-Agent: Understanding and Reasoning Over Raw Time Series via Iterative Insight Gathering", "abstract": "Large language models (LLMs) exhibit strong symbolic and compositional reasoning, yet they struggle with time series question answering as the data is typically transformed into an LLM-compatible modality, e.g., serialized text, plotted images, or compressed time series embeddings. Such conversions impose representation bottlenecks, often require cross-modal alignment or finetuning, and can exacerbate hallucination and knowledge leakage. To address these limitations, we propose TS-Agent, an agentic, tool-grounded framework that uses LLMs strictly for iterative evidence-based reasoning, while delegating statistical and structural extraction to time series analytical tools operating on raw sequences. Our framework solves time series tasks through an evidence-driven agentic process: (1) it alternates between thinking, tool execution, and observation in a ReAct-style loop, (2) records intermediate results in an explicit evidence log and corrects the reasoning trace via a self-refinement critic, and (3) enforces a final answer-verification step to prevent hallucinations and leakage. Across four benchmarks spanning time series understanding and reasoning, TS-Agent matches or exceeds strong text-based, vision-based, and time-series language model baselines, with the largest gains on reasoning tasks where multimodal LLMs are prone to hallucination and knowledge leakage in zero-shot settings.", "authors": ["Penghang Liu", "Elizabeth Fons", "Annita Vapsi", "Mohsen Ghassemi", "Svitlana Vyetrenko", "Daniel Borrajo", "Vamsi K. Potluru", "Manuela Veloso"], "categories": ["cs.AI"], "fields_of_study": ["Computer Science"], "published_date": "2025-10-08", "url": "https://arxiv.org/abs/2510.07432", "pdf_url": "https://arxiv.org/pdf/2510.07432v2", "arxiv_id": "2510.07432", "doi": null, "citation_count": 5, "influential_citation_count": 0, "has_code": false, "code_url": null, "venue": null, "quality_score": 0.2085}
{"id": "ab392f0312bc097731e85582121fad7107fc078b8503fd2274eb553e8d99d339", "sources": ["arxiv", "semantic_scholar"], "title": "ARISE: An Adaptive Resolution-Aware Metric for Test-Time Scaling Evaluation in Large Reasoning Models", "abstract": "Test-time scaling has emerged as a transformative paradigm for enhancing the performance of large reasoning models, enabling dynamic allocation of computational resources during inference. However, as the landscape of reasoning models rapidly expands, a critical question remains: how can we systematically compare and evaluate the test-time scaling capabilities across different models? In this paper, we introduce ARISE (Adaptive Resolution-aware Scaling Evaluation), a novel metric specifically designed to assess the test-time scaling effectiveness of large reasoning models. Unlike existing evaluation approaches, ARISE incorporates two key innovations: (1) sample-level awareness that effectively penalizes negative scaling behaviors where increased computation leads to performance degradation, and (2) a dynamic sampling mechanism that mitigates the impact of accuracy fluctuations and token count instability on the final assessment. We conduct comprehensive experiments evaluating state-of-the-art reasoning models across diverse domains including mathematical reasoning, code generation, and agentic tasks. Our results demonstrate that ARISE provides a reliable and fine-grained measurement of test-time scaling capabilities, revealing significant variations in scaling efficiency across models. Notably, our evaluation identifies Claude Opus as exhibiting superior scaling characteristics compared to other contemporary reasoning models.", "authors": ["Zhangyue Yin", "Qiushi Sun", "Zhiyuan Zeng", "Zhiyuan Yu", "Qipeng Guo", "Xuanjing Huang", "Xipeng Qiu"], "categories": ["cs.AI"], "fields_of_study": ["Computer Science"], "published_date": "2025-10-07", "url": "https://arxiv.org/abs/2510.06014", "pdf_url": "https://arxiv.org/pdf/2510.06014v1", "arxiv_id": "2510.06014", "doi": "10.48550/arXiv.2510.06014", "citation_count": 1, "influential_citation_count": 0, "has_code": false, "code_url": null, "venue": "arXiv.org", "quality_score": 0.3266}
{"id": "04bb0bb2ea623a194258974920cbb4f14fc46e69e16154c285cb27302b39e5e2", "sources": ["arxiv", "semantic_scholar"], "title": "Test-Time Scaling of Reasoning Models for Machine Translation", "abstract": "Test-time scaling (TTS) has enhanced the performance of Reasoning Models (RMs) on various tasks such as math and coding, yet its efficacy in machine translation (MT) remains underexplored. This paper investigates whether increased inference-time computation improves translation quality. We evaluate 12 RMs across a diverse suite of MT benchmarks spanning multiple domains, examining three scenarios: direct translation, forced-reasoning extrapolation, and post-editing. Our findings show that for general-purpose RMs, TTS provides limited and inconsistent benefits for direct translation, with performance quickly plateauing. However, the effectiveness of TTS is unlocked by domain-specific fine-tuning, which aligns a model's reasoning process with task requirements, leading to consistent improvements up to an optimal, self-determined reasoning depth. We also find that forcing a model to reason beyond its natural stopping point consistently degrades translation quality. In contrast, TTS proves highly effective in a post-editing context, reliably turning self-correction into a beneficial process. These results indicate that the value of inference-time computation in MT lies not in enhancing single-pass translation with general models, but in targeted applications like multi-step, self-correction workflows and in conjunction with task-specialized models.", "authors": ["Zihao Li", "Shaoxiong Ji", "Jörg Tiedemann"], "categories": ["cs.CL"], "fields_of_study": ["Computer Science"], "published_date": "2025-10-07", "url": "https://arxiv.org/abs/2510.06471", "pdf_url": "https://arxiv.org/pdf/2510.06471v2", "arxiv_id": "2510.06471", "doi": "10.48550/arXiv.2510.06471", "citation_count": 3, "influential_citation_count": 0, "has_code": false, "code_url": null, "venue": "Conference of the European Chapter of the Association for Computational Linguistics", "quality_score": 0.3266}
{"id": "ab937065e6d5a3521849961af1e6065cdf7a1ff385f54775f1a4dd21d97c56c0", "sources": ["arxiv", "semantic_scholar"], "title": "The Valley of Code Reasoning: Scaling Knowledge Distillation of Large Language Models", "abstract": "Distilling the thinking traces of a Large Language Model (LLM) with reasoning capabilities into a smaller model has been proven effective. Yet, there is a scarcity of work done on how model performances scale with the quantity of distillation data. In this work, we study the scaling trend of distilling competitive coding skills on two small non-reasoning LLMs. We validate the hypothesis that there is a $\\textit{valley of code reasoning}$: downstream performance on competitive coding first drops as data quantity increases, then it steadily increases in a sharper-than-log-linear fashion. Having identified the trend, we further fine-tune the models at two different distillation stages on the same data to ground conclusions on their respective learning phases. We learn that across stages in the low and medium-low data regimes, small models benefit significantly from easier coding questions than from harder ones. We also find that, surprisingly, the correctness of outputs in training data makes no difference to distillation outcomes. Our work represents a step forward in understanding the training dynamics of code reasoning distillation outside intuition", "authors": ["Muyu He", "Muhammad Ali Shafique", "Anand Kumar", "Tsach Mackey", "Nazneen Rajani"], "categories": ["cs.CL"], "fields_of_study": ["Computer Science"], "published_date": "2025-10-07", "url": "https://arxiv.org/abs/2510.06101", "pdf_url": "https://arxiv.org/pdf/2510.06101v1", "arxiv_id": "2510.06101", "doi": "10.48550/arXiv.2510.06101", "citation_count": 0, "influential_citation_count": 0, "has_code": false, "code_url": null, "venue": "arXiv.org", "quality_score": 0.3266}
{"id": "0efedbcd75c1655e621e130fbc4f660f777581155bddafbcec00a3cbd3637348", "sources": ["arxiv", "semantic_scholar"], "title": "VChain: Chain-of-Visual-Thought for Reasoning in Video Generation", "abstract": "Recent video generation models can produce smooth and visually appealing clips, but they often struggle to synthesize complex dynamics with a coherent chain of consequences. Accurately modeling visual outcomes and state transitions over time remains a core challenge. In contrast, large language and multimodal models (e.g., GPT-4o) exhibit strong visual state reasoning and future prediction capabilities. To bridge these strengths, we introduce VChain, a novel inference-time chain-of-visual-thought framework that injects visual reasoning signals from multimodal models into video generation. Specifically, VChain contains a dedicated pipeline that leverages large multimodal models to generate a sparse set of critical keyframes as snapshots, which are then used to guide the sparse inference-time visual-state adaptation of a pre-trained video generator only at these key moments. Our approach is tuning-efficient, introduces minimal overhead and avoids dense supervision. Extensive experiments on complex, multi-step scenarios show that VChain significantly enhances the quality of generated videos.", "authors": ["Ziqi Huang", "Ning Yu", "Gordon Chen", "Haonan Qiu", "Paul Debevec", "Ziwei Liu"], "categories": ["cs.CV"], "fields_of_study": ["Computer Science"], "published_date": "2025-10-06", "url": "https://arxiv.org/abs/2510.05094", "pdf_url": "https://arxiv.org/pdf/2510.05094v2", "arxiv_id": "2510.05094", "doi": "10.48550/arXiv.2510.05094", "citation_count": 22, "influential_citation_count": 0, "has_code": false, "code_url": null, "venue": "arXiv.org", "quality_score": 0.3404}
{"id": "550572696168a1b7cac328d9a248dfc79d18ca7f7cf47b8814551e02639d4782", "sources": ["arxiv", "semantic_scholar"], "title": "Thinking on the Fly: Test-Time Reasoning Enhancement via Latent Thought Policy Optimization", "abstract": "Recent advancements in Large Language Models (LLMs) have shifted from explicit Chain-of-Thought (CoT) reasoning to more efficient latent reasoning, where intermediate thoughts are represented as vectors rather than text. However, latent reasoning can be brittle on challenging, out-of-distribution tasks where robust reasoning is most critical. To overcome these limitations, we introduce Latent Thought Policy Optimization (LTPO), a parameter-free framework that enhances LLM reasoning entirely at test time, without requiring model parameter updates. LTPO treats intermediate latent \"thought\" vectors as dynamic parameters that are actively optimized for each problem instance. It employs an online policy gradient method guided by an intrinsic, confidence-based reward signal computed directly from the frozen LLM's own output distributions, eliminating the need for external supervision or expensive text generation during optimization. Extensive experiments on five reasoning benchmarks show that LTPO not only matches or surpasses strong baselines on standard tasks but also demonstrates remarkable robustness where others fail. Most notably, on highly challenging AIME benchmarks where existing latent reasoning baselines collapse to near-zero accuracy, LTPO delivers substantial improvements, showcasing a unique capability for complex reasoning.", "authors": ["Wengao Ye", "Yan Liang", "Lianlei Shan"], "categories": ["cs.CL", "cs.AI"], "fields_of_study": ["Computer Science"], "published_date": "2025-10-05", "url": "https://arxiv.org/abs/2510.04182", "pdf_url": "https://arxiv.org/pdf/2510.04182v4", "arxiv_id": "2510.04182", "doi": "10.48550/arXiv.2510.04182", "citation_count": 6, "influential_citation_count": 1, "has_code": false, "code_url": null, "venue": "arXiv.org", "quality_score": 0.3243}
{"id": "0f995e04e5a4aa896ccfd961cb22c128fbaa2790a578065c10f3d66c98976dc6", "sources": ["arxiv", "semantic_scholar"], "title": "Scaling Code-Assisted Chain-of-Thoughts and Instructions for Model Reasoning", "abstract": "Reasoning capability is pivotal for Large Language Models (LLMs) to solve complex tasks, yet achieving reliable and scalable reasoning remains challenging. While Chain-of-Thought (CoT) prompting has become a mainstream approach, existing methods often suffer from uncontrolled generation, insufficient quality, and limited diversity in reasoning paths. Recent efforts leverage code to enhance CoT by grounding reasoning in executable steps, but such methods are typically constrained to predefined mathematical problems, hindering scalability and generalizability. In this work, we propose Caco (Code-Assisted Chain-of-ThOught), a novel framework that automates the synthesis of high-quality, verifiable, and diverse instruction-CoT reasoning data through code-driven augmentation. Unlike prior work, Caco first fine-tunes a code-based CoT generator on existing math and programming solutions in a unified code format, then scales the data generation to a large amount of diverse reasoning traces. Crucially, we introduce automated validation via code execution and rule-based filtering to ensure logical correctness and structural diversity, followed by reverse-engineering filtered outputs into natural language instructions and language CoTs to enrich task adaptability. This closed-loop process enables fully automated, scalable synthesis of reasoning data with guaranteed executability. Experiments on our created Caco-1.3M dataset demonstrate that Caco-trained models achieve strong competitive performance on mathematical reasoning benchmarks, outperforming existing strong baselines. Further analysis reveals that Caco's code-anchored verification and instruction diversity contribute to superior generalization across unseen tasks. Our work establishes a paradigm for building self-sustaining, trustworthy reasoning systems without human intervention.", "authors": ["Honglin Lin", "Qizhi Pei", "Xin Gao", "Zhuoshi Pan", "Yu Li", "Juntao Li", "Conghui He", "Lijun Wu"], "categories": ["cs.CL", "cs.PL"], "fields_of_study": ["Computer Science"], "published_date": "2025-10-05", "url": "https://arxiv.org/abs/2510.04081", "pdf_url": "https://arxiv.org/pdf/2510.04081v1", "arxiv_id": "2510.04081", "doi": "10.48550/arXiv.2510.04081", "citation_count": 10, "influential_citation_count": 1, "has_code": false, "code_url": null, "venue": "arXiv.org", "quality_score": 0.3243}
{"id": "90732757a55f26c6464d272da90b41c4e8dd1195f9dcb9a1ef513a55c4559668", "sources": ["arxiv", "semantic_scholar"], "title": "Understanding the Role of Training Data in Test-Time Scaling", "abstract": "Test-time scaling improves the reasoning capabilities of large language models (LLMs) by allocating extra compute to generate longer Chains-of-Thoughts (CoTs). This enables models to tackle more complex problem by breaking them down into additional steps, backtracking, and correcting mistakes. Despite its strong performance--demonstrated by OpenAI's o1 and DeepSeek R1, the conditions in the training data under which long CoTs emerge, and when such long CoTs improve the performance, remain unclear. In this paper, we study the performance of test-time scaling for transformers trained on an in-context weight prediction task for linear regression. Our analysis provides a theoretical explanation for several intriguing observations: First, at any fixed test error, increasing test-time compute allows us to reduce the number of in-context examples (context length) in training prompts. Second, if the skills required to solve a downstream task are not sufficiently present in the training data, increasing test-time compute can harm performance. Finally, we characterize task hardness via the smallest eigenvalue of its feature covariance matrix and show that training on a diverse, relevant, and hard set of tasks results in best performance for test-time scaling. We confirm our findings with experiments on large, nonlinear transformer architectures.", "authors": ["Adel Javanmard", "Baharan Mirzasoleiman", "Vahab Mirrokni"], "categories": ["cs.AI", "cs.LG", "stat.ML"], "fields_of_study": ["Computer Science", "Mathematics"], "published_date": "2025-10-04", "url": "https://arxiv.org/abs/2510.03605", "pdf_url": "https://arxiv.org/pdf/2510.03605v2", "arxiv_id": "2510.03605", "doi": "10.48550/arXiv.2510.03605", "citation_count": 4, "influential_citation_count": 1, "has_code": false, "code_url": null, "venue": "arXiv.org", "quality_score": 0.3231}
{"id": "0c0e005ea5a1b8cefa96413a2a1735da5aeaa46c0092e7b3e38b4bbac298356b", "sources": ["arxiv", "semantic_scholar"], "title": "RoiRL: Efficient, Self-Supervised Reasoning with Offline Iterative Reinforcement Learning", "abstract": "Reinforcement learning (RL) is central to improving reasoning in large language models (LLMs) but typically requires ground-truth rewards. Test-Time Reinforcement Learning (TTRL) removes this need by using majority-vote rewards, but relies on heavy online RL and incurs substantial computational cost. We propose RoiRL: Reasoning with offline iterative Reinforcement Learning, a family of lightweight offline learning alternatives that can target the same regularized optimal policies. Unlike TTRL, RoiRL eliminates the need to maintain a reference model and instead optimizes weighted log-likelihood objectives, enabling stable training with significantly lower memory and compute requirements. Experimental results show that RoiRL trains to 2.5x faster and consistently outperforms TTRL on reasoning benchmarks, establishing a scalable path to self-improving LLMs without labels.", "authors": ["Aleksei Arzhantsev", "Otmane Sakhi", "Flavian Vasile"], "categories": ["cs.LG"], "fields_of_study": ["Computer Science"], "published_date": "2025-10-03", "url": "https://arxiv.org/abs/2510.02892", "pdf_url": "https://arxiv.org/pdf/2510.02892v1", "arxiv_id": "2510.02892", "doi": "10.48550/arXiv.2510.02892", "citation_count": 1, "influential_citation_count": 0, "has_code": false, "code_url": null, "venue": "arXiv.org", "quality_score": 0.322}
{"id": "83a245811c6cfec335c78e82cb17472128ef6b91ce6a4231303e1ade93a2128d", "sources": ["arxiv", "semantic_scholar"], "title": "Eliciting Chain-of-Thought Reasoning for Time Series Analysis using Reinforcement Learning", "abstract": "Complex numerical time series analysis often demands multi-step reasoning capabilities beyond current models' reach. Tasks like medical diagnosis and weather forecasting require sequential reasoning processes - including counterfactual analysis, logical deduction, knowledge application, and multi-modal contextual integration - that existing time series models cannot explicitly perform. While recent research has shown large language models (LLMs) can achieve sophisticated Chain-of-Thought (CoT) reasoning through reinforcement learning (RL), these advances have primarily focused on mathematical and coding domains, with LLMs still demonstrating poor performance on time series tasks. We introduce Chain Of thought for Understanding Numerical Time Series (COUNTS), the first framework that trains LLMs to perform CoT reasoning across diverse time series tasks using RL with verifiable rewards. Our approach employs a Residual Vector-Quantized VAE to create high-fidelity discrete tokens that seamlessly integrate into a pre-trained LLM's vocabulary. COUNTS undergoes a two-stage training process: first, supervised fine-tuning on time series analysis tasks to master our novel representations, followed by Group Relative Policy Optimization training on verifiable problems using prompting strategies that encourage explicit reasoning steps before producing final answers. Our experiments demonstrate that this RL-driven approach with intermediate CoT reasoning significantly enhances LLM performance across various time series analysis tasks, opening new possibilities for complex temporal data reasoning.", "authors": ["Felix Parker", "Nimeesha Chan", "Chi Zhang", "Kimia Ghobadi"], "categories": ["cs.LG"], "fields_of_study": ["Computer Science"], "published_date": "2025-10-01", "url": "https://arxiv.org/abs/2510.01116", "pdf_url": "https://arxiv.org/pdf/2510.01116v2", "arxiv_id": "2510.01116", "doi": "10.48550/arXiv.2510.01116", "citation_count": 5, "influential_citation_count": 0, "has_code": false, "code_url": null, "venue": "arXiv.org", "quality_score": 0.3197}
{"id": "5842002303bbb250f6acd4286b76df6e5a57d20d0be2ff6203afcacf98c2d269", "sources": ["arxiv", "semantic_scholar"], "title": "Typed Chain-of-Thought: A Curry-Howard Framework for Verifying LLM Reasoning", "abstract": "While Chain-of-Thought (CoT) prompting enhances the reasoning capabilities of large language models, the faithfulness of the generated rationales remains an open problem for model interpretability. We propose a novel theoretical lens for this problem grounded in the Curry-Howard correspondence, which posits a direct relationship between formal proofs and computer programs. Under this paradigm, a faithful reasoning trace is analogous to a well-typed program, where each intermediate step corresponds to a typed logical inference. We operationalise this analogy, presenting methods to extract and map the informal, natural language steps of CoT into a formal, typed proof structure. Successfully converting a CoT trace into a well-typed proof serves as a strong, verifiable certificate of its computational faithfulness, moving beyond heuristic interpretability towards formal verification. Our framework provides a methodology to transform plausible narrative explanations into formally verifiable programs, offering a path towards building more reliable and trustworthy AI systems.", "authors": ["Elija Perrier"], "categories": ["cs.AI"], "fields_of_study": ["Computer Science"], "published_date": "2025-10-01", "url": "https://arxiv.org/abs/2510.01069", "pdf_url": "https://arxiv.org/pdf/2510.01069v1", "arxiv_id": "2510.01069", "doi": "10.48550/arXiv.2510.01069", "citation_count": 3, "influential_citation_count": 0, "has_code": false, "code_url": null, "venue": "arXiv.org", "quality_score": 0.3197}
{"id": "2ca47670badc8501fb5f6092359cd95a1f59e0922fd2289021f1b4736107b428", "sources": ["arxiv", "semantic_scholar"], "title": "ThinKV: Thought-Adaptive KV Cache Compression for Efficient Reasoning Models", "abstract": "The long-output context generation of large reasoning models enables extended chain of thought (CoT) but also drives rapid growth of the key-value (KV) cache, quickly overwhelming GPU memory. To address this challenge, we propose ThinKV, a thought-adaptive KV cache compression framework. ThinKV is based on the observation that attention sparsity reveals distinct thought types with varying importance within the CoT. It applies a hybrid quantization-eviction strategy, assigning token precision by thought importance and progressively evicting tokens from less critical thoughts as reasoning trajectories evolve. Furthermore, to implement ThinKV, we design a kernel that extends PagedAttention to enable efficient reuse of evicted tokens' memory slots, eliminating compaction overheads. Extensive experiments on DeepSeek-R1-Distill, GPT-OSS, and NVIDIA AceReason across mathematics and coding benchmarks show that ThinKV achieves near-lossless accuracy with less than 5% of the original KV cache, while improving performance with up to 5.8x higher inference throughput over state-of-the-art baselines.", "authors": ["Akshat Ramachandran", "Marina Neseem", "Charbel Sakr", "Rangharajan Venkatesan", "Brucek Khailany", "Tushar Krishna"], "categories": ["cs.LG"], "fields_of_study": ["Computer Science"], "published_date": "2025-10-01", "url": "https://arxiv.org/abs/2510.01290", "pdf_url": "https://arxiv.org/pdf/2510.01290v2", "arxiv_id": "2510.01290", "doi": "10.48550/arXiv.2510.01290", "citation_count": 8, "influential_citation_count": 1, "has_code": false, "code_url": null, "venue": "arXiv.org", "quality_score": 0.3197}
{"id": "d2686c39aeb3a412985d635072ac3e58db65de49c9dc391d3e2385e8461848f0", "sources": ["arxiv", "semantic_scholar"], "title": "ReasoningBank: Scaling Agent Self-Evolving with Reasoning Memory", "abstract": "With the growing adoption of large language model agents in persistent real-world roles, they naturally encounter continuous streams of tasks. A key limitation, however, is their failure to learn from the accumulated interaction history, forcing them to discard valuable insights and repeat past errors. We propose ReasoningBank, a novel memory framework that distills generalizable reasoning strategies from an agent's self-judged successful and failed experiences. At test time, an agent retrieves relevant memories from ReasoningBank to inform its interaction and then integrates new learnings back, enabling it to become more capable over time. Building on this powerful experience learner, we further introduce memory-aware test-time scaling (MaTTS), which accelerates and diversifies this learning process by scaling up the agent's interaction experience. By allocating more compute to each task, the agent generates abundant, diverse experiences that provide rich contrastive signals for synthesizing higher-quality memory. The better memory in turn guides more effective scaling, establishing a powerful synergy between memory and test-time scaling. Across web browsing and software engineering benchmarks, ReasoningBank consistently outperforms existing memory mechanisms that store raw trajectories or only successful task routines, improving both effectiveness and efficiency; MaTTS further amplifies these gains. These findings establish memory-driven experience scaling as a new scaling dimension, enabling agents to self-evolve with emergent behaviors naturally arise. Our code can be found at https://github.com/google-research/reasoning-bank.", "authors": ["Siru Ouyang", "Jun Yan", "I-Hung Hsu", "Yanfei Chen", "Ke Jiang", "Zifeng Wang", "Rujun Han", "Long T. Le", "Samira Daruki", "Xiangru Tang", "Vishy Tirumalashetty", "George Lee", "Mahsan Rofouei", "Hangfei Lin", "Jiawei Han", "Chen-Yu Lee", "Tomas Pfister"], "categories": ["cs.AI", "cs.CL"], "fields_of_study": ["Computer Science"], "published_date": "2025-09-29", "url": "https://arxiv.org/abs/2509.25140", "pdf_url": "https://arxiv.org/pdf/2509.25140v2", "arxiv_id": "2509.25140", "doi": "10.48550/arXiv.2509.25140", "citation_count": 126, "influential_citation_count": 26, "has_code": true, "code_url": "https://github.com/google-research/reasoning-bank", "venue": "arXiv.org", "quality_score": 0.7157}
{"id": "bbc52395a40f3dc504129637afd9765ff6577fd4e67c0b5006d88af7a6824a5d", "sources": ["arxiv", "semantic_scholar"], "title": "ARS: Adaptive Reasoning Suppression for Efficient Large Reasoning Language Models", "abstract": "Large Reasoning Language Models (LRLMs or LRMs) demonstrate remarkable capabilities in complex reasoning tasks, but suffer from significant computational inefficiencies due to overthinking phenomena. Existing efficient reasoning methods face the challenge of balancing reasoning quality with inference cost reduction. We propose \\textbf{Adaptive Reasoning Suppression (ARS)}, a novel training-free approach that dynamically suppresses redundant reasoning steps while preserving accuracy through adaptive certainty monitoring. ARS introduces a multi-checkpoint certainty estimation mechanism with progressive suppression thresholds, achieving superior efficiency compared to static suppression methods. Our extensive evaluation across mathematical reasoning benchmarks using multiple model architectures demonstrates that ARS achieves up to 53%, 46.1%, and 57.9% in token, latency and energy reduction, while maintaining or improving accuracy.", "authors": ["Dongqi Zheng"], "categories": ["cs.AI", "cs.CL"], "fields_of_study": ["Computer Science"], "published_date": "2025-09-29", "url": "https://arxiv.org/abs/2510.00071", "pdf_url": "https://arxiv.org/pdf/2510.00071v2", "arxiv_id": "2510.00071", "doi": "10.48550/arXiv.2510.00071", "citation_count": 0, "influential_citation_count": 0, "has_code": false, "code_url": null, "venue": "arXiv.org", "quality_score": 0.3174}
{"id": "afff02ab49b0d8ba366fa5f5109780dfd683d163e2e12102ec145e198523840e", "sources": ["arxiv", "semantic_scholar"], "title": "AdvChain: Adversarial Chain-of-Thought Tuning for Robust Safety Alignment of Large Reasoning Models", "abstract": "Large Reasoning Models (LRMs) have demonstrated remarkable capabilities in complex problem-solving through Chain-of-Thought (CoT) reasoning. However, the multi-step nature of CoT introduces new safety challenges that extend beyond conventional language model alignment. We identify a failure mode in current safety CoT tuning methods: the \\textit{snowball effect}, where minor reasoning deviations progressively amplify throughout the thought process, leading to either harmful compliance or excessive refusal. This effect stems from models being trained to imitate perfect reasoning scripts without learning to self-correct. To address this limitation, we propose AdvChain, an alignment paradigm that teaches models dynamic self-correction through adversarial CoT tuning. Our method involves constructing a dataset containing Temptation-Correction and Hesitation-Correction samples, where models learn to recover from harmful reasoning drifts and unnecessary cautions. Extensive experiments show that AdvChain significantly enhances robustness against jailbreak attacks and CoT hijacking while substantially reducing over-refusal on benign prompts, achieving a superior safety-utility balance without compromising reasoning capabilities. Our work establishes a new direction for building more robust and reliable reasoning models.", "authors": ["Zihao Zhu", "Xinyu Wu", "Gehan Hu", "Siwei Lyu", "Ke Xu", "Baoyuan Wu"], "categories": ["cs.AI", "cs.CL"], "fields_of_study": ["Computer Science"], "published_date": "2025-09-29", "url": "https://arxiv.org/abs/2509.24269", "pdf_url": "https://arxiv.org/pdf/2509.24269v1", "arxiv_id": "2509.24269", "doi": "10.48550/arXiv.2509.24269", "citation_count": 5, "influential_citation_count": 0, "has_code": false, "code_url": null, "venue": "arXiv.org", "quality_score": 0.3174}
{"id": "5782b7c8f7947d8de5d712db30256b7dd116721658f4508c7c687231787f7b20", "sources": ["arxiv", "semantic_scholar"], "title": "Scaling LLM Test-Time Compute with Mobile NPU on Smartphones", "abstract": "Deploying Large Language Models (LLMs) on mobile devices faces the challenge of insufficient performance in smaller models and excessive resource consumption in larger ones. This paper highlights that mobile Neural Processing Units (NPUs) have underutilized computational resources, particularly their matrix multiplication units, during typical LLM inference. To leverage this wasted compute capacity, we propose applying parallel test-time scaling techniques on mobile NPUs to enhance the performance of smaller LLMs. However, this approach confronts inherent NPU challenges, including inadequate hardware support for fine-grained quantization and low efficiency in general-purpose computations. To overcome these, we introduce two key techniques: a hardware-aware tile quantization scheme that aligns group quantization with NPU memory access patterns, and efficient LUT-based replacements for complex operations such as Softmax and dequantization. We design and implement an end-to-end inference system that leverages the NPU's compute capability to support test-time scaling on Qualcomm Snapdragon platforms. Experiments show our approach brings significant speedups: up to 19.0 for mixed-precision GEMM and 2.2 for Softmax. More importantly, we demonstrate that smaller models using test-time scaling can match or exceed the accuracy of larger models, achieving a new performance-cost Pareto frontier.", "authors": ["Zixu Hao", "Jianyu Wei", "Tuowei Wang", "Minxing Huang", "Huiqiang Jiang", "Shiqi Jiang", "Ting Cao", "Ju Ren"], "categories": ["cs.DC", "cs.AI"], "fields_of_study": ["Computer Science"], "published_date": "2025-09-27", "url": "https://arxiv.org/abs/2509.23324", "pdf_url": "https://arxiv.org/pdf/2509.23324v1", "arxiv_id": "2509.23324", "doi": "10.48550/arXiv.2509.23324", "citation_count": 6, "influential_citation_count": 1, "has_code": false, "code_url": null, "venue": "European Conference on Computer Systems", "quality_score": 0.3151}
{"id": "d78efc000243b8495561c47a974f12aa2c52886e4127b18a03d58143f64be7d7", "sources": ["arxiv", "semantic_scholar"], "title": "Retrieval-of-Thought: Efficient Reasoning via Reusing Thoughts", "abstract": "Large reasoning models improve accuracy by producing long reasoning traces, but this inflates latency and cost, motivating inference-time efficiency. We propose Retrieval-of-Thought (RoT), which reuses prior reasoning as composable ``thought\" steps to guide new problems. RoT organizes steps into a thought graph with sequential and semantic edges to enable fast retrieval and flexible recombination. At inference, RoT retrieves query-relevant nodes and applies reward-guided traversal to assemble a problem-specific template that guides generation. This dynamic template reuse reduces redundant exploration and, therefore, reduces output tokens while preserving accuracy. We evaluate RoT on reasoning benchmarks with multiple models, measuring accuracy, token usage, latency, and memory overhead. Findings show small prompt growth but substantial efficiency gains, with RoT reducing output tokens by up to 40%, inference latency by 82%, and cost by 59% while maintaining accuracy. RoT establishes a scalable paradigm for efficient LRM reasoning via dynamic template construction through retrieval.", "authors": ["Ammar Ahmed", "Azal Ahmad Khan", "Ayaan Ahmad", "Sheng Di", "Zirui Liu", "Ali Anwar"], "categories": ["cs.AI", "cs.LG"], "fields_of_study": ["Computer Science"], "published_date": "2025-09-26", "url": "https://arxiv.org/abs/2509.21743", "pdf_url": "https://arxiv.org/pdf/2509.21743v2", "arxiv_id": "2509.21743", "doi": "10.48550/arXiv.2509.21743", "citation_count": 7, "influential_citation_count": 0, "has_code": false, "code_url": null, "venue": "arXiv.org", "quality_score": 0.314}
{"id": "a038be76a893bbf1f10aa17dcc1786218ec34fdee0677d6e4199f6ffab08b7c8", "sources": ["arxiv", "semantic_scholar"], "title": "Dual-Head Reasoning Distillation: Improving Classifier Accuracy with Train-Time-Only Reasoning", "abstract": "Chain-of-Thought (CoT) prompting often improves classification accuracy, but it introduces a significant throughput penalty with rationale generation (Wei et al., 2022; Cheng and Van Durme, 2024). To resolve this trade-off, we introduce Dual-Head Reasoning Distillation (DHRD), a simple training method for decoder-only language models (LMs) that adds (i) a pooled classification head used during training and inference and (ii) a reasoning head supervised by teacher rationales used only in training. We train with a loss function that is a weighted sum of label cross-entropy and token-level LM loss over input-plus-rationale sequences. On seven SuperGLUE tasks, DHRD yields relative gains of 0.65-5.47% over pooled baselines, with notably larger gains on entailment/causal tasks. Since we disable the reasoning head at test time, inference throughput matches pooled classifiers and exceeds CoT decoding on the same backbones by 96-142 times in QPS.", "authors": ["Jillian Xu", "Dylan Zhou", "Vinay Shukla", "Yang Yang", "Junrui Ruan", "Shuhuai Lin", "Wenfei Zou", "Yinxiao Liu", "Karthik Lakshmanan"], "categories": ["cs.CL", "cs.AI"], "fields_of_study": ["Computer Science"], "published_date": "2025-09-25", "url": "https://arxiv.org/abs/2509.21487", "pdf_url": "https://arxiv.org/pdf/2509.21487v2", "arxiv_id": "2509.21487", "doi": "10.48550/arXiv.2509.21487", "citation_count": 3, "influential_citation_count": 0, "has_code": false, "code_url": null, "venue": "arXiv.org", "quality_score": 0.3128}
{"id": "b1a06449b3d57105e9692591ff6e78ef29b89fa83fe3efdbc4bf906a377ed051", "sources": ["arxiv", "semantic_scholar"], "title": "A Formal Comparison Between Chain of Thought and Latent Thought", "abstract": "Chain of thought (CoT) elicits reasoning in large language models by explicitly generating intermediate tokens. In contrast, latent thought reasoning operates directly in the continuous latent space, enabling computation beyond discrete linguistic representations. While both approaches exploit iterative computation, their comparative capabilities remain underexplored. In this work, we present a formal analysis showing that latent thought admits more efficient parallel computation than inherently sequential CoT. In contrast, CoT enables approximate counting and sampling through stochastic decoding. These separations suggest the tasks for which depth-driven recursion is more suitable, thereby offering practical guidance for choosing between reasoning paradigms.", "authors": ["Kevin Xu", "Issei Sato"], "categories": ["cs.AI", "cs.CL", "cs.LG"], "fields_of_study": ["Computer Science"], "published_date": "2025-09-25", "url": "https://arxiv.org/abs/2509.25239", "pdf_url": "https://arxiv.org/pdf/2509.25239v3", "arxiv_id": "2509.25239", "doi": "10.48550/arXiv.2509.25239", "citation_count": 6, "influential_citation_count": 0, "has_code": false, "code_url": null, "venue": "arXiv.org", "quality_score": 0.3128}
{"id": "00c9b68c0160766401e6a7b0239aa011439a33fc66075627a4ca322aae7ffe35", "sources": ["arxiv", "semantic_scholar"], "title": "Expanding Reasoning Potential in Foundation Model by Learning Diverse Chains of Thought Patterns", "abstract": "Recent progress in large reasoning models for challenging mathematical reasoning has been driven by reinforcement learning (RL). Incorporating long chain-of-thought (CoT) data during mid-training has also been shown to substantially improve reasoning depth. However, current approaches often utilize CoT data indiscriminately, leaving open the critical question of which data types most effectively enhance model reasoning capabilities. In this paper, we define the foundation model's reasoning potential for the first time as the inverse of the number of independent attempts required to correctly answer the question, which is strongly correlated with the final model performance. We then propose utilizing diverse data enriched with high-value reasoning patterns to expand the reasoning potential. Specifically, we abstract atomic reasoning patterns from CoT sequences, characterized by commonality and inductive capabilities, and use them to construct a core reference set enriched with valuable reasoning patterns. Furthermore, we propose a dual-granularity algorithm involving chains of reasoning patterns and token entropy, efficiently selecting high-value CoT data (CoTP) from the data pool that aligns with the core set, thereby training models to master reasoning effectively. Only 10B-token CoTP data enables the 85A6B Mixture-of-Experts (MoE) model to improve by 9.58% on the challenging AIME 2024 and 2025, and to raise the upper bound of downstream RL performance by 7.81%.", "authors": ["Xuemiao Zhang", "Can Ren", "Chengying Tu", "Rongxiang Weng", "Shuo Wang", "Hongfei Yan", "Jingang Wang", "Xunliang Cai"], "categories": ["cs.AI", "cs.CL"], "fields_of_study": ["Computer Science"], "published_date": "2025-09-25", "url": "https://arxiv.org/abs/2509.21124", "pdf_url": "https://arxiv.org/pdf/2509.21124v3", "arxiv_id": "2509.21124", "doi": "10.48550/arXiv.2509.21124", "citation_count": 3, "influential_citation_count": 0, "has_code": false, "code_url": null, "venue": "arXiv.org", "quality_score": 0.3128}
{"id": "6c7498d45142e3780a3f1cb20c68ada3676df4d46f7692f2470bba43e2443693", "sources": ["arxiv", "semantic_scholar"], "title": "Mitigating Strategy-Selection Bias in Reasoning for More Effective Test-Time Scaling", "abstract": "Test-time scaling (TTS) has been shown to improve the performance of large language models (LLMs) by sampling and aggregating diverse reasoning paths. However, existing research has overlooked a critical issue: selection bias of reasoning strategies during scaling. Specifically, when generating reasoning processes, LLMs tend to follow certain strategies (e.g., algebraic solutions for math problems) while neglecting other valid alternatives (e.g., geometric solutions), resulting in insufficient exploration of the solution space. To further understand the impact of this bias, we present a theoretical analysis that reveals when it undermines the effectiveness of test-time scaling. Motivated by this theoretical insight, we introduce TTS-Uniform, a framework designed to mitigate the selection bias of reasoning strategies. It (i) identifies potential strategies, (ii) uniformly allocates the sampling budget across them, and (iii) filters out unstable strategies prior to aggregation. Experimental results show that TTS-Uniform significantly enhances scaling effectiveness across multiple mainstream LLMs and benchmark datasets.", "authors": ["Zongqian Wu", "Baoduo Xu", "Tianyu Li", "Zhu Sun", "Xiaofeng Zhu", "Lei Feng"], "categories": ["cs.AI"], "fields_of_study": ["Computer Science"], "published_date": "2025-09-22", "url": "https://arxiv.org/abs/2509.17905", "pdf_url": "https://arxiv.org/pdf/2509.17905v2", "arxiv_id": "2509.17905", "doi": "10.48550/arXiv.2509.17905", "citation_count": 2, "influential_citation_count": 0, "has_code": false, "code_url": null, "venue": "arXiv.org", "quality_score": 0.3094}
{"id": "8aca79dbbace566584c40b711457649dd1d5cb69acc57bf3f62582da44e06c66", "sources": ["arxiv", "semantic_scholar"], "title": "Conversational Orientation Reasoning: Egocentric-to-Allocentric Navigation with Multimodal Chain-of-Thought", "abstract": "Conversational agents must translate egocentric utterances (e.g., \"on my right\") into allocentric orientations (N/E/S/W). This challenge is particularly critical in indoor or complex facilities where GPS signals are weak and detailed maps are unavailable. While chain-of-thought (CoT) prompting has advanced reasoning in language and vision tasks, its application to multimodal spatial orientation remains underexplored. We introduce Conversational Orientation Reasoning (COR), a new benchmark designed for Traditional Chinese conversational navigation projected from real-world environments, addressing egocentric-to-allocentric reasoning in non-English and ASR-transcribed scenarios. We propose a multimodal chain-of-thought (MCoT) framework, which integrates ASR-transcribed speech with landmark coordinates through a structured three-step reasoning process: (1) extracting spatial relations, (2) mapping coordinates to absolute directions, and (3) inferring user orientation. A curriculum learning strategy progressively builds these capabilities on Taiwan-LLM-13B-v2.0-Chat, a mid-sized model representative of resource-constrained settings. Experiments show that MCoT achieves 100% orientation accuracy on clean transcripts and 98.1% with ASR transcripts, substantially outperforming unimodal and non-structured baselines. Moreover, MCoT demonstrates robustness under noisy conversational conditions, including ASR recognition errors and multilingual code-switching. The model also maintains high accuracy in cross-domain evaluation and resilience to linguistic variation, domain shift, and referential ambiguity. These findings highlight the potential of structured MCoT spatial reasoning as a path toward interpretable and resource-efficient embodied navigation.", "authors": ["Yu Ti Huang"], "categories": ["cs.LG", "cs.AI", "cs.CL", "cs.RO"], "fields_of_study": ["Computer Science"], "published_date": "2025-09-20", "url": "https://arxiv.org/abs/2509.18200", "pdf_url": "https://arxiv.org/pdf/2509.18200v1", "arxiv_id": "2509.18200", "doi": "10.48550/arXiv.2509.18200", "citation_count": 0, "influential_citation_count": 0, "has_code": false, "code_url": null, "venue": "arXiv.org", "quality_score": 0.3071}
{"id": "64a51bd6d2d8089307a8253f94dcc6b536f2f75ffd68fb17484a4c1567b1f9e9", "sources": ["arxiv", "semantic_scholar"], "title": "Slim-SC: Thought Pruning for Efficient Scaling with Self-Consistency", "abstract": "Recently, Test-Time Scaling (TTS) has gained increasing attention for improving LLM reasoning performance at test time without retraining the model. A notable TTS technique is Self-Consistency (SC), which generates multiple reasoning chains in parallel and selects the final answer via majority voting. While effective, the order-of-magnitude computational overhead limits its broad deployment. Prior attempts to accelerate SC mainly rely on model-based confidence scores or heuristics with limited empirical support. For the first time, we theoretically and empirically analyze the inefficiencies of SC and reveal actionable opportunities for improvement. Building on these insights, we propose Slim-SC, a step-wise pruning strategy that identifies and removes redundant chains using inter-chain similarity at the thought level. Experiments on three STEM reasoning datasets and two recent LLM architectures show that Slim-SC reduces inference latency and KVC usage by up to 45% and 26%, respectively, with R1-Distill, while maintaining or improving accuracy, thus offering a simple yet efficient TTS alternative for SC.", "authors": ["Colin Hong", "Xu Guo", "Anand Chaanan Singh", "Esha Choukse", "Dmitrii Ustiugov"], "categories": ["cs.CL", "cs.AI", "cs.LG"], "fields_of_study": ["Computer Science"], "published_date": "2025-09-17", "url": "https://arxiv.org/abs/2509.13990", "pdf_url": "https://arxiv.org/pdf/2509.13990v1", "arxiv_id": "2509.13990", "doi": "10.48550/arXiv.2509.13990", "citation_count": 12, "influential_citation_count": 5, "has_code": false, "code_url": null, "venue": "Conference on Empirical Methods in Natural Language Processing", "quality_score": 0.3891}
{"id": "99a0015b0a7d4c9436d99bbeeac59213b8b563ea2e6da3536d80cd6c9a01328d", "sources": ["arxiv", "semantic_scholar"], "title": "LTA-thinker: Latent Thought-Augmented Training Framework for Large Language Models on Complex Reasoning", "abstract": "Complex Reasoning in Large Language Models can be dynamically optimized using Test-Time Scaling (TTS) to mitigate Overthinking. Methods such as Coconut, SoftCoT and its variant are effective in continuous latent space inference, the core bottleneck still lies in the efficient generation and utilization of high-quality Latent Thought. Drawing from the theory of SoftCoT++ that a larger variance in the generated Latent Thought distribution more closely approximates the golden truth distribution, we propose a Latent Thought-Augmented Training Framework--LTA-Thinker, which improves distributional variance and enhances reasoning performance from two perspectives. First, LTA-Thinker constructs a Latent Thought generation architecture based on a learnable prior. This architecture aims to increase the variance distribution of generated Latent Thought Vectors in order to simplify the overall structure and raise the performance ceiling. Second, LTA-Thinker introduces a distribution-based directional optimization paradigm that jointly constrains both distribution locality and distribution scale. This mechanism improves information efficiency and computational cost through a multi-objective co-training strategy, which combines standard Supervised Fine-Tuning (SFT) loss with two novel losses: Semantic Alignment Loss, which utilizes KL divergence to ensure that the Latent Thought is highly relevant to the semantics of the question; Reasoning Focus Loss, which utilizes a contrastive learning mechanism to guide the model to focus on the most critical reasoning steps. Experiments show that LTA-thinker achieves state-of-the-art (SOTA) performance among various baselines and demonstrates a higher performance ceiling and better scaling effects.", "authors": ["Jiaqi Wang", "Binquan Ji", "Haibo Luo", "Yiyang Qi", "Ruiting Li", "Huiyan Wang", "Yuantao Han", "Cangyi Yang", "jiaxu Zhang", "Feiliang Ren"], "categories": ["cs.AI"], "fields_of_study": ["Computer Science"], "published_date": "2025-09-16", "url": "https://arxiv.org/abs/2509.12875", "pdf_url": "https://arxiv.org/pdf/2509.12875v3", "arxiv_id": "2509.12875", "doi": "10.48550/arXiv.2509.12875", "citation_count": 4, "influential_citation_count": 1, "has_code": false, "code_url": null, "venue": "arXiv.org", "quality_score": 0.3025}
{"id": "ca4eb42dea342143ca5eef4ebb2047a039102531dc9348dae72785a8be4d20e5", "sources": ["arxiv", "semantic_scholar"], "title": "Reasoning Models Can be Accurately Pruned Via Chain-of-Thought Reconstruction", "abstract": "Reasoning language models such as DeepSeek-R1 produce long chain-of-thought traces during inference time which make them costly to deploy at scale. We show that using compression techniques such as neural network pruning produces greater performance loss than in typical language modeling tasks, and in some cases can make the model slower since they cause the model to produce more thinking tokens but with worse performance. We show that this is partly due to the fact that standard LLM pruning methods often focus on input reconstruction, whereas reasoning is a decode-dominated task. We introduce a simple, drop-in fix: during pruning we jointly reconstruct activations from the input and the model's on-policy chain-of-thought traces. This \"Reasoning-Aware Compression\" (RAC) integrates seamlessly into existing pruning workflows such as SparseGPT, and boosts their performance significantly. Code reproducing the results in the paper can be found at: https://github.com/RyanLucas3/RAC", "authors": ["Ryan Lucas", "Kayhan Behdin", "Zhipeng Wang", "Qingquan Song", "Shao Tang", "Rahul Mazumder"], "categories": ["cs.AI"], "fields_of_study": ["Computer Science"], "published_date": "2025-09-15", "url": "https://arxiv.org/abs/2509.12464", "pdf_url": "https://arxiv.org/pdf/2509.12464v2", "arxiv_id": "2509.12464", "doi": "10.48550/arXiv.2509.12464", "citation_count": 1, "influential_citation_count": 0, "has_code": true, "code_url": "https://github.com/RyanLucas3/RAC", "venue": "arXiv.org", "quality_score": 0.4657}
{"id": "2af4a7e334a41eb6e5b3c79a333e3d337ba24e39da84413d54f3228899acfd64", "sources": ["arxiv", "semantic_scholar"], "title": "K2-Think: A Parameter-Efficient Reasoning System", "abstract": "K2-Think is a reasoning system that achieves state-of-the-art performance with a 32B parameter model, matching or surpassing much larger models like GPT-OSS 120B and DeepSeek v3.1. Built on the Qwen2.5 base model, our system shows that smaller models can compete at the highest levels by combining advanced post-training and test-time computation techniques. The approach is based on six key technical pillars: Long Chain-of-thought Supervised Finetuning, Reinforcement Learning with Verifiable Rewards (RLVR), Agentic planning prior to reasoning, Test-time Scaling, Speculative Decoding, and Inference-optimized Hardware, all using publicly available open-source datasets. K2-Think excels in mathematical reasoning, achieving state-of-the-art scores on public benchmarks for open-source models, while also performing strongly in other areas such as Code and Science. Our results confirm that a more parameter-efficient model like K2-Think 32B can compete with state-of-the-art systems through an integrated post-training recipe that includes long chain-of-thought training and strategic inference-time enhancements, making open-source reasoning systems more accessible and affordable. K2-Think is freely available at k2think.ai, offering best-in-class inference speeds of over 2,000 tokens per second per request via the Cerebras Wafer-Scale Engine.", "authors": ["Zhoujun Cheng", "Richard Fan", "Shibo Hao", "Taylor W. Killian", "Haonan Li", "Suqi Sun", "Hector Ren", "Alexander Moreno", "Daqian Zhang", "Tianjun Zhong", "Yuxin Xiong", "Yuanzhe Hu", "Yutao Xie", "Xudong Han", "Yuqi Wang", "Varad Pimpalkhute", "Yonghao Zhuang", "Aaryamonvikram Singh", "Xuezhi Liang", "Anze Xie", "Jianshu She", "Desai Fan", "Chengqian Gao", "Liqun Ma", "Mikhail Yurochkin", "John Maggs", "Xuezhe Ma", "Guowei He", "Zhiting Hu", "Zhengzhong Liu", "Eric P. Xing"], "categories": ["cs.LG"], "fields_of_study": ["Computer Science"], "published_date": "2025-09-09", "url": "https://arxiv.org/abs/2509.07604", "pdf_url": "https://arxiv.org/pdf/2509.07604v3", "arxiv_id": "2509.07604", "doi": "10.48550/arXiv.2509.07604", "citation_count": 8, "influential_citation_count": 0, "has_code": true, "code_url": null, "venue": "arXiv.org", "quality_score": 0.4551}
{"id": "0492f7dbb7295ba2b9666a96d705d72ac11bdb9dba3e42650f253687a2baccb1", "sources": ["arxiv", "semantic_scholar"], "title": "Test-Time Scaling in Reasoning Models Is Not Effective for Knowledge-Intensive Tasks Yet", "abstract": "Test-time scaling increases inference-time computation by allowing models to generate long reasoning chains, and has improved performance across many domains. However, in this work, we show that this approach is not yet effective for knowledge-intensive tasks. We evaluate 14 reasoning models on two knowledge-intensive benchmarks and find that increasing test-time computation does not consistently improve accuracy and often increases hallucinations. Further analysis shows that changes in hallucination rates under increased test-time computation are largely driven by models' willingness to answer. We also observe that extended reasoning can induce confirmation bias, leading to overconfident hallucinations. Finally, we provide an information-theoretic account: compute-only test-time scaling is a post-processing of a fixed trained model and therefore cannot increase information about the ground-truth answer beyond what is already encoded in the model, explaining its limited gains on knowledge-intensive tasks. Code and data are available at https://github.com/XuZhao0/tts-knowledge", "authors": ["James Xu Zhao", "Bryan Hooi", "See-Kiong Ng"], "categories": ["cs.AI", "cs.CL", "cs.LG"], "fields_of_study": ["Computer Science"], "published_date": "2025-09-08", "url": "https://arxiv.org/abs/2509.06861", "pdf_url": "https://arxiv.org/pdf/2509.06861v2", "arxiv_id": "2509.06861", "doi": "10.48550/arXiv.2509.06861", "citation_count": 10, "influential_citation_count": 0, "has_code": true, "code_url": "https://github.com/XuZhao0/tts-knowledge", "venue": "arXiv.org", "quality_score": 0.4533}
{"id": "30a81d504c8af4c2229e4982d7e971d2c8ca9c567cecdf2ecb292af21d961fa4", "sources": ["arxiv", "semantic_scholar"], "title": "Less is More Tokens: Efficient Math Reasoning via Difficulty-Aware Chain-of-Thought Distillation", "abstract": "Chain-of-thought reasoning, while powerful, can produce unnecessarily verbose output for simpler problems. We present a framework for difficulty-aware reasoning that teaches models to dynamically adjust reasoning depth based on problem complexity. Remarkably, we show that models can be endowed with such dynamic inference pathways without any architectural modifications; we simply post-train on data that is carefully curated to include chain-of-thought traces that are proportional in length to problem difficulty. Our analysis reveals that post-training via supervised fine-tuning (SFT) primarily captures patterns like reasoning length and format, while direct preference optimization (DPO) preserves reasoning accuracy, with their combination reducing length and maintaining or improving performance. Both quantitative metrics and qualitative assessments confirm that models can learn to \"think proportionally\", reasoning minimally on simple problems while maintaining depth for complex ones.", "authors": ["Abdul Waheed", "Chancharik Mitra", "Laurie Z. Wang", "Deva Ramanan", "Bhiksha Raj"], "categories": ["cs.CL"], "fields_of_study": ["Computer Science"], "published_date": "2025-09-05", "url": "https://arxiv.org/abs/2509.05226", "pdf_url": "https://arxiv.org/pdf/2509.05226v1", "arxiv_id": "2509.05226", "doi": "10.48550/arXiv.2509.05226", "citation_count": 0, "influential_citation_count": 0, "has_code": false, "code_url": null, "venue": "arXiv.org", "quality_score": 0.2899}
{"id": "29d52e9bc0c8b1dea677661900835320f195e3c6de789226a6c0cd6c948b183f", "sources": ["arxiv", "semantic_scholar"], "title": "Chain or tree? Re-evaluating complex reasoning from the perspective of a matrix of thought", "abstract": "Large Language Models (LLMs) face significant accuracy degradation due to insufficient reasoning ability when dealing with complex and abstract tasks. Thought structures such as Chain of Thought (CoT) and Tree of Thought (ToT) focus on enhancing the reasoning capability of LLMs. However, they suffer from inherent drawbacks such as redundancy within the same layer of the tree structure and the singularity of the paths in the chain structure. Some studies have utilized Retrieval-Augmented Generation (RAG) methods to enhance CoT and ToT in mitigating hallucinations in LLMs, yet the fundamental shortcomings of the thought structures still persist. Furthermore, when dealing with multi-entity and multi-hop information, the retrieved verification knowledge often contains large amounts of fragmented, superficial, or even erroneous data, misleading the reasoning process of LLMs. To address these issues, we propose the Matrix of Thought (MoT), a novel and efficient thought structure for LLMs. MoT explores problems in both horizontal and vertical dimensions through a \"column-cell communication\" mechanism, enabling LLMs to actively engage in multi-strategy and deep thinking while reducing redundancy in the thought nodes within the column cells, thereby enhancing the reasoning capability of LLMs. Additionally, through a fact-correction mechanism, it leverages the knowledge graph triples retrieved by RAG and the original text to construct knowledge units and correct erroneous answers. To validate the effectiveness of this method, we conducted extensive experiments in three tasks: 24-point game, question answering evaluation, and proposition writing.The results demonstrate that our framework outperforms state-of-the-art methods, with reasoning time only 14.4\\% of that of the baseline method, proving its efficiency and accuracy. The code for framework is available at https://github.com/lyfiter/mtqa.", "authors": ["Fengxiao Tang", "Yufeng Li", "Zongzong Wu", "Ming Zhao"], "categories": ["cs.CL", "cs.AI"], "fields_of_study": ["Computer Science"], "published_date": "2025-09-04", "url": "https://arxiv.org/abs/2509.03918", "pdf_url": "https://arxiv.org/pdf/2509.03918v2", "arxiv_id": "2509.03918", "doi": "10.48550/arXiv.2509.03918", "citation_count": 1, "influential_citation_count": 0, "has_code": true, "code_url": "https://github.com/lyfiter/mtqa", "venue": "arXiv.org", "quality_score": 0.4462}
{"id": "4c97fb75a345424b91c894df5551ab0f00d354a9dc99f1988028ba5873d78081", "sources": ["arxiv", "semantic_scholar"], "title": "Why Does Reasoning Length Converge? Unveiling the Underfitting-Overfitting Trade-off in Chain-of-Thought", "abstract": "Test-time scaling, primarily manifested through multi-step Chain-of-Thought (CoT) reasoning via Reinforcement Learning (RL), has emerged as a pivotal paradigm for enhancing the reasoning capabilities of Large Language Models (LLMs). However, a significant theoretical gap persists: traditional token-level analysis fails to capture the macroscopic dynamics of reasoning-level scaling. To address this, we introduce CoT-Space, a novel theoretical framework that recasts the reasoning process from a discrete token-prediction task to an optimization process within a continuous, reasoning-level semantic space. By modeling the reasoning trajectory from both noise and risk perspectives and revitalizing foundational principles from classical learning theory, we demonstrate that the observed convergence to an optimal CoT length is a natural consequence of the fundamental trade-off between underfitting and overfitting. We further utilize RL as a tool to elicit and verify these results in our experiments. Our findings provide a mechanistic explanation for the internal test-time scaling via RL, offering a principled theoretical foundation to optimize reasoning trajectories in modern LLMs.", "authors": ["Zeyu Gan", "Hao Yi", "Yong Liu"], "categories": ["cs.AI", "cs.CL"], "fields_of_study": ["Computer Science"], "published_date": "2025-09-04", "url": "https://arxiv.org/abs/2509.04027", "pdf_url": "https://arxiv.org/pdf/2509.04027v4", "arxiv_id": "2509.04027", "doi": null, "citation_count": 2, "influential_citation_count": 1, "has_code": false, "code_url": null, "venue": null, "quality_score": 0.1837}
{"id": "e013a8d45247186efb8662c143dabd60f1efc453be131dc361e864180b3af797", "sources": ["arxiv", "semantic_scholar"], "title": "ParaThinker: Native Parallel Thinking as a New Paradigm to Scale LLM Test-time Compute", "abstract": "Recent advances in Large Language Models (LLMs) have been driven by test-time compute scaling - a strategy that improves reasoning by generating longer, sequential thought processes. While effective, this approach encounters a significant bottleneck as computation increases, where further computation offers only marginal performance gains. We argue this ceiling is not an inherent limit of the model's capability but a flaw in the scaling strategy itself, a phenomenon we term \"Tunnel Vision\", where a model's imperfect initial steps lock it into a suboptimal reasoning path. To overcome this, we introduce a new scaling paradigm: native thought parallelism. We present ParaThinker, an end-to-end framework that trains an LLM to generate multiple, diverse reasoning paths in parallel and synthesize them into a superior final answer. By exploring different lines of thoughts simultaneously, ParaThinker effectively sidesteps the Tunnel Vision issue and unlocks the model's latent reasoning potential. Our approach demonstrates that scaling compute in parallel (width) is a more effective and efficient way to superior reasoning than simply scaling sequentially (depth). On challenging reasoning benchmarks, ParaThinker achieves substantial accuracy improvements over sequential LLMs (12.3% for 1.5B and 7.5% for 7B models on average with 8 parallel paths), while adding only negligible latency overhead (7.1%). This enables smaller models to surpass much larger counterparts and establishes parallel thinking as a critical, efficient dimension for scaling future LLMs.", "authors": ["Hao Wen", "Yifan Su", "Feifei Zhang", "Yunxin Liu", "Yunhao Liu", "Ya-Qin Zhang", "Yuanchun Li"], "categories": ["cs.CL", "cs.AI"], "fields_of_study": ["Computer Science"], "published_date": "2025-08-30", "url": "https://arxiv.org/abs/2509.04475", "pdf_url": "https://arxiv.org/pdf/2509.04475v1", "arxiv_id": "2509.04475", "doi": "10.48550/arXiv.2509.04475", "citation_count": 34, "influential_citation_count": 1, "has_code": false, "code_url": null, "venue": "arXiv.org", "quality_score": 0.386}
{"id": "b686dc1fffbef85733df570f6eb505850111cfdf23d02217f88e789114384153", "sources": ["arxiv", "semantic_scholar"], "title": "SQL-of-Thought: Multi-agentic Text-to-SQL with Guided Error Correction", "abstract": "Converting natural language queries into SQL queries is a crucial challenge in both industry and academia, aiming to increase access to databases and large-scale applications. This work examines how in-context learning and chain-of-thought can be utilized to develop a robust solution for text-to-SQL systems. We propose SQL-of-Thought: a multi-agent framework that decomposes the Text2SQL task into schema linking, subproblem identification, query plan generation, SQL generation, and a guided correction loop. Unlike prior systems that rely only on execution-based static correction, we introduce taxonomy-guided dynamic error modification informed by in-context learning. SQL-of-Thought achieves state-of-the-art results on the Spider dataset and its variants, combining guided error taxonomy with reasoning-based query planning.", "authors": ["Saumya Chaturvedi", "Aman Chadha", "Laurent Bindschaedler"], "categories": ["cs.DB", "cs.LG"], "fields_of_study": ["Computer Science"], "published_date": "2025-08-30", "url": "https://arxiv.org/abs/2509.00581", "pdf_url": "https://arxiv.org/pdf/2509.00581v2", "arxiv_id": "2509.00581", "doi": "10.48550/arXiv.2509.00581", "citation_count": 8, "influential_citation_count": 0, "has_code": true, "code_url": "https://github.com/shollercoaster/SQL-of-Thought", "venue": "arXiv.org", "quality_score": 0.4374}
{"id": "18e6c7cbbdc83e0f2541cda34743167c9e711306604f9d9dbb80641e2a9f5fbb", "sources": ["arxiv", "semantic_scholar"], "title": "Explainable Chain-of-Thought Reasoning: An Empirical Analysis on State-Aware Reasoning Dynamics", "abstract": "Recent advances in chain-of-thought (CoT) prompting have enabled large language models (LLMs) to perform multi-step reasoning. However, the explainability of such reasoning remains limited, with prior work primarily focusing on local token-level attribution, such that the high-level semantic roles of reasoning steps and their transitions remain underexplored. In this paper, we introduce a state-aware transition framework that abstracts CoT trajectories into structured latent dynamics. Specifically, to capture the evolving semantics of CoT reasoning, each reasoning step is represented via spectral analysis of token-level embeddings and clustered into semantically coherent latent states. To characterize the global structure of reasoning, we model their progression as a Markov chain, yielding a structured and interpretable view of the reasoning process. This abstraction supports a range of analyses, including semantic role identification, temporal pattern visualization, and consistency evaluation.", "authors": ["Sheldon Yu", "Yuxin Xiong", "Junda Wu", "Xintong Li", "Tong Yu", "Xiang Chen", "Ritwik Sinha", "Jingbo Shang", "Julian McAuley"], "categories": ["cs.CL", "cs.AI"], "fields_of_study": ["Computer Science"], "published_date": "2025-08-29", "url": "https://arxiv.org/abs/2509.00190", "pdf_url": "https://arxiv.org/pdf/2509.00190v2", "arxiv_id": "2509.00190", "doi": "10.48550/arXiv.2509.00190", "citation_count": 15, "influential_citation_count": 1, "has_code": false, "code_url": null, "venue": "Conference on Empirical Methods in Natural Language Processing", "quality_score": 0.301}
{"id": "3e1d8ffb4fdac49a01c0ec39a7d0d4d27cdf8974f894cd3489f8abed95610559", "sources": ["arxiv", "semantic_scholar"], "title": "FastTTS: Accelerating Test-Time Scaling for Edge LLM Reasoning", "abstract": "Recent advances in reasoning Large Language Models (LLMs) are driving the emergence of agentic AI systems. Edge deployment of LLM agents near end users is increasingly necessary to protect data privacy, enable offline use, and provide responsive interaction with local context. However, strict memory constraints on edge devices limit deployment to smaller LLMs, whose reasoning capabilities are much weaker than those of large cloud models, hindering practical deployment of edge agentic AI. Test-Time Scaling (TTS) offers a promising solution by allocating more compute during inference to enhance the reasoning capability of edge LLMs. However, current TTS methods introduce heavy hardware performance overhead on resource-constrained devices, making them impractical for real applications. To address this challenge, we present FastTTS, a serving system that enables fast and efficient TTS for memory-constrained LLM reasoning. After analyzing common patterns across various TTS methods and identifying their performance bottlenecks, we introduce three novel techniques: i) Speculative Beam Extension, which mitigates system stragglers caused by irregular reasoning paths, ii) Asymmetric Multi-Model Memory Allocation, which dynamically balances memory usage between token generation and reasoning-step verification, and iii) Dynamic Prefix-Aware Scheduling, which optimizes reasoning execution to maximize KV-cache reuse across search paths. FastTTS offers a plug-and-play third-party library on top of vLLM, enabling edge LLMs on a single consumer GPU to match cloud-model accuracy and cloud-measured latency. Comprehensive evaluation shows that FastTTS achieves an average 2.2x higher goodput and reduces latency by 38%--68% compared to the vLLM baseline; it pushes the boundaries of low-latency TTS on memory-constrained edge devices and highlights the potential for democratizing agentic AI.", "authors": ["Hao Mark Chen", "Zhiwen Mo", "Guanxi Lu", "Shuang Liang", "Lingxiao Ma", "Wayne Luk", "Hongxiang Fan"], "categories": ["cs.LG"], "fields_of_study": ["Computer Science"], "published_date": "2025-08-29", "url": "https://arxiv.org/abs/2509.00195", "pdf_url": "https://arxiv.org/pdf/2509.00195v2", "arxiv_id": "2509.00195", "doi": "10.1145/3779212.3790161", "citation_count": 4, "influential_citation_count": 0, "has_code": false, "code_url": null, "venue": "International Conference on Architectural Support for Programming Languages and Operating Systems", "quality_score": 0.2819}
{"id": "7e171e8a3cda6079b8a13779d715a8f1298ff68ba8ed9d293b3ec8deb371a605", "sources": ["arxiv", "semantic_scholar"], "title": "Logical Reasoning with Outcome Reward Models for Test-Time Scaling", "abstract": "Logical reasoning is a critical benchmark for evaluating the capabilities of large language models (LLMs), as it reflects their ability to derive valid conclusions from given premises. While the combination of test-time scaling with dedicated outcome or process reward models has opened up new avenues to enhance LLMs performance in complex reasoning tasks, this space is under-explored in deductive logical reasoning. We present a set of Outcome Reward Models (ORMs) for deductive reasoning. To train the ORMs we mainly generate data using Chain-of-Thought (CoT) with single and multiple samples. Additionally, we propose a novel tactic to further expand the type of errors covered in the training dataset of the ORM. In particular, we propose an echo generation technique that leverages LLMs' tendency to reflect incorrect assumptions made in prompts to extract additional training data, covering previously unexplored error types. While a standard CoT chain may contain errors likely to be made by the reasoner, the echo strategy deliberately steers the model toward incorrect reasoning. We show that ORMs trained on CoT and echo-augmented data demonstrate improved performance on the FOLIO, JustLogic, and ProverQA datasets across four different LLMs.", "authors": ["Ramya Keerthy Thatikonda", "Wray Buntine", "Ehsan Shareghi"], "categories": ["cs.CL", "cs.AI"], "fields_of_study": ["Computer Science"], "published_date": "2025-08-27", "url": "https://arxiv.org/abs/2508.19903", "pdf_url": "https://arxiv.org/pdf/2508.19903v1", "arxiv_id": "2508.19903", "doi": "10.48550/arXiv.2508.19903", "citation_count": 3, "influential_citation_count": 0, "has_code": false, "code_url": null, "venue": "Conference on Empirical Methods in Natural Language Processing", "quality_score": 0.2796}
{"id": "6cc1babd17cd8d1c8b1575d9dd60d47875a79fb6797e20468de44c1a46a506de", "sources": ["arxiv", "semantic_scholar"], "title": "MIRAGE: Scaling Test-Time Inference with Parallel Graph-Retrieval-Augmented Reasoning Chains", "abstract": "Large reasoning models (LRMs) have shown significant progress in test-time scaling through chain-of-thought prompting. Current approaches like search-o1 integrate retrieval augmented generation (RAG) into multi-step reasoning processes but rely on a single, linear reasoning chain while incorporating unstructured textual information in a flat, context-agnostic manner. As a result, these approaches can lead to error accumulation throughout the reasoning chain, which significantly limits its effectiveness in medical question-answering (QA) tasks where both accuracy and traceability are critical requirements. To address these challenges, we propose MIRAGE (Multi-chain Inference with Retrieval-Augmented Graph Exploration), a novel test-time scalable reasoning framework that performs dynamic multi-chain inference over structured medical knowledge graphs. Specifically, MIRAGE 1) decomposes complex queries into entity-grounded sub-questions, 2) executes parallel inference chains, 3) retrieves evidence adaptively via neighbor expansion and multi-hop traversal, and 4) integrates answers using cross-chain verification to resolve contradictions. Experiments on three medical QA benchmarks (GenMedGPT-5k, CMCQA, and ExplainCPE) show that MIRAGE consistently outperforms GPT-4o, Tree-of-Thought variants, and other retrieval-augmented baselines in both automatic and human evaluations. Additionally, MIRAGE improves interpretability by generating explicit reasoning chains that trace each factual claim to concrete chains within the knowledge graph, making it well-suited for complex medical reasoning scenarios. The code will be available for further research.", "authors": ["Kaiwen Wei", "Rui Shan", "Dongsheng Zou", "Jianzhong Yang", "Bi Zhao", "Junnan Zhu", "Jiang Zhong"], "categories": ["cs.CL"], "fields_of_study": ["Computer Science"], "published_date": "2025-08-25", "url": "https://arxiv.org/abs/2508.18260", "pdf_url": "https://arxiv.org/pdf/2508.18260v2", "arxiv_id": "2508.18260", "doi": "10.48550/arXiv.2508.18260", "citation_count": 5, "influential_citation_count": 0, "has_code": false, "code_url": null, "venue": "AAAI Conference on Artificial Intelligence", "quality_score": 0.2773}
{"id": "12e113a469c6bb3f602dbd684aa3f930c8e142c5bc54d3e72950a40f7e67e892", "sources": ["arxiv", "semantic_scholar"], "title": "Explain Before You Answer: A Survey on Compositional Visual Reasoning", "abstract": "Compositional visual reasoning has emerged as a key research frontier in multimodal AI, aiming to endow machines with the human-like ability to decompose visual scenes, ground intermediate concepts, and perform multi-step logical inference. While early surveys focus on monolithic vision-language models or general multimodal reasoning, a dedicated synthesis of the rapidly expanding compositional visual reasoning literature is still missing. We fill this gap with a comprehensive survey spanning 2023 to 2025 that systematically reviews 260+ papers from top venues (CVPR, ICCV, NeurIPS, ICML, ACL, etc.). We first formalize core definitions and describe why compositional approaches offer advantages in cognitive alignment, semantic fidelity, robustness, interpretability, and data efficiency. Next, we trace a five-stage paradigm shift: from prompt-enhanced language-centric pipelines, through tool-enhanced LLMs and tool-enhanced VLMs, to recently minted chain-of-thought reasoning and unified agentic VLMs, highlighting their architectural designs, strengths, and limitations. We then catalog 60+ benchmarks and corresponding metrics that probe compositional visual reasoning along dimensions such as grounding accuracy, chain-of-thought faithfulness, and high-resolution perception. Drawing on these analyses, we distill key insights, identify open challenges (e.g., limitations of LLM-based reasoning, hallucination, a bias toward deductive reasoning, scalable supervision, tool integration, and benchmark limitations), and outline future directions, including world-model integration, human-AI collaborative reasoning, and richer evaluation protocols. By offering a unified taxonomy, historical roadmap, and critical outlook, this survey aims to serve as a foundational reference and inspire the next generation of compositional visual reasoning research.", "authors": ["Fucai Ke", "Joy Hsu", "Zhixi Cai", "Zixian Ma", "Xin Zheng", "Xindi Wu", "Sukai Huang", "Weiqing Wang", "Pari Delir Haghighi", "Gholamreza Haffari", "Ranjay Krishna", "Jiajun Wu", "Hamid Rezatofighi"], "categories": ["cs.CV", "cs.AI"], "fields_of_study": ["Computer Science"], "published_date": "2025-08-24", "url": "https://arxiv.org/abs/2508.17298", "pdf_url": "https://arxiv.org/pdf/2508.17298v2", "arxiv_id": "2508.17298", "doi": "10.48550/arXiv.2508.17298", "citation_count": 19, "influential_citation_count": 0, "has_code": true, "code_url": "https://github.com/pokerme7777/Compositional-Visual-Reasoning-Survey", "venue": "arXiv.org", "quality_score": 0.4268}
{"id": "fe3cc1eb181218206473ca643fde67da0d09d48a4cb78e926417fe48631d76c8", "sources": ["arxiv", "semantic_scholar"], "title": "Beyond Memorization: Extending Reasoning Depth with Recurrence, Memory and Test-Time Compute Scaling", "abstract": "Reasoning is a core capability of large language models, yet how multi-step reasoning is learned and executed remains unclear. We study this question in a controlled cellular-automata (1dCA) framework that excludes memorisation by using disjoint training and test rules. Given a short state sequence, the model is required to infer the hidden local rule and then chain it to predict multiple future steps. Our evaluation shows that LLMs largely fail to reliably solve a natural-language proxy of the proposed task. We find that most neural architectures trained from scratch can learn rule inference and achieve high next-step accuracy, but performance drops sharply as the required number of intermediate reasoning steps increases. Experiments show that increasing model depth is crucial, and extending effective depth via recurrence, memory, or test-time compute improves results but remains bounded. The code is available on github: https://github.com/RodkinIvan/associative-recurrent-memory-transformer/tree/ACT", "authors": ["Ivan Rodkin", "Daniil Orel", "Konstantin Smirnov", "Arman Bolatov", "Bilal Elbouardi", "Besher Hassan", "Yuri Kuratov", "Aydar Bulatov", "Preslav Nakov", "Timothy Baldwin", "Artem Shelmanov", "Mikhail Burtsev"], "categories": ["cs.LG", "cs.AI"], "fields_of_study": ["Computer Science"], "published_date": "2025-08-22", "url": "https://arxiv.org/abs/2508.16745", "pdf_url": "https://arxiv.org/pdf/2508.16745v3", "arxiv_id": "2508.16745", "doi": "10.48550/arXiv.2508.16745", "citation_count": 3, "influential_citation_count": 0, "has_code": true, "code_url": "https://github.com/RodkinIvan/associative-recurrent-memory-transformer/tree/ACT", "venue": "arXiv.org", "quality_score": 0.4232}
{"id": "e028184a233cb5504e3e44b17892e7107fdafad3ad8c31c3b8f3a47eb94f0b8a", "sources": ["arxiv", "semantic_scholar"], "title": "Scaling Group Inference for Diverse and High-Quality Generation", "abstract": "Generative models typically sample outputs independently, and recent inference-time guidance and scaling algorithms focus on improving the quality of individual samples. However, in real-world applications, users are often presented with a set of multiple images (e.g., 4-8) for each prompt, where independent sampling tends to lead to redundant results, limiting user choices and hindering idea exploration. In this work, we introduce a scalable group inference method that improves both the diversity and quality of a group of samples. We formulate group inference as a quadratic integer assignment problem: candidate outputs are modeled as graph nodes, and a subset is selected to optimize sample quality (unary term) while maximizing group diversity (binary term). To substantially improve runtime efficiency, we progressively prune the candidate set using intermediate predictions, allowing our method to scale up to large candidate sets. Extensive experiments show that our method significantly improves group diversity and quality compared to independent sampling baselines and recent inference algorithms. Our framework generalizes across a wide range of tasks, including text-to-image, image-to-image, image prompting, and video generation, enabling generative models to treat multiple outputs as cohesive groups rather than independent samples.", "authors": ["Gaurav Parmar", "Or Patashnik", "Daniil Ostashev", "Kuan-Chieh Wang", "Kfir Aberman", "Srinivasa Narasimhan", "Jun-Yan Zhu"], "categories": ["cs.CV", "cs.GR", "cs.LG"], "fields_of_study": ["Computer Science"], "published_date": "2025-08-21", "url": "https://arxiv.org/abs/2508.15773", "pdf_url": "https://arxiv.org/pdf/2508.15773v1", "arxiv_id": "2508.15773", "doi": "10.48550/arXiv.2508.15773", "citation_count": 9, "influential_citation_count": 2, "has_code": true, "code_url": "https://github.com/GaParmar/group-inference", "venue": "arXiv.org", "quality_score": 0.4215}
{"id": "1bda3db8a0bf2821fba6cffc8f42c117f43545e5b4c0f1822a35f3839dd11fb7", "sources": ["arxiv", "semantic_scholar"], "title": "Long Chain-of-Thought Reasoning Across Languages", "abstract": "While large reasoning models have shown remarkable ability to generate long chains-of-thought (CoTs) in English, we still lack understanding of how these long-form reasoning abilities transfer to the vast majority of the world's languages. In this work, we systematically investigate four key stages of model development--scaling, pretraining, post-training, and inference--to understand how long CoT capabilities extend beyond English. We compare two reasoning settings across nine non-English target languages: En-CoT, where models process target-language inputs, but reason in English; and Target-CoT, where models both process inputs and generate long CoTs in the target language. We find that scaling reasoning model size improves multilingual task performance in En-CoT, but Target-CoT performance lags behind. This gap widens for tasks requiring long, multi-step CoTs such as mathematical reasoning. Shifting to pretraining, we find that adding a specialized reasoning stage enhances En-CoT performance but degrades Target-CoT, whereas broad multilingual pretraining improves both modes simultaneously. Given the scarcity of high-quality reasoning traces in languages other than English, we explore synthetic data curation approaches for post-training. We demonstrate that fine-tuning on reasoning traces automatically translated from gold English traces outperforms fine-tuning on target-language traces distilled from large reasoning models. Finally, we report disparities in inference efficiency between languages and uncover language-specific failure modes in CoTs. We release models, datasets, and code to foster further research.", "authors": ["Josh Barua", "Seun Eisape", "Kayo Yin", "Alane Suhr"], "categories": ["cs.CL", "cs.AI", "cs.LG"], "fields_of_study": ["Computer Science"], "published_date": "2025-08-20", "url": "https://arxiv.org/abs/2508.14828", "pdf_url": "https://arxiv.org/pdf/2508.14828v3", "arxiv_id": "2508.14828", "doi": "10.48550/arXiv.2508.14828", "citation_count": 12, "influential_citation_count": 2, "has_code": false, "code_url": null, "venue": "arXiv.org", "quality_score": 0.2785}
{"id": "9b702849268b396d69430d346d1fc37c871b358ef9a9a657d7f4e0222c79d9ac", "sources": ["arxiv", "semantic_scholar"], "title": "Multimodal Chain of Continuous Thought for Latent-Space Reasoning in Vision-Language Models", "abstract": "Many reasoning techniques for large multimodal models adapt language model approaches, such as Chain-of-Thought (CoT) prompting, which express reasoning as word sequences. While effective for text, these methods are suboptimal for multimodal contexts, struggling to align audio, visual, and textual information dynamically. To explore an alternative paradigm, we propose the Multimodal Chain of Continuous Thought (MCOUT), which enables reasoning directly in a joint latent space rather than in natural language. In MCOUT, the reasoning state is represented as a continuous hidden vector, iteratively refined and aligned with visual and textual embeddings, inspired by human reflective cognition. We develop two variants: MCOUT-Base, which reuses the language model`s last hidden state as the continuous thought for iterative reasoning, and MCOUT-Multi, which integrates multimodal latent attention to strengthen cross-modal alignment between visual and textual features. Experiments on benchmarks including MMMU, ScienceQA, and MMStar show that MCOUT consistently improves multimodal reasoning, yielding up to 8.23% accuracy gains over strong baselines and improving BLEU scores up to 8.27% across multiple-choice and open-ended tasks. These findings highlight latent continuous reasoning as a promising direction for advancing LMMs beyond language-bound CoT, offering a scalable framework for human-like reflective multimodal inference. Code is available at https://github.com/Hanhpt23/OmniMod.", "authors": ["Tan-Hanh Pham", "Chris Ngo"], "categories": ["cs.CV"], "fields_of_study": ["Computer Science"], "published_date": "2025-08-18", "url": "https://arxiv.org/abs/2508.12587", "pdf_url": "https://arxiv.org/pdf/2508.12587v2", "arxiv_id": "2508.12587", "doi": "10.48550/arXiv.2508.12587", "citation_count": 13, "influential_citation_count": 0, "has_code": true, "code_url": "https://github.com/Hanhpt23/OmniMod", "venue": "arXiv.org", "quality_score": 0.4161}
{"id": "88a920267208facc180d3264fe20b7f0a23b46ada21bac9b4ead64fb22b99851", "sources": ["arxiv", "semantic_scholar"], "title": "Non-Interactive Symbolic-Aided Chain-of-Thought for Logical Reasoning", "abstract": "This work introduces Symbolic-Aided Chain-of-Thought (CoT), an improved approach to standard CoT, for logical reasoning in large language models (LLMs). The key idea is to integrate lightweight symbolic representations into few-shot prompts, structuring the inference steps with a consistent strategy to make reasoning patterns more explicit within a non-interactive reasoning process. By incorporating these symbolic structures, Symbolic-Aided CoT preserves the generalizability of standard prompting techniques while enhancing the transparency, interpretability, and analyzability of LLM logical reasoning. Extensive experiments on four well-known logical reasoning benchmarks -- ProofWriter, FOLIO, ProntoQA, and LogicalDeduction, which cover diverse reasoning tasks and scenarios -- demonstrate the effectiveness of the proposed approach, particularly in complex reasoning tasks that require navigating multiple constraints or rules. Notably, Symbolic-Aided CoT consistently improves LLMs' reasoning capabilities across various model sizes and significantly outperforms conventional CoT on three out of four datasets, ProofWriter, ProntoQA, and LogicalDeduction.", "authors": ["Phuong Minh Nguyen", "Tien Huu Dang", "Naoya Inoue"], "categories": ["cs.AI", "cs.CL"], "fields_of_study": ["Computer Science"], "published_date": "2025-08-17", "url": "https://arxiv.org/abs/2508.12425", "pdf_url": "https://arxiv.org/pdf/2508.12425v2", "arxiv_id": "2508.12425", "doi": null, "citation_count": 4, "influential_citation_count": 1, "has_code": false, "code_url": null, "venue": "Pacific Asia Conference on Language, Information and Computation", "quality_score": 0.2681}
{"id": "bae7c2161c2fb16ca0db16c02c502f470699434adf95015b1abda1b330aa26a2", "sources": ["arxiv", "semantic_scholar"], "title": "Audio Flamingo Sound-CoT Technical Report: Improving Chain-of-Thought Reasoning in Sound Understanding", "abstract": "Chain-of-thought reasoning has demonstrated significant improvements in large language models and vision language models, yet its potential for audio language models remains largely unexplored. In this technical report, we take a preliminary step towards closing this gap. For better assessment of sound reasoning, we propose AF-Reasoning-Eval, a benchmark targeting common-sense reasoning and the ability to discriminate among closely related choices. To prepare training corpus for sound reasoning abilities, we propose automatic pipelines that transform existing audio question answering and classification data into explicit reasoning chains, yielding AF-CoT-Train with 1.24M samples. We study the effect of finetuning Audio Flamingo series on AF-CoT-Train and observe considerable improvements on several reasoning benchmarks, validating the effectiveness of chain-of-thought finetuning on advanced sound understanding.", "authors": ["Zhifeng Kong", "Arushi Goel", "Joao Felipe Santos", "Sreyan Ghosh", "Rafael Valle", "Wei Ping", "Bryan Catanzaro"], "categories": ["cs.SD", "cs.LG"], "fields_of_study": ["Computer Science"], "published_date": "2025-08-15", "url": "https://arxiv.org/abs/2508.11818", "pdf_url": "https://arxiv.org/pdf/2508.11818v1", "arxiv_id": "2508.11818", "doi": "10.48550/arXiv.2508.11818", "citation_count": 7, "influential_citation_count": 0, "has_code": false, "code_url": null, "venue": "arXiv.org", "quality_score": 0.2658}
{"id": "73d80a2933fac18ba1b826fe0e88eb7e847722ca361118917fc45cff55351941", "sources": ["arxiv", "semantic_scholar"], "title": "Diagnosing Memorization in Chain-of-Thought Reasoning, One Token at a Time", "abstract": "Large Language Models (LLMs) perform well on reasoning benchmarks but often fail when inputs alter slightly, raising concerns about the extent to which their success relies on memorization. This issue is especially acute in Chain-of-Thought (CoT) reasoning, where spurious memorized patterns can trigger intermediate errors that cascade into incorrect final answers. We introduce STIM, a novel framework for Source-aware Token-level Identification of Memorization, which attributes each token in a reasoning chain to one of multiple memorization sources - local, mid-range, or long-range - based on their statistical co-occurrence with the token in the pretraining corpus. Our token-level analysis across tasks and distributional settings reveals that models rely more on memorization in complex or long-tail cases, and that local memorization is often the dominant driver of errors, leading to up to 67% of wrong tokens. We also show that memorization scores from STIM can be effective in predicting the wrong tokens in the wrong reasoning step. STIM offers a powerful tool for diagnosing and improving model reasoning and can generalize to other structured step-wise generation tasks.", "authors": ["Huihan Li", "You Chen", "Siyuan Wang", "Yixin He", "Ninareh Mehrabi", "Rahul Gupta", "Xiang Ren"], "categories": ["cs.CL", "cs.AI"], "fields_of_study": ["Computer Science"], "published_date": "2025-08-04", "url": "https://arxiv.org/abs/2508.02037", "pdf_url": "https://arxiv.org/pdf/2508.02037v2", "arxiv_id": "2508.02037", "doi": "10.48550/arXiv.2508.02037", "citation_count": 6, "influential_citation_count": 0, "has_code": false, "code_url": null, "venue": "Conference on Empirical Methods in Natural Language Processing", "quality_score": 0.2532}
{"id": "179e309c9e6d1410c0ae8607b9b598e8f94dae8aace574cd5f3fde377b72ebd0", "sources": ["arxiv", "semantic_scholar"], "title": "Is Chain-of-Thought Reasoning of LLMs a Mirage? A Data Distribution Lens", "abstract": "Chain-of-Thought (CoT) prompting has been shown to be effective in eliciting structured reasoning (i.e., CoT reasoning) from large language models (LLMs). Regardless of its popularity, recent studies expose its failures in some reasoning tasks, raising fundamental questions about the nature of CoT reasoning. In this work, we propose a data distribution lens to understand when and why CoT reasoning succeeds or fails. We hypothesize that CoT reasoning reflects a structured inductive bias learned from in-distribution data, enabling models to conditionally generate reasoning trajectories that approximate those observed during training. As such, the effectiveness of CoT reasoning is fundamentally governed by the nature and degree of distribution discrepancy between training data and test queries. Guided by this lens, we dissect CoT reasoning via three dimensions: task, length, and format. To test the hypothesis, we introduce DataAlchemy, an abstract and fully controllable environment that trains LLMs from scratch and systematically probes them under various distribution conditions. Through rigorous controlled experiments, we reveal that CoT reasoning is a brittle mirage when it is pushed beyond training distributions, emphasizing the ongoing challenge of achieving genuine and generalizable reasoning.", "authors": ["Chengshuai Zhao", "Zhen Tan", "Pingchuan Ma", "Dawei Li", "Bohan Jiang", "Yancheng Wang", "Yingzhen Yang", "Huan Liu"], "categories": ["cs.AI", "cs.CL", "cs.LG"], "fields_of_study": ["Computer Science"], "published_date": "2025-08-02", "url": "https://arxiv.org/abs/2508.01191", "pdf_url": "https://arxiv.org/pdf/2508.01191v6", "arxiv_id": "2508.01191", "doi": "10.48550/arXiv.2508.01191", "citation_count": 63, "influential_citation_count": 1, "has_code": false, "code_url": null, "venue": "arXiv.org", "quality_score": 0.4515}
{"id": "effc3aea874528668ead6080fbcdf84c4a654b5d35c31894f0bdf5e81fd44861", "sources": ["arxiv", "semantic_scholar"], "title": "Cognitive Chain-of-Thought (CoCoT): Structured Multimodal Reasoning about Social Situations", "abstract": "Chain-of-Thought (CoT) prompting helps models think step by step. But naive CoT breaks down in visually grounded social tasks, where models must perceive, understand, and judge all at once; bridging perception with norm-grounded reasoning. Recent work has introduced structured reasoning for multi-turn agent planning and visual QA, decomposing tasks into sequential sub-goals. To extend this to single-shot multimodal social reasoning, we introduce Cognitive Chain-of-Thought (CoCoT), a reasoning framework that structures vision-language-model (VLM) reasoning through three cognitively inspired stages: Perception (extract grounded facts), Situation (infer situations), and Norm (applying social norms). Evaluation across multiple distinct tasks such as multimodal intent disambiguation, multimodal theory of mind, social commonsense reasoning, and safety instruction following, shows consistent improvements (5.9% to 4.6% on average). We further explore the utility of CoCoT for improving models' reasoning through training and show that supervised fine-tuning on CoCoT-structured traces yields 5-6% improvements without explicit CoCoT prompting at inference, demonstrating that models internalize the structured reasoning pattern rather than merely following instructions. We show that structuring model reasoning through cognitively grounded stages enhances interpretability and social alignment, laying the groundwork for more reliable multimodal systems. All code and data will be released publicly.", "authors": ["Eunkyu Park", "Wesley Hanwen Deng", "Gunhee Kim", "Motahhare Eslami", "Maarten Sap"], "categories": ["cs.CL", "cs.AI", "cs.CY"], "fields_of_study": ["Computer Science"], "published_date": "2025-07-27", "url": "https://arxiv.org/abs/2507.20409", "pdf_url": "https://arxiv.org/pdf/2507.20409v2", "arxiv_id": "2507.20409", "doi": null, "citation_count": 3, "influential_citation_count": 0, "has_code": false, "code_url": null, "venue": null, "quality_score": 0.1553}
{"id": "43ce0d240f107a7c5604388796262378ed72538176e5be5ab5e116e3f12acaeb", "sources": ["arxiv", "semantic_scholar"], "title": "How does Chain of Thought Think? Mechanistic Interpretability of Chain-of-Thought Reasoning with Sparse Autoencoding", "abstract": "Chain-of-thought (CoT) prompting boosts Large Language Models accuracy on multi-step tasks, yet whether the generated \"thoughts\" reflect the true internal reasoning process is unresolved. We present the first feature-level causal study of CoT faithfulness. Combining sparse autoencoders with activation patching, we extract monosemantic features from Pythia-70M and Pythia-2.8B while they tackle GSM8K math problems under CoT and plain (noCoT) prompting. Swapping a small set of CoT-reasoning features into a noCoT run raises answer log-probabilities significantly in the 2.8B model, but has no reliable effect in 70M, revealing a clear scale threshold. CoT also leads to significantly higher activation sparsity and feature interpretability scores in the larger model, signalling more modular internal computation. For example, the model's confidence in generating correct answers improves from 1.2 to 4.3. We introduce patch-curves and random-feature patching baselines, showing that useful CoT information is not only present in the top-K patches but widely distributed. Overall, our results indicate that CoT can induce more interpretable internal structures in high-capacity LLMs, validating its role as a structured prompting method.", "authors": ["Xi Chen", "Aske Plaat", "Niki van Stein"], "categories": ["cs.CL", "cs.AI"], "fields_of_study": ["Computer Science"], "published_date": "2025-07-24", "url": "https://arxiv.org/abs/2507.22928", "pdf_url": "https://arxiv.org/pdf/2507.22928v1", "arxiv_id": "2507.22928", "doi": "10.48550/arXiv.2507.22928", "citation_count": 12, "influential_citation_count": 0, "has_code": false, "code_url": null, "venue": "AAAI Conference on Artificial Intelligence", "quality_score": 0.2785}
{"id": "cd1bca4a3bdda049aa316be52fa7d120b61cea886a53c41e3e723fe37fe89c81", "sources": ["arxiv", "semantic_scholar"], "title": "Zebra-CoT: A Dataset for Interleaved Vision Language Reasoning", "abstract": "Humans often use visual aids, for example diagrams or sketches, when solving complex problems. Training multimodal models to do the same, known as Visual Chain of Thought (Visual CoT), is challenging due to: (1) poor off-the-shelf visual CoT performance, which hinders reinforcement learning, and (2) the lack of high-quality visual CoT training data. We introduce $\\textbf{Zebra-CoT}$, a diverse large-scale dataset with 182,384 samples, containing logically coherent interleaved text-image reasoning traces. We focus on four categories of tasks where sketching or visual reasoning is especially natural, spanning scientific questions such as geometry, physics, and algorithms; 2D visual reasoning tasks like visual search and jigsaw puzzles; 3D reasoning tasks including 3D multi-hop inference, embodied and robot planning; visual logic problems and strategic games like chess. Fine-tuning the Anole-7B model on the Zebra-CoT training corpus results in an improvement of +12% in our test-set accuracy and yields up to +13% performance gain on standard VLM benchmark evaluations. Fine-tuning Bagel-7B yields a model that generates high-quality interleaved visual reasoning chains, underscoring Zebra-CoT's effectiveness for developing multimodal reasoning abilities. We open-source our dataset and models to support development and evaluation of visual CoT.", "authors": ["Ang Li", "Charles Wang", "Deqing Fu", "Kaiyu Yue", "Zikui Cai", "Wang Bill Zhu", "Ollie Liu", "Peng Guo", "Willie Neiswanger", "Furong Huang", "Tom Goldstein", "Micah Goldblum"], "categories": ["cs.CV", "cs.CL", "cs.LG"], "fields_of_study": ["Computer Science"], "published_date": "2025-07-22", "url": "https://arxiv.org/abs/2507.16746", "pdf_url": "https://arxiv.org/pdf/2507.16746v2", "arxiv_id": "2507.16746", "doi": "10.48550/arXiv.2507.16746", "citation_count": 53, "influential_citation_count": 8, "has_code": true, "code_url": null, "venue": "arXiv.org", "quality_score": 0.4771}
{"id": "9b2d27f5160fd0564bfe3f51ebec8be38072697626ac16aab0b04dfd9cd61e57", "sources": ["arxiv", "semantic_scholar"], "title": "Chart-R1: Chain-of-Thought Supervision and Reinforcement for Advanced Chart Reasoner", "abstract": "Chart reasoning presents unique challenges due to its inherent complexity -- requiring precise numerical comprehension, multi-level visual understanding, and logical inference across interconnected data elements. Existing vision-language models often struggle with such reasoning tasks, particularly when handling multi-subchart scenarios and numerical sensitivity. To address these challenges, we introduce Chart-R1, a chart-domain vision-language model that leverages reinforcement fine-tuning for advanced chart reasoning. We first propose a programmatic data synthesis approach to generate high-quality step-by-step reasoning data with verifiable answer formats, covering diverse chart types and complexity levels. Our two-stage training strategy includes: (1) Chart-COT, which decomposes complex reasoning into interpretable subtasks through chain-of-thought supervision, and (2) Chart-RFT, which employs group relative policy optimization with numerically sensitive rewards tailored for chart-specific reasoning. Experiments on open-source benchmarks and our proposed ChartRQA dataset demonstrate that Chart-R1 significantly outperforms existing chart-domain methods and rivals large-scale open/closed-source models.", "authors": ["Lei Chen", "Xuanle Zhao", "Zhixiong Zeng", "Jing Huang", "Yufeng Zhong", "Lin Ma"], "categories": ["cs.AI", "cs.CV"], "fields_of_study": ["Computer Science"], "published_date": "2025-07-21", "url": "https://arxiv.org/abs/2507.15509", "pdf_url": "https://arxiv.org/pdf/2507.15509v3", "arxiv_id": "2507.15509", "doi": "10.48550/arXiv.2507.15509", "citation_count": 18, "influential_citation_count": 3, "has_code": true, "code_url": null, "venue": "arXiv.org", "quality_score": 0.3666}
{"id": "31ca141634e3463ee72f5ba8229482f203d57bde0289711a31f38d139465489a", "sources": ["arxiv", "semantic_scholar"], "title": "Inverse Scaling in Test-Time Compute", "abstract": "We construct evaluation tasks where extending the reasoning length of Large Reasoning Models (LRMs) deteriorates performance, exhibiting an inverse scaling relationship between test-time compute and accuracy. Our evaluation tasks span four categories: simple counting tasks with distractors, regression tasks with spurious features, deduction tasks with constraint tracking, and advanced AI risks. We identify five distinct failure modes when models reason for longer: 1) Claude models become increasingly distracted by irrelevant information; 2) OpenAI o-series models resist distractors but overfit to problem framings; 3) models shift from reasonable priors to spurious correlations; 4) all models show difficulties in maintaining focus on complex deductive tasks; and 5) extended reasoning may amplify concerning behaviors, with Claude Sonnet 4 showing increased expressions of self-preservation. These findings suggest that while test-time compute scaling remains promising for improving model capabilities, it may inadvertently reinforce problematic reasoning patterns. Our results demonstrate the importance of evaluating models across diverse reasoning lengths to identify and address these failure modes in LRMs.", "authors": ["Aryo Pradipta Gema", "Alexander Hägele", "Runjin Chen", "Andy Arditi", "Jacob Goldman-Wetzler", "Kit Fraser-Taliente", "Henry Sleight", "Linda Petrini", "Julian Michael", "Beatrice Alex", "Pasquale Minervini", "Yanda Chen", "Joe Benton", "Ethan Perez"], "categories": ["cs.AI", "cs.CL"], "fields_of_study": ["Computer Science"], "published_date": "2025-07-19", "url": "https://arxiv.org/abs/2507.14417", "pdf_url": "https://arxiv.org/pdf/2507.14417v2", "arxiv_id": "2507.14417", "doi": "10.48550/arXiv.2507.14417", "citation_count": 42, "influential_citation_count": 4, "has_code": false, "code_url": null, "venue": "Transactions on Machine Learning Research (TMLR); 12/2025; https://openreview.net/forum?id=NXgyHW1c7M", "quality_score": 0.4084}
{"id": "35a9e4610f64c528007bdff7bcf4d687a40045dc8ff09f1c6f9f70ea125f1fd1", "sources": ["arxiv", "semantic_scholar"], "title": "Change of Thought: Adaptive Test-Time Computation", "abstract": "Transformers evaluated in a single, fixed-depth pass are provably limited in expressive power to the constant-depth circuit class TC0. Running a Transformer autoregressively removes that ceiling -- first in next-token prediction and, more recently, in chain-of-thought reasoning. Both regimes rely on feedback loops that decode internal states into tokens only to re-encode them in subsequent steps. While this \"thinking aloud\" mirrors human reasoning, biological brains iterate without externalising intermediate states as language. To boost the expressive power of encoder Transformers without resorting to token-level autoregression, we introduce the SELF-Transformer: an encoder layer that iteratively refines its own attention weights to a fixed point. Instead of producing -- in one pass -- the alignment matrix that remixes the input sequence, the SELF-Transformer iteratively updates that matrix internally, scaling test-time computation with input difficulty. This adaptivity yields up to 20\\% accuracy gains on encoder-style benchmarks without increasing parameter count, demonstrating that input-adaptive alignment at test time offers substantial benefits for only a modest extra compute budget. Self-Transformers thus recover much of the expressive power of iterative reasoning while preserving the simplicity of pure encoder architectures.", "authors": ["Mrinal Mathur", "Mike Doan", "Barak Pearlmutter", "Sergey Plis"], "categories": ["cs.LG", "cs.AI"], "fields_of_study": ["Computer Science"], "published_date": "2025-07-17", "url": "https://arxiv.org/abs/2507.13569", "pdf_url": "https://arxiv.org/pdf/2507.13569v1", "arxiv_id": "2507.13569", "doi": "10.48550/arXiv.2507.13569", "citation_count": 0, "influential_citation_count": 0, "has_code": false, "code_url": null, "venue": "arXiv.org", "quality_score": 0.2326}
{"id": "281fd7127af366a09d190285b1c71ba5f9dfdf6cc119ce9a6a5ae662f7d361ad", "sources": ["arxiv", "semantic_scholar"], "title": "Thought Purity: A Defense Framework For Chain-of-Thought Attack", "abstract": "Large Reasoning Models (LRMs) leverage Chain-of-Thought (CoT) reasoning to solve complex tasks, but this explicit reasoning process introduces a critical vulnerability: adversarial manipulation of the thought chain itself, known as Chain-of-Thought Attacks (CoTA). Such attacks subtly corrupt the reasoning path to produce erroneous outputs, challenging conventional defenses that often sacrifice model utility for safety. To address this, we propose Thought Purity(TP), a defense framework that shifts from passive refusal to active reasoning recovery. TP integrates a safety-aware data pipeline with reinforcement learning, employing a dual-reward mechanism to teach models to dynamically identify and isolate malicious logic while preserving correct reasoning. Experiments on multiple model families demonstrate that TP significantly reduces the attack success rate of CoTA while maintaining or enhancing the model's performance on benign tasks.", "authors": ["Zihao Xue", "Zhen Bi", "Long Ma", "Zhenlin Hu", "Yan Wang", "Xueshu Chen", "Zhenfang Liu", "Kang Zhao", "Jie Xiao", "Jungang Lou"], "categories": ["cs.LG", "cs.AI", "cs.CE", "cs.CR"], "fields_of_study": ["Computer Science"], "published_date": "2025-07-16", "url": "https://arxiv.org/abs/2507.12314", "pdf_url": "https://arxiv.org/pdf/2507.12314v3", "arxiv_id": "2507.12314", "doi": null, "citation_count": 3, "influential_citation_count": 0, "has_code": false, "code_url": null, "venue": null, "quality_score": 0.1505}
{"id": "d740622f0f6db43f74c93cd0cd20fd9f9a1d903487e3a1d40ab6a61a87e90ba3", "sources": ["arxiv", "semantic_scholar"], "title": "MindJourney: Test-Time Scaling with World Models for Spatial Reasoning", "abstract": "Spatial reasoning in 3D space is central to human cognition and indispensable for embodied tasks such as navigation and manipulation. However, state-of-the-art vision-language models (VLMs) struggle frequently with tasks as simple as anticipating how a scene will look after an egocentric motion: they perceive 2D images but lack an internal model of 3D dynamics. We therefore propose MindJourney, a test-time scaling framework that grants a VLM with this missing capability by coupling it to a controllable world model based on video diffusion. The VLM iteratively sketches a concise camera trajectory, while the world model synthesizes the corresponding view at each step. The VLM then reasons over this multi-view evidence gathered during the interactive exploration. Without any fine-tuning, our MindJourney achieves over an average 7.7% performance boost on the representative spatial reasoning benchmark SAT, showing that pairing VLMs with world models for test-time scaling offers a simple, plug-and-play route to robust 3D reasoning. Meanwhile, our method also improves upon the test-time inference VLMs trained through reinforcement learning, which demonstrates the potential of our method that utilizes world models for test-time scaling.", "authors": ["Yuncong Yang", "Jiageng Liu", "Zheyuan Zhang", "Siyuan Zhou", "Reuben Tan", "Jianwei Yang", "Yilun Du", "Chuang Gan"], "categories": ["cs.CV", "cs.AI", "cs.RO"], "fields_of_study": ["Computer Science"], "published_date": "2025-07-16", "url": "https://arxiv.org/abs/2507.12508", "pdf_url": "https://arxiv.org/pdf/2507.12508v2", "arxiv_id": "2507.12508", "doi": "10.48550/arXiv.2507.12508", "citation_count": 36, "influential_citation_count": 8, "has_code": false, "code_url": null, "venue": "arXiv.org", "quality_score": 0.4771}
{"id": "6d62ba5b51ae7da0a21a77d38051a21380dae36a90c34e0a9f5720d38407d653", "sources": ["arxiv", "semantic_scholar"], "title": "Deep Hidden Cognition Facilitates Reliable Chain-of-Thought Reasoning", "abstract": "Chain of Thought (CoT) reasoning has demonstrated remarkable deep reasoning capabilities in both large language models (LLMs) and multimodal large language models (MLLMs). However, its reliability is often undermined by the accumulation of errors in intermediate steps. This paper introduces an novel approach to calibrate the CoT reasoning accuracy by leveraging the model's intrinsic veracity encoding. We discover that specific attention head activations reliably reflect the truthfulness of reasoning steps in CoT. Based on this insight, we train a confidence predictor to evaluate the correctness of each reasoning step using these truthfulness-sensitive activations, dynamically selecting the most plausible reasoning path via beam search. Experimental results demonstrate that our method significantly outperforms the state-of-the-art baselines (e.g., Few-Shot CoT, Self-Consistency, and Self-Evaluation Guided Beam Search) across the mathematical, symbolic, and commonsense reasoning tasks, exhibiting superior accuracy and reliability in both unimodal and multimodal settings. We further validate the approach on large reasoning models, confirming its applicability to specialized reasoning models. Additionally, we explore the role of the model's self-correction ability in CoT reasoning. This work provides a novel reliability improvement path for CoT reasoning with broad application potential.", "authors": ["Zijun Chen", "Wenbo Hu", "Richang Hong"], "categories": ["cs.AI"], "fields_of_study": ["Computer Science"], "published_date": "2025-07-14", "url": "https://arxiv.org/abs/2507.10007", "pdf_url": "https://arxiv.org/pdf/2507.10007v2", "arxiv_id": "2507.10007", "doi": "10.48550/arXiv.2507.10007", "citation_count": 1, "influential_citation_count": 0, "has_code": false, "code_url": null, "venue": "AAAI Conference on Artificial Intelligence", "quality_score": 0.2292}
{"id": "12a0e62ca123d9297716257bb090e4c865761f8e2b35e275d6c1f65d6790e9e6", "sources": ["arxiv", "semantic_scholar"], "title": "Corvid: Improving Multimodal Large Language Models Towards Chain-of-Thought Reasoning", "abstract": "Recent advancements in multimodal large language models (MLLMs) have demonstrated exceptional performance in multimodal perception and understanding. However, leading open-source MLLMs exhibit significant limitations in complex and structured reasoning, particularly in tasks requiring deep reasoning for decision-making and problem-solving. In this work, we present Corvid, an MLLM with enhanced chain-of-thought (CoT) reasoning capabilities. Architecturally, Corvid incorporates a hybrid vision encoder for informative visual representation and a meticulously designed connector (GateMixer) to facilitate cross-modal alignment. To enhance Corvid's CoT reasoning capabilities, we introduce MCoT-Instruct-287K, a high-quality multimodal CoT instruction-following dataset, refined and standardized from diverse public reasoning sources. Leveraging this dataset, we fine-tune Corvid with a two-stage CoT-formatted training approach to progressively enhance its step-by-step reasoning abilities. Furthermore, we propose an effective inference-time scaling strategy that enables Corvid to mitigate over-reasoning and under-reasoning through self-verification. Extensive experiments demonstrate that Corvid outperforms existing o1-like MLLMs and state-of-the-art MLLMs with similar parameter scales, with notable strengths in mathematical reasoning and science problem-solving. Project page: https://mm-vl.github.io/corvid.", "authors": ["Jingjing Jiang", "Chao Ma", "Xurui Song", "Hanwang Zhang", "Jun Luo"], "categories": ["cs.CV"], "fields_of_study": ["Computer Science"], "published_date": "2025-07-10", "url": "https://arxiv.org/abs/2507.07424", "pdf_url": "https://arxiv.org/pdf/2507.07424v1", "arxiv_id": "2507.07424", "doi": "10.1109/ICCV51701.2025.00291", "citation_count": 12, "influential_citation_count": 1, "has_code": true, "code_url": null, "venue": "IEEE International Conference on Computer Vision", "quality_score": 0.3471}
{"id": "bfe7899fbcfee94915ee121fd09f4802316835979167c831b2f9de82c351d7b6", "sources": ["arxiv", "semantic_scholar"], "title": "Video-RTS: Rethinking Reinforcement Learning and Test-Time Scaling for Efficient and Enhanced Video Reasoning", "abstract": "Despite advances in reinforcement learning (RL)-based video reasoning with large language models (LLMs), data collection and fine-tuning remain significant challenges. These methods often rely on large-scale supervised fine-tuning (SFT) with extensive video data and long Chain-of-Thought (CoT) annotations, making them costly and hard to scale. To address this, we present Video-RTS, a new approach to improve video reasoning capability with drastically improved data efficiency by combining data-efficient RL with a video-adaptive test-time scaling (TTS) strategy. Building on observations about the data scaling, we skip the resource-intensive SFT step and employ efficient pure-RL training with output-based rewards, requiring no additional annotations or extensive fine-tuning. Furthermore, to utilize computational resources more efficiently, we introduce a sparse-to-dense video TTS strategy that improves inference by iteratively adding frames based on output consistency. We validate our approach on multiple video reasoning benchmarks, showing that Video-RTS surpasses existing video reasoning models by 2.4% in accuracy using only 3.6% training samples. Specifically, Video-RTS achieves a 4.2% improvement on Video-Holmes, a recent and challenging video reasoning benchmark. Notably, our pure RL training and adaptive video TTS offer complementary strengths, enabling Video-RTS's strong reasoning performance.", "authors": ["Ziyang Wang", "Jaehong Yoon", "Shoubin Yu", "Md Mohaiminul Islam", "Gedas Bertasius", "Mohit Bansal"], "categories": ["cs.CV", "cs.AI", "cs.CL"], "fields_of_study": ["Computer Science"], "published_date": "2025-07-09", "url": "https://arxiv.org/abs/2507.06485", "pdf_url": "https://arxiv.org/pdf/2507.06485v2", "arxiv_id": "2507.06485", "doi": "10.48550/arXiv.2507.06485", "citation_count": 27, "influential_citation_count": 3, "has_code": false, "code_url": null, "venue": "Conference on Empirical Methods in Natural Language Processing", "quality_score": 0.3618}
{"id": "fc9179fdf89afc8100004c0112920249c6d37742a76a5680838a4da3fc7739c8", "sources": ["arxiv", "semantic_scholar"], "title": "ESTR-CoT: Towards Explainable and Accurate Event Stream based Scene Text Recognition with Chain-of-Thought Reasoning", "abstract": "Event stream based scene text recognition is a newly arising research topic in recent years which performs better than the widely used RGB cameras in extremely challenging scenarios, especially the low illumination, fast motion. Existing works either adopt end-to-end encoder-decoder framework or large language models for enhanced recognition, however, they are still limited by the challenges of insufficient interpretability and weak contextual logical reasoning. In this work, we propose a novel chain-of-thought reasoning based event stream scene text recognition framework, termed ESTR-CoT. Specifically, we first adopt the vision encoder EVA-CLIP (ViT-G/14) to transform the input event stream into tokens and utilize a Llama tokenizer to encode the given generation prompt. A Q-former is used to align the vision token to the pre-trained large language model Vicuna-7B and output both the answer and chain-of-thought (CoT) reasoning process simultaneously. Our framework can be optimized using supervised fine-tuning in an end-to-end manner. In addition, we also propose a large-scale CoT dataset to train our framework via a three stage processing (i.e., generation, polish, and expert verification). This dataset provides a solid data foundation for the development of subsequent reasoning-based large models. Extensive experiments on three event stream STR benchmark datasets (i.e., EventSTR, WordArt*, IC15*) fully validated the effectiveness and interpretability of our proposed framework. The source code and pre-trained models will be released on https://github.com/Event-AHU/ESTR-CoT.", "authors": ["Xiao Wang", "Jingtao Jiang", "Qiang Chen", "Lan Chen", "Lin Zhu", "Yaowei Wang", "Yonghong Tian", "Jin Tang"], "categories": ["cs.CV", "cs.AI", "cs.CL"], "fields_of_study": ["Computer Science"], "published_date": "2025-07-02", "url": "https://arxiv.org/abs/2507.02200", "pdf_url": "https://arxiv.org/pdf/2507.02200v1", "arxiv_id": "2507.02200", "doi": "10.48550/arXiv.2507.02200", "citation_count": 3, "influential_citation_count": 0, "has_code": true, "code_url": "https://github.com/Event-AHU/ESTR-CoT", "venue": "arXiv.org", "quality_score": 0.3329}
{"id": "e516d087e2824030897e449179b33cff3d7e95db9c974cb36f61976783599574", "sources": ["arxiv", "semantic_scholar"], "title": "Reasoning on a Budget: A Survey of Adaptive and Controllable Test-Time Compute in LLMs", "abstract": "Large language models (LLMs) have rapidly progressed into general-purpose agents capable of solving a broad spectrum of tasks. However, current models remain inefficient at reasoning: they apply fixed inference-time compute regardless of task complexity, often overthinking simple problems while underthinking hard ones. This survey presents a comprehensive review of efficient test-time compute (TTC) strategies, which aim to improve the computational efficiency of LLM reasoning. We introduce a two-tiered taxonomy that distinguishes between L1-controllability, methods that operate under fixed compute budgets, and L2-adaptiveness, methods that dynamically scale inference based on input difficulty or model confidence. We benchmark leading proprietary LLMs across diverse datasets, highlighting critical trade-offs between reasoning performance and token usage. Compared to prior surveys on efficient reasoning, our review emphasizes the practical control, adaptability, and scalability of TTC methods. Finally, we discuss emerging trends such as hybrid thinking models and identify key challenges for future work towards making LLMs more computationally efficient, robust, and responsive to user constraints.", "authors": ["Mohammad Ali Alomrani", "Yingxue Zhang", "Derek Li", "Qianyi Sun", "Soumyasundar Pal", "Zhanguang Zhang", "Yaochen Hu", "Rohan Deepak Ajwani", "Antonios Valkanas", "Raika Karimi", "Peng Cheng", "Yunzhou Wang", "Pengyi Liao", "Hanrui Huang", "Bin Wang", "Jianye Hao", "Mark Coates"], "categories": ["cs.AI", "cs.LG"], "fields_of_study": ["Computer Science"], "published_date": "2025-07-02", "url": "https://arxiv.org/abs/2507.02076", "pdf_url": "https://arxiv.org/pdf/2507.02076v1", "arxiv_id": "2507.02076", "doi": "10.48550/arXiv.2507.02076", "citation_count": 21, "influential_citation_count": 6, "has_code": false, "code_url": null, "venue": "arXiv.org", "quality_score": 0.4225}
{"id": "f8b216cb0bf198ed159f315758893fdbfd6b9ba8a816a443d3dabf747791706f", "sources": ["arxiv", "semantic_scholar"], "title": "Latent Chain-of-Thought? Decoding the Depth-Recurrent Transformer", "abstract": "Chain-of-thought (CoT) reasoning has enabled transformer-based language models to excel at complex mathematics and multi-step planning. However, in standard decoder-only architectures, these reasoning steps are externalized in natural language, improving interpretability at the cost of efficiency. To capture reasoning that is not easily represented in words, many works have explored recurrent architectures that aim to internalize reasoning in latent space, potentially supporting latent CoT. In this paper, we investigate whether such reasoning structures emerge in Huginn-3.5B, a depth-recurrent Transformer that reuses layers at inference time without increasing parameter count. We examine the model's internal behavior on arithmetic tasks using a suite of probing techniques including the Logit Lens and Coda Lens. Our findings reveal limited evidence of interpretable latent CoT by tracking rank trajectories of final and intermediate result tokens. Furthermore, we uncover significant probing inconsistencies across recurrent blocks, where the interpretability of hidden states depends heavily on both the layer index and the decoding method. Finally, we empirically show that increasing recurrence depth yields only marginal gains and falls well short of models that explicitly externalize reasoning steps. The code is available at https://github.com/wenquanlu/huginn-latent-cot.", "authors": ["Wenquan Lu", "Yuechuan Yang", "Kyle Lee", "Yanshu Li", "Enqi Liu"], "categories": ["cs.CL", "cs.AI", "cs.LG"], "fields_of_study": ["Computer Science"], "published_date": "2025-07-02", "url": "https://arxiv.org/abs/2507.02199", "pdf_url": "https://arxiv.org/pdf/2507.02199v2", "arxiv_id": "2507.02199", "doi": "10.48550/arXiv.2507.02199", "citation_count": 10, "influential_citation_count": 0, "has_code": true, "code_url": "https://github.com/wenquanlu/huginn-latent-cot", "venue": "arXiv.org", "quality_score": 0.3329}
{"id": "81a24c1bec0812e7133a1ff984a3ce8a2d9cc81f0cfd1333d94b463ba38f8b73", "sources": ["arxiv", "semantic_scholar"], "title": "Can Large Language Models Develop Strategic Reasoning? Post-training Insights from Learning Chess", "abstract": "While reinforcement learning (RL) for large language models (LLMs) has shown promise in mathematical reasoning, strategic reasoning for LLMs using RL remains largely unexplored. We investigate whether LLMs can develop strategic reasoning capabilities through RL in chess. To this end, we leverage a chess-pretrained action-value network to provide dense reward on the LLM's output move quality, which can be seen as a form of knowledge distillation. Our experiments show that our distillation-based dense rewards often outperform sparse binary rewards. However, surprisingly, all models plateau far below expert levels. We provide SFT and RL ablations on chess reasoning training and find evidence that this limitation stems from a deficit in the pretrained models' internal understanding of chess-a deficit which RL alone may not be able to fully overcome. The code is available at https://github.com/krafton-ai/Chess-R1.", "authors": ["Dongyoon Hwang", "Hojoon Lee", "Jaegul Choo", "Dongmin Park", "Jongho Park"], "categories": ["cs.AI", "cs.LG"], "fields_of_study": ["Computer Science"], "published_date": "2025-07-01", "url": "https://arxiv.org/abs/2507.00726", "pdf_url": "https://arxiv.org/pdf/2507.00726v3", "arxiv_id": "2507.00726", "doi": "10.48550/arXiv.2507.00726", "citation_count": 6, "influential_citation_count": 2, "has_code": true, "code_url": "https://github.com/krafton-ai/Chess-R1", "venue": "arXiv.org", "quality_score": 0.3311}
{"id": "66fecae8014ebed1002d49d032ded53fd2fa8413f3c8ed1ea6f4be4a4619853f", "sources": ["arxiv", "semantic_scholar"], "title": "Interactive Reasoning: Visualizing and Controlling Chain-of-Thought Reasoning in Large Language Models", "abstract": "The output quality of large language models (LLMs) can be improved via \"reasoning\": generating segments of chain-of-thought (CoT) content to further condition the model prior to producing user-facing output. While these chains contain valuable information, they are verbose and lack explicit organization, making them tedious to review. Moreover, they lack opportunities for user feedback, such as to remove unwanted considerations, add desired ones, or clarify unclear assumptions. We introduce Interactive Reasoning, an interaction design that visualizes chain-of-thought outputs as a hierarchy of topics and enables user review and modification. We implement interactive reasoning in Hippo, a prototype for AI-assisted decision making in the face of uncertain trade-offs. In a user study with 16 participants, we find that interactive reasoning in Hippo allows users to quickly identify and interrupt erroneous generations, efficiently steer the model towards customized responses, and better understand both model reasoning and model outputs. Our work contributes to a new paradigm that incorporates user oversight into LLM reasoning processes.", "authors": ["Rock Yuren Pang", "K. J. Kevin Feng", "Shangbin Feng", "Chu Li", "Weijia Shi", "Yulia Tsvetkov", "Jeffrey Heer", "Katharina Reinecke"], "categories": ["cs.HC", "cs.AI"], "fields_of_study": ["Computer Science"], "published_date": "2025-06-30", "url": "https://arxiv.org/abs/2506.23678", "pdf_url": "https://arxiv.org/pdf/2506.23678v1", "arxiv_id": "2506.23678", "doi": "10.48550/arXiv.2506.23678", "citation_count": 14, "influential_citation_count": 0, "has_code": false, "code_url": null, "venue": "International Conference on Intelligent User Interfaces", "quality_score": 0.294}
{"id": "d5ff59c843c3bdbf8805570bbb824474fd815a3f4d87cc72ffd15e27350dea3d", "sources": ["arxiv", "semantic_scholar"], "title": "Accurate Parameter-Efficient Test-Time Adaptation for Time Series Forecasting", "abstract": "Real-world time series often exhibit a non-stationary nature, degrading the performance of pre-trained forecasting models. Test-Time Adaptation (TTA) addresses this by adjusting models during inference, but existing methods typically update the full model, increasing memory and compute costs. We propose PETSA, a parameter-efficient method that adapts forecasters at test time by only updating small calibration modules on the input and output. PETSA uses low-rank adapters and dynamic gating to adjust representations without retraining. To maintain accuracy despite limited adaptation capacity, we introduce a specialized loss combining three components: (1) a robust term, (2) a frequency-domain term to preserve periodicity, and (3) a patch-wise structural term for structural alignment. PETSA improves the adaptability of various forecasting backbones while requiring fewer parameters than baselines. Experimental results on benchmark datasets show that PETSA achieves competitive or better performance across all horizons. Our code is available at: https://github.com/BorealisAI/PETSA", "authors": ["Heitor R. Medeiros", "Hossein Sharifi-Noghabi", "Gabriel L. Oliveira", "Saghar Irandoust"], "categories": ["cs.LG", "cs.AI"], "fields_of_study": ["Computer Science"], "published_date": "2025-06-29", "url": "https://arxiv.org/abs/2506.23424", "pdf_url": "https://arxiv.org/pdf/2506.23424v1", "arxiv_id": "2506.23424", "doi": "10.48550/arXiv.2506.23424", "citation_count": 8, "influential_citation_count": 2, "has_code": true, "code_url": "https://github.com/BorealisAI/PETSA", "venue": "arXiv.org", "quality_score": 0.3276}
{"id": "2c01f8fe09ef8b4d28c1e29d28d367374d7eb191e01de36a5a2f331ba52e63f4", "sources": ["arxiv", "semantic_scholar"], "title": "ECCoT: A Framework for Enhancing Effective Cognition via Chain of Thought in Large Language Model", "abstract": "In the era of large-scale artificial intelligence, Large Language Models (LLMs) have made significant strides in natural language processing. However, they often lack transparency and generate unreliable outputs, raising concerns about their interpretability. To address this, the Chain of Thought (CoT) prompting method structures reasoning into step-by-step deductions. Yet, not all reasoning chains are valid, and errors can lead to unreliable conclusions. We propose ECCoT, an End-to-End Cognitive Chain of Thought Validation Framework, to evaluate and refine reasoning chains in LLMs. ECCoT integrates the Markov Random Field-Embedded Topic Model (MRF-ETM) for topic-aware CoT generation and Causal Sentence-BERT (CSBert) for causal reasoning alignment. By filtering ineffective chains using structured ordering statistics, ECCoT improves interpretability, reduces biases, and enhances the trustworthiness of LLM-based decision-making. Key contributions include the introduction of ECCoT, MRF-ETM for topic-driven CoT generation, and CSBert for causal reasoning enhancement. Code is released at: https://github.com/erwinmsmith/ECCoT.git.", "authors": ["Zhenke Duan", "Jiqun Pan", "Jiani Tu", "Xiaoyi Wang", "Yanqing Wang"], "categories": ["cs.CL", "cs.AI"], "fields_of_study": ["Computer Science"], "published_date": "2025-06-24", "url": "https://arxiv.org/abs/2506.19599", "pdf_url": "https://arxiv.org/pdf/2506.19599v1", "arxiv_id": "2506.19599", "doi": "10.48550/arXiv.2506.19599", "citation_count": 0, "influential_citation_count": 0, "has_code": true, "code_url": "https://github.com/erwinmsmith/ECCoT.git", "venue": "Annual Meeting of the Cognitive Science Society", "quality_score": 0.3187}
{"id": "6ffb54bb6f0e1546c2bab09630a618972155bb18ee33164b2916b7eb2895f4ae", "sources": ["arxiv", "semantic_scholar"], "title": "Parallel Continuous Chain-of-Thought with Jacobi Iteration", "abstract": "Continuous chain-of-thought has been shown to be effective in saving reasoning tokens for large language models. By reasoning with continuous latent thought tokens, continuous CoT is able to perform implicit reasoning in a compact manner. However, the sequential dependencies between latent thought tokens spoil parallel training, leading to long training time. In this paper, we propose Parallel Continuous Chain-of-Thought (PCCoT), which performs Jacobi iteration on the latent thought tokens, updating them iteratively in parallel instead of sequentially and thus improving both training and inference efficiency of continuous CoT. Experiments demonstrate that by choosing the proper number of iterations, we are able to achieve comparable or even better performance while saving nearly 50% of the training and inference time. Moreover, PCCoT shows better stability and robustness in the training process. Our code is available at https://github.com/whyNLP/PCCoT.", "authors": ["Haoyi Wu", "Zhihao Teng", "Kewei Tu"], "categories": ["cs.CL"], "fields_of_study": ["Computer Science"], "published_date": "2025-06-23", "url": "https://arxiv.org/abs/2506.18582", "pdf_url": "https://arxiv.org/pdf/2506.18582v2", "arxiv_id": "2506.18582", "doi": "10.48550/arXiv.2506.18582", "citation_count": 16, "influential_citation_count": 3, "has_code": true, "code_url": "https://github.com/whyNLP/PCCoT", "venue": "Conference on Empirical Methods in Natural Language Processing", "quality_score": 0.317}
{"id": "67527d9bf6849dbcc56c8696694f884d5d9bc77108d0be3801ea1083713fdece", "sources": ["arxiv", "semantic_scholar"], "title": "Thought Anchors: Which LLM Reasoning Steps Matter?", "abstract": "Current frontier large-language models rely on reasoning to achieve state-of-the-art performance. Many existing interpretability are limited in this area, as standard methods have been designed to study single forward passes of a model rather than the multi-token computational steps that unfold during reasoning. We argue that analyzing reasoning traces at the sentence level is a promising approach to understanding reasoning processes. We introduce a black-box method that measures each sentence's counterfactual importance by repeatedly sampling replacement sentences from the model, filtering for semantically different ones, and continuing the chain of thought from that point onwards to quantify the sentence's impact on the distribution of final answers. We discover that certain sentences can have an outsized impact on the trajectory of the reasoning trace and final answer. We term these sentences \\textit{thought anchors}. These are generally planning or uncertainty management sentences, and specialized attention heads consistently attend from subsequent sentences to thought anchors. We further show that examining sentence-sentence causal links within a reasoning trace gives insight into a model's behavior. Such information can be used to predict a problem's difficulty and the extent different question domains involve sequential or diffuse reasoning. As a proof-of-concept, we demonstrate that our techniques together provide a practical toolkit for analyzing reasoning models by conducting a detailed case study of how the model solves a difficult math problem, finding that our techniques yield a consistent picture of the reasoning trace's structure. We provide an open-source tool (thought-anchors.com) for visualizing the outputs of our methods on further problems. The convergence across our methods shows the potential of sentence-level analysis for a deeper understanding of reasoning models.", "authors": ["Paul C. Bogdan", "Uzay Macar", "Neel Nanda", "Arthur Conmy"], "categories": ["cs.LG", "cs.AI", "cs.CL"], "fields_of_study": ["Computer Science"], "published_date": "2025-06-23", "url": "https://arxiv.org/abs/2506.19143", "pdf_url": "https://arxiv.org/pdf/2506.19143v4", "arxiv_id": "2506.19143", "doi": "10.48550/arXiv.2506.19143", "citation_count": 105, "influential_citation_count": 16, "has_code": true, "code_url": null, "venue": "arXiv.org", "quality_score": 0.6152}
{"id": "4a735fe87a31099720c17963faa5eca12b4a400d69b116b16a1a88242b77cad6", "sources": ["arxiv", "semantic_scholar"], "title": "Aha Moment Revisited: Are VLMs Truly Capable of Self Verification in Inference-time Scaling?", "abstract": "Inference time techniques such as decoding time scaling and self refinement have been shown to substantially improve mathematical reasoning in large language models (LLMs), largely attributed to emergent self correction and self verification behaviors often elicited through reinforcement learning (RL). In this work, we ask whether the same recipe transfers to vision language models (VLMs), especially RL finetuned variants that claim strong visual mathematical reasoning. Through extensive evaluation, we reach three main findings that differ markedly from text only models. First, generation time capability matters more than verification and refinement: simple majority voting consistently and substantially outperforms verification centric strategies such as best of N with self verification. Second, behaviors often associated with RL tuned models at inference time, such as the 'Aha moment,' do not yield reliable reasoning performance improvements. Third, visual information is not effectively integrated into the model's self verification process. Overall, our analysis highlights a key limitation: current RL trained VLMs derive limited benefit from self verification in the visual modality, which constrains the effectiveness of inference time scaling for visual mathematical reasoning.", "authors": ["Mingyuan Wu", "Meitang Li", "Jingcheng Yang", "Jize Jiang", "Kaizhuo Yan", "Zhaoheng Li", "Hanchao Yu", "Minjia Zhang", "Klara Nahrstedt"], "categories": ["cs.LG"], "fields_of_study": ["Computer Science"], "published_date": "2025-06-20", "url": "https://arxiv.org/abs/2506.17417", "pdf_url": "https://arxiv.org/pdf/2506.17417v3", "arxiv_id": "2506.17417", "doi": "10.48550/arXiv.2506.17417", "citation_count": 4, "influential_citation_count": 0, "has_code": false, "code_url": null, "venue": "arXiv.org", "quality_score": 0.2017}
{"id": "44af5c9852a2f54e772f707603f5456bb2b5292d3ec43c93d3e5b9c389b5d838", "sources": ["arxiv", "semantic_scholar"], "title": "On Training-Test (Mis)alignment in Unsupervised Combinatorial Optimization: Observation, Empirical Exploration, and Analysis", "abstract": "In unsupervised combinatorial optimization (UCO), during training, one aims to have continuous decisions that are promising in a probabilistic sense for each training instance, which enables end-to-end training on initially discrete and non-differentiable problems. At the test time, for each test instance, starting from continuous decisions, derandomization is typically applied to obtain the final deterministic decisions. Researchers have developed more and more powerful test-time derandomization schemes to enhance the empirical performance and the theoretical guarantee of UCO methods. However, we notice a misalignment between training and testing in the existing UCO methods. Consequently, lower training losses do not necessarily entail better post-derandomization performance, even for the training instances without any data distribution shift. Empirically, we indeed observe such undesirable cases. We explore a preliminary idea to better align training and testing in UCO by including a differentiable version of derandomization into training. Our empirical exploration shows that such an idea indeed improves training-test alignment, but also introduces nontrivial challenges into training.", "authors": ["Fanchen Bu", "Kijung Shin"], "categories": ["cs.LG", "cs.AI", "cs.DM", "math.PR"], "fields_of_study": ["Computer Science", "Mathematics"], "published_date": "2025-06-20", "url": "https://arxiv.org/abs/2506.16732", "pdf_url": "https://arxiv.org/pdf/2506.16732v1", "arxiv_id": "2506.16732", "doi": "10.48550/arXiv.2506.16732", "citation_count": 0, "influential_citation_count": 0, "has_code": false, "code_url": null, "venue": "arXiv.org", "quality_score": 0.2017}
{"id": "eec6afbd5fd5e4db8cb3aa89e5136b093dcb97aa680977c2120cc1ce57e3c228", "sources": ["arxiv", "semantic_scholar"], "title": "FinCoT: Grounding Chain-of-Thought in Expert Financial Reasoning", "abstract": "This paper presents FinCoT, a structured chain-of-thought (CoT) prompting framework that embeds domain-specific expert financial reasoning blueprints to guide large language models' behaviors. We identify three main prompting styles in financial NLP (FinNLP): (1) standard prompting (zero-shot), (2) unstructured CoT (free-form reasoning), and (3) structured CoT (with explicitly structured reasoning steps). Prior work has mainly focused on the first two, while structured CoT remains underexplored and lacks domain expertise incorporation. Therefore, we evaluate all three prompting approaches across ten CFA-style financial domains and introduce FinCoT as the first structured finance-specific prompting approach incorporating blueprints from domain experts. FinCoT improves the accuracy of a general-purpose model, Qwen3-8B-Base, from 63.2% to 80.5%, and boosts Fin-R1 (7B), a finance-specific model, from 65.7% to 75.7%, while reducing output length by up to 8.9x and 1.16x compared to structured CoT methods, respectively. We find that FinCoT proves most effective for models lacking financial post-training. Our findings show that FinCoT does not only improve performance and reduce inference costs but also yields more interpretable and expert-aligned reasoning traces.", "authors": ["Natapong Nitarach", "Warit Sirichotedumrong", "Panop Pitchayarthorn", "Pittawat Taveekitworachai", "Potsawee Manakul", "Kunat Pipatanakul"], "categories": ["cs.CL"], "fields_of_study": ["Computer Science"], "published_date": "2025-06-19", "url": "https://arxiv.org/abs/2506.16123", "pdf_url": "https://arxiv.org/pdf/2506.16123v5", "arxiv_id": "2506.16123", "doi": "10.48550/arXiv.2506.16123", "citation_count": 2, "influential_citation_count": 0, "has_code": false, "code_url": null, "venue": null, "quality_score": 0.1276}
{"id": "707eb9668896e8eeec44646d4ab2a0e1d2cce2a1b29de2e0e2b38ff4090188b1", "sources": ["arxiv", "semantic_scholar"], "title": "Fractional Reasoning via Latent Steering Vectors Improves Inference Time Compute", "abstract": "Test-time compute has emerged as a powerful paradigm for improving the performance of large language models (LLMs), where generating multiple outputs or refining individual chains can significantly boost answer accuracy. However, existing methods like Best-of-N, majority voting, and self-reflection typically apply reasoning in a uniform way across inputs, overlooking the fact that different problems may require different levels of reasoning depth. In this work, we propose Fractional Reasoning, a training-free and model-agnostic framework that enables continuous control over reasoning intensity at inference time, going beyond the limitations of fixed instructional prompts. Our method operates by extracting the latent steering vector associated with deeper reasoning and reapplying it with a tunable scaling factor, allowing the model to tailor its reasoning process to the complexity of each input. This supports two key modes of test-time scaling: (1) improving output quality in breadth-based strategies (e.g., Best-of-N, majority voting), and (2) enhancing the correctness of individual reasoning chains in depth-based strategies (e.g., self-reflection). Experiments on GSM8K, MATH500, and GPQA demonstrate that Fractional Reasoning consistently improves performance across diverse reasoning tasks and models.", "authors": ["Sheng Liu", "Tianlang Chen", "Pan Lu", "Haotian Ye", "Yizheng Chen", "Lei Xing", "James Zou"], "categories": ["cs.LG", "cs.AI", "cs.CL", "eess.SP"], "fields_of_study": ["Computer Science", "Engineering"], "published_date": "2025-06-18", "url": "https://arxiv.org/abs/2506.15882", "pdf_url": "https://arxiv.org/pdf/2506.15882v2", "arxiv_id": "2506.15882", "doi": "10.48550/arXiv.2506.15882", "citation_count": 11, "influential_citation_count": 1, "has_code": false, "code_url": null, "venue": "arXiv.org", "quality_score": 0.2698}
{"id": "c5f04d0f8afd4ccd77cebe5a04da53cf03a3af0a8377c2eb5f0c6573b0e0cd7d", "sources": ["arxiv", "semantic_scholar"], "title": "Scaling Test-time Compute for LLM Agents", "abstract": "Scaling test time compute has shown remarkable success in improving the reasoning abilities of large language models (LLMs). In this work, we conduct the first systematic exploration of applying test-time scaling methods to language agents and investigate the extent to which it improves their effectiveness. Specifically, we explore different test-time scaling strategies, including: (1) parallel sampling algorithms; (2) sequential revision strategies; (3) verifiers and merging methods; (4)strategies for diversifying rollouts.We carefully analyze and ablate the impact of different design strategies on applying test-time scaling on language agents, and have follow findings: 1. Scaling test time compute could improve the performance of agents. 2. Knowing when to reflect is important for agents. 3. Among different verification and result merging approaches, the list-wise method performs best. 4. Increasing diversified rollouts exerts a positive effect on the agent's task performance.", "authors": ["King Zhu", "Hanhao Li", "Siwei Wu", "Tianshun Xing", "Dehua Ma", "Xiangru Tang", "Minghao Liu", "Jian Yang", "Jiaheng Liu", "Yuchen Eleanor Jiang", "Changwang Zhang", "Chenghua Lin", "Jun Wang", "Ge Zhang", "Wangchunshu Zhou"], "categories": ["cs.AI"], "fields_of_study": ["Computer Science"], "published_date": "2025-06-15", "url": "https://arxiv.org/abs/2506.12928", "pdf_url": "https://arxiv.org/pdf/2506.12928v1", "arxiv_id": "2506.12928", "doi": "10.48550/arXiv.2506.12928", "citation_count": 35, "influential_citation_count": 0, "has_code": false, "code_url": null, "venue": "arXiv.org", "quality_score": 0.3891}
{"id": "841c7b9dcf65216dc01a424ddba0d8c0046863ed3150bf3bdda719ade88eff12", "sources": ["arxiv", "semantic_scholar"], "title": "MemeMind: A Large-Scale Multimodal Dataset with Chain-of-Thought Reasoning for Harmful Meme Detection", "abstract": "As a multimodal medium combining images and text, memes frequently convey implicit harmful content through metaphors and humor, rendering the detection of harmful memes a complex and challenging task. Although recent studies have made progress in detection accuracy and interpretability, large-scale, high-quality datasets for harmful memes remain scarce, and current methods still struggle to capture implicit risks and nuanced semantics. Thus, we construct MemeMind, a large-scale harmful meme dataset. Aligned with the international standards and the context of internet, MemeMind provides detailed Chain-of-Thought (CoT) reasoning annotations to support fine-grained analysis of implicit intentions in memes. Based on this dataset, we further propose MemeGuard, a reasoning-oriented multimodal detection framework that significantly improves both the accuracy of harmful meme detection and the interpretability of model decisions. Extensive experimental results demonstrate that MemeGuard outperforms existing state-of-the-art methods on the MemeMind dataset, establishing a solid foundation for future research in harmful meme detection. The complete dataset and code will be released upon acceptance.", "authors": ["Hexiang Gu", "Qifan Yu", "Yuan Liu", "Zikang Li", "Saihui Hou", "Jian Zhao", "Zhaofeng He"], "categories": ["cs.CL", "cs.AI", "cs.CV"], "fields_of_study": ["Computer Science"], "published_date": "2025-06-15", "url": "https://arxiv.org/abs/2506.18919", "pdf_url": "https://arxiv.org/pdf/2506.18919v4", "arxiv_id": "2506.18919", "doi": "10.48550/arXiv.2506.18919", "citation_count": 6, "influential_citation_count": 0, "has_code": false, "code_url": null, "venue": "arXiv.org", "quality_score": 0.2113}
{"id": "5050fd5d7ff58fa0c6d3550c133a09b1ca4fa17e0c4e0f5598f6d06e417dbe24", "sources": ["arxiv", "semantic_scholar"], "title": "Unveiling Confirmation Bias in Chain-of-Thought Reasoning", "abstract": "Chain-of-thought (CoT) prompting has been widely adopted to enhance the reasoning capabilities of large language models (LLMs). However, the effectiveness of CoT reasoning is inconsistent across tasks with different reasoning types. This work presents a novel perspective to understand CoT behavior through the lens of \\textit{confirmation bias} in cognitive psychology. Specifically, we examine how model internal beliefs, approximated by direct question-answering probabilities, affect both reasoning generation ($Q \\to R$) and reasoning-guided answer prediction ($QR \\to A$) in CoT. By decomposing CoT into a two-stage process, we conduct a thorough correlation analysis in model beliefs, rationale attributes, and stage-wise performance. Our results provide strong evidence of confirmation bias in LLMs, such that model beliefs not only skew the reasoning process but also influence how rationales are utilized for answer prediction. Furthermore, the interplay between task vulnerability to confirmation bias and the strength of beliefs also provides explanations for CoT effectiveness across reasoning tasks and models. Overall, this study provides a valuable insight for the needs of better prompting strategies that mitigate confirmation bias to enhance reasoning performance. Code is available at \\textit{https://github.com/yuewan2/biasedcot}.", "authors": ["Yue Wan", "Xiaowei Jia", "Xiang Lorraine Li"], "categories": ["cs.LG", "cs.AI"], "fields_of_study": ["Computer Science"], "published_date": "2025-06-14", "url": "https://arxiv.org/abs/2506.12301", "pdf_url": "https://arxiv.org/pdf/2506.12301v1", "arxiv_id": "2506.12301", "doi": "10.48550/arXiv.2506.12301", "citation_count": 10, "influential_citation_count": 0, "has_code": true, "code_url": "https://github.com/yuewan2/biasedcot}", "venue": "Annual Meeting of the Association for Computational Linguistics", "quality_score": 0.301}
{"id": "e50a269be053faa103856dff40701467bbadcd6f038964aed2fc4baa2470cf60", "sources": ["arxiv", "semantic_scholar"], "title": "Thought Graph Traversal for Test-time Scaling in Chest X-ray VLLMs", "abstract": "Test-time scaling offers a promising way to improve the reasoning performance of vision-language large models (VLLMs) without additional training. In this paper, we explore a simple but effective approach for applying test-time scaling to chest X-ray report generation. Specifically, we introduce a lightweight Thought Graph Traversal (TGT) framework that guides the model to reason through organ-specific findings in a medically coherent order. This framework integrates structured medical priors into the prompt, enabling deeper and more logical analysis with no changes to the underlying model. To further enhance reasoning depth, we apply a reasoning budget forcing strategy that adjusts the model's inference depth at test time by dynamically extending its generation process. This simple yet powerful combination allows a frozen radiology VLLM to self-correct and generate more accurate, consistent chest X-ray reports. Our method outperforms baseline prompting approaches on standard benchmarks, and also reveals dataset biases through traceable reasoning paths. Code and prompts are open-sourced for reproducibility at https://github.com/glerium/Thought-Graph-Traversal", "authors": ["Yue Yao", "Zelin Wen", "Yan Tong", "Xinyu Tian", "Xuqing Li", "Xiao Ma", "Dongliang Xu", "Tom Gedeon"], "categories": ["cs.CV"], "fields_of_study": ["Computer Science"], "published_date": "2025-06-13", "url": "https://arxiv.org/abs/2506.11989", "pdf_url": "https://arxiv.org/pdf/2506.11989v3", "arxiv_id": "2506.11989", "doi": "10.1016/j.patcog.2026.113639", "citation_count": 5, "influential_citation_count": 0, "has_code": true, "code_url": "https://github.com/glerium/Thought-Graph-Traversal", "venue": "Pattern Recognition", "quality_score": 0.2993}
{"id": "e35eea937f51cede7d82dce239cffd4c42461da6b1980784616be385a80fabec", "sources": ["arxiv", "semantic_scholar"], "title": "EasyARC: Evaluating Vision Language Models on True Visual Reasoning", "abstract": "Building on recent advances in language-based reasoning models, we explore multimodal reasoning that integrates vision and text. Existing multimodal benchmarks primarily test visual extraction combined with text-based reasoning, lacking true visual reasoning with more complex interactions between vision and language. Inspired by the ARC challenge, we introduce EasyARC, a vision-language benchmark requiring multi-image, multi-step reasoning, and self-correction. EasyARC is procedurally generated, fully verifiable, and scalable, making it ideal for reinforcement learning (RL) pipelines. The generators incorporate progressive difficulty levels, enabling structured evaluation across task types and complexities. We benchmark state-of-the-art vision-language models and analyze their failure modes. We argue that EasyARC sets a new standard for evaluating true reasoning and test-time scaling capabilities in vision-language models. We open-source our benchmark dataset and evaluation code.", "authors": ["Mert Unsal", "Aylin Akkus"], "categories": ["cs.CV", "cs.LG"], "fields_of_study": ["Computer Science"], "published_date": "2025-06-13", "url": "https://arxiv.org/abs/2506.11595", "pdf_url": "https://arxiv.org/pdf/2506.11595v1", "arxiv_id": "2506.11595", "doi": "10.48550/arXiv.2506.11595", "citation_count": 9, "influential_citation_count": 1, "has_code": true, "code_url": null, "venue": "arXiv.org", "quality_score": 0.2993}
{"id": "297616f31cbb2902d31fec7abce98954df8e5e5f13214d1cc519c279884d6917", "sources": ["arxiv", "semantic_scholar"], "title": "DART: Distilling Autoregressive Reasoning to Silent Thought", "abstract": "Chain-of-Thought (CoT) reasoning has significantly advanced Large Language Models (LLMs) in solving complex tasks. However, its autoregressive paradigm leads to significant computational overhead, hindering its deployment in latency-sensitive applications. To address this, we propose \\textbf{DART} (\\textbf{D}istilling \\textbf{A}utoregressive \\textbf{R}easoning to Silent \\textbf{T}hought), a self-distillation framework that enables LLMs to replace autoregressive CoT with non-autoregressive Silent Thought (ST). Specifically, DART introduces two training pathways: the CoT pathway for traditional reasoning and the ST pathway for generating answers directly from a few ST tokens. The ST pathway utilizes a lightweight Reasoning Evolvement Module (REM) to align its hidden states with the CoT pathway, enabling the ST tokens to evolve into informative embeddings. During inference, only the ST pathway is activated, leveraging evolving ST tokens to deliver the answer directly. Extensive experimental results demonstrate that DART offers significant performance gains compared with existing non-autoregressive baselines without extra inference latency, serving as a feasible alternative for efficient reasoning.", "authors": ["Nan Jiang", "Ziming Wu", "De-Chuan Zhan", "Fuming Lai", "Shaobing Lian"], "categories": ["cs.CL"], "fields_of_study": ["Computer Science"], "published_date": "2025-06-13", "url": "https://arxiv.org/abs/2506.11752", "pdf_url": "https://arxiv.org/pdf/2506.11752v2", "arxiv_id": "2506.11752", "doi": "10.48550/arXiv.2506.11752", "citation_count": 6, "influential_citation_count": 3, "has_code": false, "code_url": null, "venue": "Conference on Empirical Methods in Natural Language Processing", "quality_score": 0.301}
{"id": "d5a2ae774e380d7b7a7b9059c68129956fa782992ddf4e864aac7dc6231561b8", "sources": ["arxiv", "semantic_scholar"], "title": "Causal Sufficiency and Necessity Improves Chain-of-Thought Reasoning", "abstract": "Chain-of-Thought (CoT) prompting plays an indispensable role in endowing large language models (LLMs) with complex reasoning capabilities. However, CoT currently faces two fundamental challenges: (1) Sufficiency, which ensures that the generated intermediate inference steps comprehensively cover and substantiate the final conclusion; and (2) Necessity, which identifies the inference steps that are truly indispensable for the soundness of the resulting answer. We propose a causal framework that characterizes CoT reasoning through the dual lenses of sufficiency and necessity. Incorporating causal Probability of Sufficiency and Necessity allows us not only to determine which steps are logically sufficient or necessary to the prediction outcome, but also to quantify their actual influence on the final reasoning outcome under different intervention scenarios, thereby enabling the automated addition of missing steps and the pruning of redundant ones. Extensive experimental results on various mathematical and commonsense reasoning benchmarks confirm substantial improvements in reasoning efficiency and reduced token usage without sacrificing accuracy. Our work provides a promising direction for improving LLM reasoning performance and cost-effectiveness.", "authors": ["Xiangning Yu", "Zhuohan Wang", "Linyi Yang", "Haoxuan Li", "Anjie Liu", "Xiao Xue", "Jun Wang", "Mengyue Yang"], "categories": ["cs.CL", "cs.AI", "math.ST", "stat.ME"], "fields_of_study": ["Computer Science", "Mathematics"], "published_date": "2025-06-11", "url": "https://arxiv.org/abs/2506.09853", "pdf_url": "https://arxiv.org/pdf/2506.09853v3", "arxiv_id": "2506.09853", "doi": "10.48550/arXiv.2506.09853", "citation_count": 13, "influential_citation_count": 1, "has_code": false, "code_url": null, "venue": "arXiv.org", "quality_score": 0.2865}
{"id": "9a6813e06af8d4d4df326d675665ecf03c164c8fb705dd9bfe787e78803de474", "sources": ["arxiv", "semantic_scholar"], "title": "A Theory of Inference Compute Scaling: Reasoning through Directed Stochastic Skill Search", "abstract": "Large language models (LLMs) demand considerable computational, energy, and financial resources during both training and deployment. While scaling laws for training have guided much of the field's recent progress, inference costs now represent a significant and growing component of the overall resource burden, particularly for reasoning-focused models. Existing characterizations of compute-optimality that consider model size, dataset size, and inference tokens in isolation or in fixed combinations risk overlooking more efficient operating points. We introduce directed stochastic skill search (DS3), a general framework that represents inference as stochastic traversal over a learned skill graph. From a simplified yet expressive instantiation, we derive closed-form expressions for task success and compute cost across a wide range of inference strategies -- including chain-of-thought (CoT) and tree-of-thought (ToT) -- enabling comparative analysis as a function of task difficulty and model capability. To that end, we extend a prior first-principles tripartite graph framework of LLM training to incorporate inference, and separately bridge DS3 with empirical methods that characterize LLM scaling behavior. We theoretically recover empirically observed patterns, including: linear accuracy scaling with logarithmic compute; variation in preferred inference strategies as a function of task difficulty and model capability; emergent behavior elicited by reasoning even when performance plateaus under parameter scaling; and both best-of-N (BoN) and majority voting behavior captured within a unified analytical framework. By explicitly characterizing training-inference interdependencies, our framework deepens theoretical understanding and supports principled algorithmic design and resource allocation.", "authors": ["Austin R. Ellis-Mohr", "Anuj K. Nayak", "Lav R. Varshney"], "categories": ["cs.LG", "cs.AI", "cs.CY", "cs.PF"], "fields_of_study": ["Medicine", "Computer Science"], "published_date": "2025-06-10", "url": "https://arxiv.org/abs/2507.00004", "pdf_url": "https://arxiv.org/pdf/2507.00004v2", "arxiv_id": "2507.00004", "doi": "10.48550/arXiv.2507.00004", "citation_count": 3, "influential_citation_count": 0, "has_code": false, "code_url": null, "venue": null, "quality_score": 0.1505}
{"id": "d9105a2b7db8ee24e6e379f2edc2c9c72ea773aa4a3c6e67f330f2de2291bcd5", "sources": ["arxiv", "semantic_scholar"], "title": "Thinking vs. Doing: Agents that Reason by Scaling Test-Time Interaction", "abstract": "The current paradigm of test-time scaling relies on generating long reasoning traces (\"thinking\" more) before producing a response. In agent problems that require interaction, this can be done by generating thinking traces before acting in the world. However, this process does not allow agents to acquire new information from the environment or adapt their behavior over time. In this work, we propose to scale test-time interaction, an untapped dimension of test-time scaling that increases the agent's interaction horizon to enable running rich behaviors such as exploration, backtracking, and dynamic re-planning within a single rollout. To demonstrate the promise of this scaling dimension, we study the domain of web agents. We first show that even prompting-based interaction scaling without any training can improve task success on web benchmarks non-trivially. Building on this, we introduce TTI (Test-Time Interaction), a curriculum-based online reinforcement learning (RL) approach that trains agents by adaptively adjusting their rollout lengths. Using a Gemma 3 12B model, TTI produces state-of-the-art open-source, open-data web agents on WebVoyager and WebArena benchmarks. We further show that TTI enables agents to balance exploration and exploitation adaptively. Our results establish interaction scaling as a powerful, complementary axis to scaling per-step compute, offering new avenues for training adaptive agents.", "authors": ["Junhong Shen", "Hao Bai", "Lunjun Zhang", "Yifei Zhou", "Amrith Setlur", "Shengbang Tong", "Diego Caples", "Nan Jiang", "Tong Zhang", "Ameet Talwalkar", "Aviral Kumar"], "categories": ["cs.LG", "cs.AI"], "fields_of_study": ["Computer Science"], "published_date": "2025-06-09", "url": "https://arxiv.org/abs/2506.07976", "pdf_url": "https://arxiv.org/pdf/2506.07976v2", "arxiv_id": "2506.07976", "doi": "10.48550/arXiv.2506.07976", "citation_count": 34, "influential_citation_count": 2, "has_code": true, "code_url": null, "venue": "arXiv.org", "quality_score": 0.386}
{"id": "4424117ebb6aa31bfbe6049397ae8fd0ad4700ccadfe7a52b07ae0eb11ec34de", "sources": ["arxiv", "semantic_scholar"], "title": "Temporalizing Confidence: Evaluation of Chain-of-Thought Reasoning with Signal Temporal Logic", "abstract": "Large Language Models (LLMs) have shown impressive performance in mathematical reasoning tasks when guided by Chain-of-Thought (CoT) prompting. However, they tend to produce highly confident yet incorrect outputs, which poses significant risks in domains like education, where users may lack the expertise to assess reasoning steps. To address this, we propose a structured framework that models stepwise confidence as a temporal signal and evaluates it using Signal Temporal Logic (STL). In particular, we define formal STL-based constraints to capture desirable temporal properties and compute robustness scores that serve as structured, interpretable confidence estimates. Our approach also introduces a set of uncertainty reshaping strategies to enforce smoothness, monotonicity, and causal consistency across the reasoning trajectory. Experiments show that our approach consistently improves calibration metrics and provides more reliable uncertainty estimates than conventional confidence aggregation and post-hoc calibration.", "authors": ["Zhenjiang Mao", "Artem Bisliouk", "Rohith Reddy Nama", "Ivan Ruchkin"], "categories": ["cs.LG", "cs.AI"], "fields_of_study": ["Computer Science"], "published_date": "2025-06-09", "url": "https://arxiv.org/abs/2506.08243", "pdf_url": "https://arxiv.org/pdf/2506.08243v1", "arxiv_id": "2506.08243", "doi": "10.48550/arXiv.2506.08243", "citation_count": 9, "influential_citation_count": 0, "has_code": false, "code_url": null, "venue": "Workshop on Innovative Use of NLP for Building Educational Applications", "quality_score": 0.25}
{"id": "d8ef199644ac8edfa91b31da024db82fdecc3a2bbf7d67ad738ac90a15c4b566", "sources": ["arxiv", "semantic_scholar"], "title": "Saffron-1: Safety Inference Scaling", "abstract": "Existing safety assurance research has primarily focused on training-phase alignment to instill safe behaviors into LLMs. However, recent studies have exposed these methods' susceptibility to diverse jailbreak attacks. Concurrently, inference scaling has significantly advanced LLM reasoning capabilities but remains unexplored in the context of safety assurance. Addressing this gap, our work pioneers inference scaling for robust and effective LLM safety against emerging threats. We reveal that conventional inference scaling techniques, despite their success in reasoning tasks, perform poorly in safety contexts, even falling short of basic approaches like Best-of-N Sampling. We attribute this inefficiency to a newly identified challenge, the exploration--efficiency dilemma, arising from the high computational overhead associated with frequent process reward model (PRM) evaluations. To overcome this dilemma, we propose SAFFRON, a novel inference scaling paradigm tailored explicitly for safety assurance. Central to our approach is the introduction of a multifurcation reward model (MRM) that significantly reduces the required number of reward model evaluations. To operationalize this paradigm, we further propose: (i) a partial supervision training objective for MRM, (ii) a conservative exploration constraint to prevent out-of-distribution explorations, and (iii) a Trie-based key--value caching strategy that facilitates cache sharing across sequences during tree search. Extensive experiments validate the effectiveness of our method. Additionally, we publicly release our trained multifurcation reward model (Saffron-1) and the accompanying token-level safety reward dataset (Safety4M) to accelerate future research in LLM safety. Our code, model, and data are publicly available at https://github.com/q-rz/saffron , and our project homepage is at https://q-rz.github.io/p/saffron .", "authors": ["Ruizhong Qiu", "Gaotang Li", "Tianxin Wei", "Jingrui He", "Hanghang Tong"], "categories": ["cs.LG", "cs.AI", "cs.CR"], "fields_of_study": ["Computer Science"], "published_date": "2025-06-06", "url": "https://arxiv.org/abs/2506.06444", "pdf_url": "https://arxiv.org/pdf/2506.06444v2", "arxiv_id": "2506.06444", "doi": null, "citation_count": 0, "influential_citation_count": 0, "has_code": true, "code_url": "https://github.com/q-rz/saffron", "venue": null, "quality_score": 0.2194}
{"id": "36da977f574e35453d1eaf35ac930d7be2bc8d010b92ddbcd62f751b91cdddb1", "sources": ["arxiv", "semantic_scholar"], "title": "ScaleRTL: Scaling LLMs with Reasoning Data and Test-Time Compute for Accurate RTL Code Generation", "abstract": "Recent advances in large language models (LLMs) have enabled near-human performance on software coding benchmarks, but their effectiveness in RTL code generation remains limited due to the scarcity of high-quality training data. While prior efforts have fine-tuned LLMs for RTL tasks, they do not fundamentally overcome the data bottleneck and lack support for test-time scaling due to their non-reasoning nature. In this work, we introduce ScaleRTL, the first reasoning LLM for RTL coding that scales up both high-quality reasoning data and test-time compute. Specifically, we curate a diverse set of long chain-of-thought reasoning traces averaging 56K tokens each, resulting in a dataset of 3.5B tokens that captures rich RTL knowledge. Fine-tuning a general-purpose reasoning model on this corpus yields ScaleRTL that is capable of deep RTL reasoning. Subsequently, we further enhance the performance of ScaleRTL through a novel test-time scaling strategy that extends the reasoning process via iteratively reflecting on and self-correcting previous reasoning steps. Experimental results show that ScaleRTL achieves state-of-the-art performance on VerilogEval and RTLLM, outperforming 18 competitive baselines by up to 18.4% on VerilogEval and 12.7% on RTLLM.", "authors": ["Chenhui Deng", "Yun-Da Tsai", "Guan-Ting Liu", "Zhongzhi Yu", "Haoxing Ren"], "categories": ["cs.AR", "cs.AI"], "fields_of_study": ["Computer Science"], "published_date": "2025-06-05", "url": "https://arxiv.org/abs/2506.05566", "pdf_url": "https://arxiv.org/pdf/2506.05566v2", "arxiv_id": "2506.05566", "doi": "10.1109/MLCAD65511.2025.11189212", "citation_count": 24, "influential_citation_count": 3, "has_code": false, "code_url": null, "venue": "Workshop on Machine Learning for CAD", "quality_score": 0.3495}
{"id": "fc12903d542ba38ab160953fb75812e8cedb54d05b019b6ff01bc223c6eb7a12", "sources": ["arxiv", "semantic_scholar"], "title": "Revisiting Test-Time Scaling: A Survey and a Diversity-Aware Method for Efficient Reasoning", "abstract": "Test-Time Scaling (TTS) improves the reasoning performance of Large Language Models (LLMs) by allocating additional compute during inference. We conduct a structured survey of TTS methods and categorize them into sampling-based, search-based, and trajectory optimization strategies. We observe that reasoning-optimized models often produce less diverse outputs, which limits TTS effectiveness. To address this, we propose ADAPT (A Diversity Aware Prefix fine-Tuning), a lightweight method that applies prefix tuning with a diversity-focused data strategy. Experiments on mathematical reasoning tasks show that ADAPT reaches 80% accuracy using eight times less compute than strong baselines. Our findings highlight the essential role of generative diversity in maximizing TTS effectiveness.", "authors": ["Ho-Lam Chung", "Teng-Yun Hsiao", "Hsiao-Ying Huang", "Chunerh Cho", "Jian-Ren Lin", "Zhang Ziwei", "Yun-Nung Chen"], "categories": ["cs.CL"], "fields_of_study": ["Computer Science"], "published_date": "2025-06-05", "url": "https://arxiv.org/abs/2506.04611", "pdf_url": "https://arxiv.org/pdf/2506.04611v1", "arxiv_id": "2506.04611", "doi": "10.48550/arXiv.2506.04611", "citation_count": 5, "influential_citation_count": 1, "has_code": false, "code_url": null, "venue": "arXiv.org", "quality_score": 0.1945}
{"id": "f300968d241bf8ee30fa9a40d210dd3e8188607c2ce034ef5fcbaf232e268c95", "sources": ["arxiv", "semantic_scholar"], "title": "Reason-to-Recommend: Using Interaction-of-Thought Reasoning to Enhance LLM Recommendation", "abstract": "Driven by advances in Large Language Models (LLMs), integrating them into recommendation tasks has gained interest due to their strong semantic understanding and prompt flexibility. Prior work encoded user-item interactions or metadata into prompts for recommendations. In parallel, LLM reasoning, boosted by test-time scaling and reinforcement learning, has excelled in fields like mathematics and code, where reasoning traces and correctness signals are clear, enabling high performance and interpretability. However, directly applying these reasoning methods to recommendation is ineffective because user feedback is implicit and lacks reasoning supervision. To address this, we propose $\\textbf{R2Rec}$, a reasoning-enhanced recommendation framework that samples interaction chains from the user-item graph and converts them into structured interaction-of-thoughts via a progressive masked prompting strategy, with each thought representing stepwise reasoning grounded in interaction context. This allows LLMs to simulate step-by-step decision-making based on implicit patterns. We design a two-stage training pipeline: supervised fine-tuning teaches basic reasoning from high-quality traces, and reinforcement learning refines reasoning via reward signals, alleviating sparse explicit supervision. Experiments on three real-world datasets show R2Rec outperforms classical and LLM-based baselines with an average $\\textbf{10.48%}$ improvement in HitRatio@1 and $\\textbf{131.81%}$ gain over the original LLM. Furthermore, the explicit reasoning chains enhance interpretability by revealing the decision process. Our code is available at: https://anonymous.4open.science/r/R2Rec-7C5D.", "authors": ["Keyu Zhao", "Fengli Xu", "Yong Li"], "categories": ["cs.IR", "cs.AI"], "fields_of_study": ["Computer Science"], "published_date": "2025-06-05", "url": "https://arxiv.org/abs/2506.05069", "pdf_url": "https://arxiv.org/pdf/2506.05069v2", "arxiv_id": "2506.05069", "doi": "10.48550/arXiv.2506.05069", "citation_count": 18, "influential_citation_count": 0, "has_code": true, "code_url": null, "venue": "arXiv.org", "quality_score": 0.3197}
{"id": "d3e23781871ec09d0ff6572742955b90ae28dfce0fd61901742a120eb8b4101f", "sources": ["arxiv", "semantic_scholar"], "title": "Dissecting Long-Chain-of-Thought Reasoning Models: An Empirical Study", "abstract": "Despite recent progress in training long-chain-of-thought reasoning models via scaling reinforcement learning (RL), its underlying training dynamics remain poorly understood, and several counterintuitive behaviors persist. This work focuses on three key aspects: (1) We systematically analyze the roles of positive and negative samples in scaling RL, revealing that positive samples mainly facilitate precise fitting to the training data, whereas negative samples significantly enhance generalization and robustness. Interestingly, while positive samples are essential for convergence in the zero-RL setting, training on negative samples alone suffices to attain strong reasoning performance and even better generalization in cold-start scenarios. (2) We identify substantial data inefficiency in group relative policy optimization, where over half of the samples yield zero advantage. To address this, we explore two strategies, including relative length rewards and offline sample injection, to leverage these data better and enhance reasoning efficiency and capability. (3) We investigate unstable performance across various reasoning models and benchmarks, attributing instability to uncertain problems with ambiguous outcomes, and demonstrate that greedy decoding can distort evaluation by flipping the correctness of responses. Our code is available at: https://github.com/takagi97/Dissect-Long-Reason-Models.", "authors": ["Yongyu Mu", "Jiali Zeng", "Bei Li", "Xinyan Guan", "Fandong Meng", "Jie Zhou", "Tong Xiao", "Jingbo Zhu"], "categories": ["cs.LG", "cs.CL"], "fields_of_study": ["Computer Science"], "published_date": "2025-06-05", "url": "https://arxiv.org/abs/2506.04913", "pdf_url": "https://arxiv.org/pdf/2506.04913v2", "arxiv_id": "2506.04913", "doi": null, "citation_count": 2, "influential_citation_count": 0, "has_code": true, "code_url": "https://github.com/takagi97/Dissect-Long-Reason-Models", "venue": null, "quality_score": 0.218}
{"id": "d7c6b9299beaaffe573271262ffc834d3842ff5ad142918d416cc49574345f71", "sources": ["arxiv", "semantic_scholar"], "title": "Reason from Future: Reverse Thought Chain Enhances LLM Reasoning", "abstract": "It has been demonstrated that carefully designed reasoning paradigms, like Chain-of-Thought (CoT) and Tree-of-Thought (ToT), can enhance the reasoning capabilities of small language models by detailed thinking and extensive thought searching, unbounded branching factors in the searching space create prohibitive reasoning consumption. However these methods fall into the trap of local optimum reasoning, which means the model lacks a global perspective while solving problems. We propose a novel reasoning paradigm called Reason from Future (RFF), which generates reasoning paths by bidirectional reasoning that combines top-down planning with bottom-up reasoning accumulation. The essence of RFF lies in its reverse reasoning mechanism, which prioritizes core logical relationships and imposes goal-oriented constraints on intermediate steps, thereby reducing the searching space and mitigating error accumulation inherent in sequential forward reasoning. Empirical evaluations across diverse experiments demonstrate that RFF outperforms conventional paradigms with higher accuracy and less searching space to solve complex tasks.", "authors": ["Yinlong Xu", "Yanzhao Zheng", "Shuoshuo Sun", "Shuaihan Huang", "Baohua Dong", "Hangcheng Zhu", "Ruohui Huang", "Gang Yu", "Hongxia Xu", "Jian Wu"], "categories": ["cs.AI"], "fields_of_study": ["Computer Science"], "published_date": "2025-06-04", "url": "https://arxiv.org/abs/2506.03673", "pdf_url": "https://arxiv.org/pdf/2506.03673v1", "arxiv_id": "2506.03673", "doi": "10.48550/arXiv.2506.03673", "citation_count": 5, "influential_citation_count": 0, "has_code": false, "code_url": null, "venue": "Annual Meeting of the Association for Computational Linguistics", "quality_score": 0.1945}
{"id": "52af86fb662de021d026ce1d49116b599b93cbc4529c7323d9a557663f89e353", "sources": ["arxiv", "semantic_scholar"], "title": "Does Thinking More always Help? Mirage of Test-Time Scaling in Reasoning Models", "abstract": "Recent trends in test-time scaling for reasoning models (e.g., OpenAI o1, DeepSeek R1) have led to a popular belief that extending thinking traces using prompts like \"Wait\" or \"Let me rethink\" can improve performance. This raises a natural question: Does thinking more at test-time truly lead to better reasoning? To answer this question, we perform a detailed empirical study across models and benchmarks, which reveals a consistent pattern of initial performance improvements from additional thinking followed by a decline, due to \"overthinking\". To understand this non-monotonic trend, we consider a simple probabilistic model, which reveals that additional thinking increases output variance-creating an illusion of improved reasoning while ultimately undermining precision. Thus, observed gains from \"more thinking\" are not true indicators of improved reasoning, but artifacts stemming from the connection between model uncertainty and evaluation metric. This suggests that test-time scaling through extended thinking is not an effective way to utilize the inference thinking budget. Recognizing these limitations, we introduce an alternative test-time scaling approach, parallel thinking, inspired by Best-of-N sampling. Our method generates multiple independent reasoning paths within the same inference budget and selects the most consistent response via majority vote, achieving up to 20% higher accuracy compared to extended thinking. This provides a simple yet effective mechanism for test-time scaling of reasoning models.", "authors": ["Soumya Suvra Ghosal", "Souradip Chakraborty", "Avinash Reddy", "Yifu Lu", "Mengdi Wang", "Dinesh Manocha", "Furong Huang", "Mohammad Ghavamzadeh", "Amrit Singh Bedi"], "categories": ["cs.AI", "cs.CL"], "fields_of_study": ["Computer Science"], "published_date": "2025-06-04", "url": "https://arxiv.org/abs/2506.04210", "pdf_url": "https://arxiv.org/pdf/2506.04210v3", "arxiv_id": "2506.04210", "doi": null, "citation_count": 29, "influential_citation_count": 2, "has_code": false, "code_url": null, "venue": null, "quality_score": 0.3693}
{"id": "0b9608956ee8a8bb4c93a625c28f0da44ccce61a7e845acec5b491d1aa97357d", "sources": ["arxiv", "semantic_scholar"], "title": "Enhancing Large Language Models with Neurosymbolic Reasoning for Multilingual Tasks", "abstract": "Large language models (LLMs) often struggle to perform multi-target reasoning in long-context scenarios where relevant information is scattered across extensive documents. To address this challenge, we introduce NeuroSymbolic Augmented Reasoning (NSAR), which combines the benefits of neural and symbolic reasoning during inference. NSAR explicitly extracts symbolic facts from text and generates executable Python code to handle complex reasoning steps. Through extensive experiments across seven languages and diverse context lengths, we demonstrate that NSAR significantly outperforms both a vanilla RAG baseline and advanced prompting strategies in accurately identifying and synthesizing multiple pieces of information. Our results highlight the effectiveness of combining explicit symbolic operations with neural inference for robust, interpretable, and scalable reasoning in multilingual settings.", "authors": ["Sina Bagheri Nezhad", "Ameeta Agrawal"], "categories": ["cs.CL"], "fields_of_study": ["Computer Science"], "published_date": "2025-06-03", "url": "https://arxiv.org/abs/2506.02483", "pdf_url": "https://arxiv.org/pdf/2506.02483v1", "arxiv_id": "2506.02483", "doi": "10.48550/arXiv.2506.02483", "citation_count": 6, "influential_citation_count": 0, "has_code": false, "code_url": null, "venue": "International Workshop on Neural-Symbolic Learning and Reasoning", "quality_score": 0.2113}
{"id": "9c9e28dcb7f78518cdebb299afa335bdb5565367ff00dbc28b386463e393be97", "sources": ["arxiv", "semantic_scholar"], "title": "GeoChain: Multimodal Chain-of-Thought for Geographic Reasoning", "abstract": "This paper introduces GeoChain, a large-scale benchmark for evaluating step-by-step geographic reasoning in multimodal large language models (MLLMs). Leveraging 1.46 million Mapillary street-level images, GeoChain pairs each image with a 21-step chain-of-thought (CoT) question sequence (over 30 million Q&A pairs). These sequences guide models from coarse attributes to fine-grained localization across four reasoning categories - visual, spatial, cultural, and precise geolocation - annotated by difficulty. Images are also enriched with semantic segmentation (150 classes) and a visual locatability score. Our benchmarking of contemporary MLLMs (GPT-4.1 variants, Claude 3.7, Gemini 2.5 variants) on a diverse 2,088-image subset reveals consistent challenges: models frequently exhibit weaknesses in visual grounding, display erratic reasoning, and struggle to achieve accurate localization, especially as the reasoning complexity escalates. GeoChain offers a robust diagnostic methodology, critical for fostering significant advancements in complex geographic reasoning within MLLMs.", "authors": ["Sahiti Yerramilli", "Nilay Pande", "Rynaa Grover", "Jayant Sravan Tamarapalli"], "categories": ["cs.AI", "cs.CV", "cs.LG"], "fields_of_study": ["Computer Science"], "published_date": "2025-06-01", "url": "https://arxiv.org/abs/2506.00785", "pdf_url": "https://arxiv.org/pdf/2506.00785v3", "arxiv_id": "2506.00785", "doi": "10.48550/arXiv.2506.00785", "citation_count": 21, "influential_citation_count": 1, "has_code": false, "code_url": null, "venue": "Conference on Empirical Methods in Natural Language Processing", "quality_score": 0.3356}
{"id": "028ab388ad0641f66266658cb0540a934c0601430f620ed4e8560fd5d768281c", "sources": ["arxiv", "semantic_scholar"], "title": "SCOUT: Teaching Pre-trained Language Models to Enhance Reasoning via Flow Chain-of-Thought", "abstract": "Chain of Thought (CoT) prompting improves the reasoning performance of large language models (LLMs) by encouraging step by step thinking. However, CoT-based methods depend on intermediate reasoning steps, which limits scalability and generalization. Recent work explores recursive reasoning, where LLMs reuse internal layers across iterations to refine latent representations without explicit CoT supervision. While promising, these approaches often require costly pretraining and lack a principled framework for how reasoning should evolve across iterations. We address this gap by introducing Flow Chain of Thought (Flow CoT), a reasoning paradigm that models recursive inference as a progressive trajectory of latent cognitive states. Flow CoT frames each iteration as a distinct cognitive stage deepening reasoning across iterations without relying on manual supervision. To realize this, we propose SCOUT (Stepwise Cognitive Optimization Using Teachers), a lightweight fine tuning framework that enables Flow CoT style reasoning without the need for pretraining. SCOUT uses progressive distillation to align each iteration with a teacher of appropriate capacity, and a cross attention based retrospective module that integrates outputs from previous iterations while preserving the models original computation flow. Experiments across eight reasoning benchmarks show that SCOUT consistently improves both accuracy and explanation quality, achieving up to 1.8% gains under fine tuning. Qualitative analyses further reveal that SCOUT enables progressively deeper reasoning across iterations refining both belief formation and explanation granularity. These results not only validate the effectiveness of SCOUT, but also demonstrate the practical viability of Flow CoT as a scalable framework for enhancing reasoning in LLMs.", "authors": ["Guanghao Li", "Wenhao Jiang", "Mingfeng Chen", "Yan Li", "Hao Yu", "Shuting Dong", "Tao Ren", "Ming Tang", "Chun Yuan"], "categories": ["cs.AI"], "fields_of_study": ["Computer Science"], "published_date": "2025-05-30", "url": "https://arxiv.org/abs/2505.24181", "pdf_url": "https://arxiv.org/pdf/2505.24181v1", "arxiv_id": "2505.24181", "doi": "10.48550/arXiv.2505.24181", "citation_count": 4, "influential_citation_count": 0, "has_code": false, "code_url": null, "venue": "arXiv.org", "quality_score": 0.1776}
{"id": "e735009457cb01c4d607f84b1a91a5c4e939f20062a31e548ca7cf244f35863a", "sources": ["arxiv", "semantic_scholar"], "title": "REASONING GYM: Reasoning Environments for Reinforcement Learning with Verifiable Rewards", "abstract": "We introduce Reasoning Gym (RG), a library of reasoning environments for reinforcement learning with verifiable rewards. It provides over 100 data generators and verifiers spanning multiple domains including algebra, arithmetic, computation, cognition, geometry, graph theory, logic, and various common games. Its key innovation is the ability to generate virtually infinite training data with adjustable complexity, unlike most previous reasoning datasets, which are typically fixed. This procedural generation approach allows for continuous evaluation across varying difficulty levels. Our experimental results demonstrate the efficacy of RG in both evaluating and reinforcement learning of reasoning models.", "authors": ["Zafir Stojanovski", "Oliver Stanley", "Joe Sharratt", "Richard Jones", "Abdulhakeem Adefioye", "Jean Kaddour", "Andreas Köpf"], "categories": ["cs.LG", "cs.AI", "cs.CL"], "fields_of_study": ["Computer Science"], "published_date": "2025-05-30", "url": "https://arxiv.org/abs/2505.24760", "pdf_url": "https://arxiv.org/pdf/2505.24760v2", "arxiv_id": "2505.24760", "doi": "10.48550/arXiv.2505.24760", "citation_count": 88, "influential_citation_count": 21, "has_code": true, "code_url": "https://github.com/open-thought/reasoning-gym", "venue": "arXiv.org", "quality_score": 0.6712}
{"id": "66cc1b145a91c6d26e6bf95f717da596df5cc9550088adc4356c6a0670e18bdf", "sources": ["arxiv", "semantic_scholar"], "title": "A*-Thought: Efficient Reasoning via Bidirectional Compression for Low-Resource Settings", "abstract": "Large Reasoning Models (LRMs) achieve superior performance by extending the thought length. However, a lengthy thinking trajectory leads to reduced efficiency. Most of the existing methods are stuck in the assumption of overthinking and attempt to reason efficiently by compressing the Chain-of-Thought, but this often leads to performance degradation. To address this problem, we introduce A*-Thought, an efficient tree search-based unified framework designed to identify and isolate the most essential thoughts from the extensive reasoning chains produced by these models. It formulates the reasoning process of LRMs as a search tree, where each node represents a reasoning span in the giant reasoning space. By combining the A* search algorithm with a cost function specific to the reasoning path, it can efficiently compress the chain of thought and determine a reasoning path with high information density and low cost. In addition, we also propose a bidirectional importance estimation mechanism, which further refines this search process and enhances its efficiency beyond uniform sampling. Extensive experiments on several advanced math tasks show that A*-Thought effectively balances performance and efficiency over a huge search space. Specifically, A*-Thought can improve the performance of QwQ-32B by 2.39$\\times$ with low-budget and reduce the length of the output token by nearly 50% with high-budget. The proposed method is also compatible with several other LRMs, demonstrating its generalization capability. The code can be accessed at: https://github.com/AI9Stars/AStar-Thought.", "authors": ["Xiaoang Xu", "Shuo Wang", "Xu Han", "Zhenghao Liu", "Huijia Wu", "Peipei Li", "Zhiyuan Liu", "Maosong Sun", "Zhaofeng He"], "categories": ["cs.CL"], "fields_of_study": ["Computer Science"], "published_date": "2025-05-30", "url": "https://arxiv.org/abs/2505.24550", "pdf_url": "https://arxiv.org/pdf/2505.24550v2", "arxiv_id": "2505.24550", "doi": "10.48550/arXiv.2505.24550", "citation_count": 7, "influential_citation_count": 0, "has_code": true, "code_url": "https://github.com/AI9Stars/AStar-Thought", "venue": "arXiv.org", "quality_score": 0.2745}
{"id": "9c9a4e27b80d1f2ca58b3821f52bd8658550cebfaa21add73ebc4dce72650170", "sources": ["arxiv", "semantic_scholar"], "title": "Inference-time Scaling of Diffusion Models through Classical Search", "abstract": "Classical search algorithms have long underpinned modern artificial intelligence. In this work, we tackle the challenge of inference-time control in diffusion models -- adapting generated outputs to meet diverse test-time objectives -- using principles from classical search. We propose a general framework that orchestrates local and global search to efficiently navigate the generative space. It employs a theoretically grounded local search via annealed Langevin MCMC and performs compute-efficient global exploration using breadth-first and depth-first tree search. We evaluate our approach on a range of challenging domains, including planning, offline reinforcement learning, and image generation. Across all tasks, we observe significant gains in both performance and efficiency. These results show that classical search provides a principled and practical foundation for inference-time scaling in diffusion models. Project page at https://diffusion-inference-scaling.github.io/.", "authors": ["Xiangcheng Zhang", "Haowei Lin", "Haotian Ye", "James Zou", "Jianzhu Ma", "Yitao Liang", "Yilun Du"], "categories": ["cs.LG", "stat.ML"], "fields_of_study": ["Computer Science", "Mathematics"], "published_date": "2025-05-29", "url": "https://arxiv.org/abs/2505.23614", "pdf_url": "https://arxiv.org/pdf/2505.23614v2", "arxiv_id": "2505.23614", "doi": "10.48550/arXiv.2505.23614", "citation_count": 46, "influential_citation_count": 3, "has_code": false, "code_url": null, "venue": "arXiv.org", "quality_score": 0.418}
{"id": "18a79a2582cb530033041a5e199c037d9e4f8a76e2710adaa6a13bf209043bd9", "sources": ["arxiv", "semantic_scholar"], "title": "Continuous Chain of Thought Enables Parallel Exploration and Reasoning", "abstract": "Modern language models generate chain-of-thought traces by autoregressively sampling tokens from a finite vocabulary. While this discrete sampling has achieved remarkable success, conducting chain-of-thought with continuously-valued tokens (CoT2) offers a richer and more expressive alternative. Our work provides new theoretical guarantees and algorithms for CoT2, motivated by logical reasoning tasks that inherently require search capabilities. Theoretically, we establish how CoT2 facilitates the model to track multiple discrete traces in parallel; and quantify the level of achievable parallelism and its benefits for inference efficiency. We also provide a CoT2-based one-layer transformer construction that solves the combinatorial \"subset sum problem\" given a sufficient embedding dimension. These insights arise from a novel and effective supervision strategy where we match the language model outputs to the empirical token distributions of a set of target traces. Complementing this, we introduce sampling strategies that unlock policy optimization methods for CoT2. Our primary strategy samples and composes $K$ discrete tokens at each decoding step to control the level of parallelism. Experiments confirm that (i) the optimal level of parallelism is governed by the embedding dimension, (ii) our continuous supervision strategy can outperform alternative methods, and (iii) policy optimization with CoT2 indeed improves the performance of the model beyond its initial discrete or continuous supervision.", "authors": ["Halil Alperen Gozeten", "M. Emrullah Ildiz", "Xuechen Zhang", "Hrayr Harutyunyan", "Ankit Singh Rawat", "Samet Oymak"], "categories": ["cs.LG"], "fields_of_study": ["Computer Science"], "published_date": "2025-05-29", "url": "https://arxiv.org/abs/2505.23648", "pdf_url": "https://arxiv.org/pdf/2505.23648v3", "arxiv_id": "2505.23648", "doi": "10.48550/arXiv.2505.23648", "citation_count": 26, "influential_citation_count": 3, "has_code": false, "code_url": null, "venue": "arXiv.org", "quality_score": 0.3578}
{"id": "31d7c7bcfe428496f911318f56a7ed3f84d96046ab8ed46736db9733769356da", "sources": ["arxiv", "semantic_scholar"], "title": "Table-R1: Inference-Time Scaling for Table Reasoning", "abstract": "In this work, we present the first study to explore inference-time scaling on table reasoning tasks. We develop and evaluate two post-training strategies to enable inference-time scaling: distillation from frontier model reasoning traces and reinforcement learning with verifiable rewards (RLVR). For distillation, we introduce a large-scale dataset of reasoning traces generated by DeepSeek-R1, which we use to fine-tune LLMs into the Table-R1-SFT model. For RLVR, we propose task-specific verifiable reward functions and apply the GRPO algorithm to obtain the Table-R1-Zero model. We evaluate our Table-R1-series models across diverse table reasoning tasks, including short-form QA, fact verification, and free-form QA. Notably, the Table-R1-Zero model matches or exceeds the performance of GPT-4.1 and DeepSeek-R1, while using only a 7B-parameter LLM. It also demonstrates strong generalization to out-of-domain datasets. Extensive ablation and qualitative analyses reveal the benefits of instruction tuning, model architecture choices, and cross-task generalization, as well as emergence of essential table reasoning skills during RL training.", "authors": ["Zheyuan Yang", "Lyuhao Chen", "Arman Cohan", "Yilun Zhao"], "categories": ["cs.CL"], "fields_of_study": ["Computer Science"], "published_date": "2025-05-29", "url": "https://arxiv.org/abs/2505.23621", "pdf_url": "https://arxiv.org/pdf/2505.23621v2", "arxiv_id": "2505.23621", "doi": "10.48550/arXiv.2505.23621", "citation_count": 19, "influential_citation_count": 1, "has_code": false, "code_url": null, "venue": "Conference on Empirical Methods in Natural Language Processing", "quality_score": 0.3253}
{"id": "3283a0c8be227defc515253036ef9441aeaa695e7ad29cc6007ed486bd93728b", "sources": ["arxiv", "semantic_scholar"], "title": "Argus: Vision-Centric Reasoning with Grounded Chain-of-Thought", "abstract": "Recent advances in multimodal large language models (MLLMs) have demonstrated remarkable capabilities in vision-language tasks, yet they often struggle with vision-centric scenarios where precise visual focus is needed for accurate reasoning. In this paper, we introduce Argus to address these limitations with a new visual attention grounding mechanism. Our approach employs object-centric grounding as visual chain-of-thought signals, enabling more effective goal-conditioned visual attention during multimodal reasoning tasks. Evaluations on diverse benchmarks demonstrate that Argus excels in both multimodal reasoning tasks and referring object grounding tasks. Extensive analysis further validates various design choices of Argus, and reveals the effectiveness of explicit language-guided visual region-of-interest engagement in MLLMs, highlighting the importance of advancing multimodal intelligence from a visual-centric perspective. Project page: https://yunzeman.github.io/argus/", "authors": ["Yunze Man", "De-An Huang", "Guilin Liu", "Shiwei Sheng", "Shilong Liu", "Liang-Yan Gui", "Jan Kautz", "Yu-Xiong Wang", "Zhiding Yu"], "categories": ["cs.CV"], "fields_of_study": ["Computer Science"], "published_date": "2025-05-29", "url": "https://arxiv.org/abs/2505.23766", "pdf_url": "https://arxiv.org/pdf/2505.23766v1", "arxiv_id": "2505.23766", "doi": "10.1109/CVPR52734.2025.01331", "citation_count": 42, "influential_citation_count": 3, "has_code": false, "code_url": null, "venue": "Computer Vision and Pattern Recognition", "quality_score": 0.4084}
{"id": "3776f5ba916f3985980e989391b4c5497d71eea25ec803374241d6c63eca6910", "sources": ["arxiv", "semantic_scholar"], "title": "What Makes a Good Reasoning Chain? Uncovering Structural Patterns in Long Chain-of-Thought Reasoning", "abstract": "Recent advances in reasoning with large language models (LLMs) have popularized Long Chain-of-Thought (LCoT), a strategy that encourages deliberate and step-by-step reasoning before producing a final answer. While LCoTs have enabled expert-level performance in complex tasks, how the internal structures of their reasoning chains drive, or even predict, the correctness of final answers remains a critical yet underexplored question. In this work, we present LCoT2Tree, an automated framework that converts sequential LCoTs into hierarchical tree structures and thus enables deeper structural analysis of LLM reasoning. Using graph neural networks (GNNs), we reveal that structural patterns extracted by LCoT2Tree, including exploration, backtracking, and verification, serve as stronger predictors of final performance across a wide range of tasks and models. Leveraging an explainability technique, we further identify critical thought patterns such as over-branching that account for failures. Beyond diagnostic insights, the structural patterns by LCoT2Tree support practical applications, including improving Best-of-N decoding effectiveness. Overall, our results underscore the critical role of internal structures of reasoning chains, positioning LCoT2Tree as a powerful tool for diagnosing, interpreting, and improving reasoning in LLMs.", "authors": ["Gangwei Jiang", "Yahui Liu", "Zhaoyi Li", "Qi Wang", "Fuzheng Zhang", "Linqi Song", "Ying Wei", "Defu Lian"], "categories": ["cs.AI"], "fields_of_study": ["Computer Science"], "published_date": "2025-05-28", "url": "https://arxiv.org/abs/2505.22148", "pdf_url": "https://arxiv.org/pdf/2505.22148v1", "arxiv_id": "2505.22148", "doi": "10.48550/arXiv.2505.22148", "citation_count": 26, "influential_citation_count": 3, "has_code": false, "code_url": null, "venue": "Conference on Empirical Methods in Natural Language Processing", "quality_score": 0.3578}
{"id": "648186da23f0f3271191b8e3e99f8bc63e4e7e9ba9cae200b9888506c3330ff4", "sources": ["arxiv", "semantic_scholar"], "title": "On Learning Verifiers and Implications to Chain-of-Thought Reasoning", "abstract": "Chain-of-Thought reasoning has emerged as a powerful approach for solving complex mathematical and logical problems. However, it can often veer off track through incorrect or unsubstantiated inferences. Formal mathematical reasoning, which can be checked with a formal verifier, is one approach to addressing this issue. However, currently LLMs are simply not good enough to solve complex problems in a formal way, and even just formalizing an informal problem statement can be challenging. Motivated by this fact, in this work we consider the problem of learning reliable verifiers for natural language Chain-of-Thought reasoning. That is, given a problem statement and step-by-step solution in natural language, the aim of the verifier is to output [Yes] if the reasoning steps in the solution are all valid, and [No] otherwise. In this work we give a formal PAC-learning framework for studying this problem. We propose and analyze several natural verification goals, at different levels of strength, in this framework. We provide sample complexity upper-bounds for learning verifiers satisfying these goals, as well as lower-bound and impossibility results for learning other natural verification objectives without additional assumptions.", "authors": ["Maria-Florina Balcan", "Avrim Blum", "Zhiyuan Li", "Dravyansh Sharma"], "categories": ["cs.LG"], "fields_of_study": ["Computer Science"], "published_date": "2025-05-28", "url": "https://arxiv.org/abs/2505.22650", "pdf_url": "https://arxiv.org/pdf/2505.22650v2", "arxiv_id": "2505.22650", "doi": null, "citation_count": 5, "influential_citation_count": 0, "has_code": false, "code_url": null, "venue": null, "quality_score": 0.1945}
{"id": "94568d0d754a1897e67539084fb653b216c7855401187537fa2bf2421a0e31ba", "sources": ["arxiv", "semantic_scholar"], "title": "Let Me Think! A Long Chain-of-Thought Can Be Worth Exponentially Many Short Ones", "abstract": "Inference-time computation has emerged as a promising scaling axis for improving large language model reasoning. However, despite yielding impressive performance, the optimal allocation of inference-time computation remains poorly understood. A central question is whether to prioritize sequential scaling (e.g., longer chains of thought) or parallel scaling (e.g., majority voting across multiple short chains of thought). In this work, we seek to illuminate the landscape of test-time scaling by demonstrating the existence of reasoning settings where sequential scaling offers an exponential advantage over parallel scaling. These settings are based on graph connectivity problems in challenging distributions of graphs. We validate our theoretical findings with comprehensive experiments across a range of language models, including models trained from scratch for graph connectivity with different chain of thought strategies as well as large reasoning models.", "authors": ["Parsa Mirtaheri", "Ezra Edelman", "Samy Jelassi", "Eran Malach", "Enric Boix-Adsera"], "categories": ["cs.LG", "cs.AI", "cs.CL"], "fields_of_study": ["Computer Science"], "published_date": "2025-05-27", "url": "https://arxiv.org/abs/2505.21825", "pdf_url": "https://arxiv.org/pdf/2505.21825v2", "arxiv_id": "2505.21825", "doi": "10.48550/arXiv.2505.21825", "citation_count": 7, "influential_citation_count": 1, "has_code": false, "code_url": null, "venue": "arXiv.org", "quality_score": 0.2258}
{"id": "f854dec708cd8009c14a992ec60313488980e6e40f67ebc3d4daa5deae75651a", "sources": ["arxiv", "semantic_scholar"], "title": "Leveraging Large Language Models for Bengali Math Word Problem Solving with Chain of Thought Reasoning", "abstract": "Solving Bengali Math Word Problems (MWPs) remains a major challenge in natural language processing (NLP) due to the language's low-resource status and the multi-step reasoning required. Existing models struggle with complex Bengali MWPs, largely because no human-annotated Bengali dataset has previously addressed this task. This gap has limited progress in Bengali mathematical reasoning. To address this, we created SOMADHAN, a dataset of 8792 complex Bengali MWPs with manually written, step-by-step solutions. We designed this dataset to support reasoning-focused evaluation and model development in a linguistically underrepresented context. Using SOMADHAN, we evaluated a range of large language models (LLMs) - including GPT-4o, GPT-3.5 Turbo, LLaMA series models, Deepseek, and Qwen - through both zero-shot and few-shot prompting with and without Chain of Thought (CoT) reasoning. CoT prompting consistently improved performance over standard prompting, especially in tasks requiring multi-step logic. LLaMA-3.3 70B achieved the highest accuracy of 88% with few-shot CoT prompting. We also applied Low-Rank Adaptation (LoRA) to fine-tune models efficiently, enabling them to adapt to Bengali MWPs with minimal computational cost. Our work fills a critical gap in Bengali NLP by providing a high-quality reasoning dataset and a scalable framework for solving complex MWPs. We aim to advance equitable research in low-resource languages and enhance reasoning capabilities in educational and language technologies.", "authors": ["Bidyarthi Paul", "Jalisha Jashim Era", "Mirazur Rahman Zim", "Tahmid Sattar Aothoi", "Faisal Muhammad Shah"], "categories": ["cs.CL", "cs.LG"], "fields_of_study": ["Computer Science"], "published_date": "2025-05-27", "url": "https://arxiv.org/abs/2505.21354", "pdf_url": "https://arxiv.org/pdf/2505.21354v2", "arxiv_id": "2505.21354", "doi": "10.48550/arXiv.2505.21354", "citation_count": 1, "influential_citation_count": 1, "has_code": false, "code_url": null, "venue": "arXiv.org", "quality_score": 0.1742}
{"id": "ca620655d9c3d4c60f4e68b9fbba172d82f35bbd91276e9e1dbc825608650650", "sources": ["arxiv", "semantic_scholar"], "title": "Efficient Reasoning via Chain of Unconscious Thought", "abstract": "Large Reasoning Models (LRMs) achieve promising performance but compromise token efficiency due to verbose reasoning processes. Unconscious Thought Theory (UTT) posits that complex problems can be solved more efficiently through internalized cognitive processes. Inspired by UTT, we propose a new reasoning paradigm, termed Chain of Unconscious Thought (CoUT), to improve the token efficiency of LRMs by guiding them to mimic human unconscious thought and internalize reasoning processes. Concretely, we first prompt the model to internalize the reasoning by thinking in the hidden layer. Then, we design a bag of token-efficient strategies to further help models reduce unnecessary tokens yet preserve the performance. Our work reveals that models may possess beneficial unconscious thought, enabling improved efficiency without sacrificing performance. Extensive experiments demonstrate the effectiveness of CoUT. Remarkably, it surpasses CoT by reducing token usage by 47.62% while maintaining comparable accuracy, as shown in Figure 1. The code of CoUT is available at this link: https://github.com/Rohan-GRH/CoUT", "authors": ["Ruihan Gong", "Yue Liu", "Wenjie Qu", "Mingzhe Du", "Yufei He", "Yingwei Ma", "Yulin Chen", "Xiang Liu", "Yi Wen", "Xinfeng Li", "Ruidong Wang", "Xinzhong Zhu", "Bryan Hooi", "Jiaheng Zhang"], "categories": ["cs.CL"], "fields_of_study": ["Computer Science"], "published_date": "2025-05-26", "url": "https://arxiv.org/abs/2505.19756", "pdf_url": "https://arxiv.org/pdf/2505.19756v1", "arxiv_id": "2505.19756", "doi": "10.48550/arXiv.2505.19756", "citation_count": 7, "influential_citation_count": 1, "has_code": true, "code_url": "https://github.com/Rohan-GRH/CoUT", "venue": "arXiv.org", "quality_score": 0.2674}
{"id": "7920f6bba51609f3b8915614dd36fccaf547ca136a5b3be7fb728436d5b2006f", "sources": ["arxiv", "semantic_scholar"], "title": "Scaling over Scaling: Exploring Test-Time Scaling Plateau in Large Reasoning Models", "abstract": "Large reasoning models (LRMs) have exhibited the capacity of enhancing reasoning performance via internal test-time scaling. Building upon this, a promising direction is to further scale test-time compute to unlock even greater reasoning capabilities. However, as we push these scaling boundaries, systematically understanding the practical limits and achieving optimal resource allocation becomes a critical challenge. In this paper, we investigate the scaling plateau of test-time scaling and introduce the Test-Time Scaling Performance Model (TTSPM). We theoretically analyze two fundamental paradigms for such extended scaling, parallel scaling and sequential scaling, from a probabilistic modeling perspective. Our primary contribution is the derivation of the saturation point on the scaling budget for both strategies, identifying thresholds beyond which additional computation yields diminishing returns. Remarkably, despite their distinct mechanisms, both paradigms converge to a unified mathematical structure in their upper bounds. We empirically validate our theoretical findings on challenging reasoning benchmarks, including AIME, MATH-500, and GPQA, demonstrating the practical utility of these bounds for test-time resource allocation. We hope that this work provides insights into the cost-benefit trade-offs of test-time scaling, guiding the development of more resource-efficient inference strategies for large reasoning models.", "authors": ["Jian Wang", "Boyan Zhu", "Chak Tou Leong", "Yongqi Li", "Wenjie Li"], "categories": ["cs.AI", "cs.CL", "cs.LG"], "fields_of_study": ["Computer Science"], "published_date": "2025-05-26", "url": "https://arxiv.org/abs/2505.20522", "pdf_url": "https://arxiv.org/pdf/2505.20522v2", "arxiv_id": "2505.20522", "doi": "10.48550/arXiv.2505.20522", "citation_count": 7, "influential_citation_count": 0, "has_code": false, "code_url": null, "venue": "arXiv.org", "quality_score": 0.2258}
{"id": "81f822ed99c2550db10c94fc634cdfeadda4791e8b6d2986b4613b9ea8c9fa98", "sources": ["arxiv", "semantic_scholar"], "title": "Large Language Models' Reasoning Stalls: An Investigation into the Capabilities of Frontier Models", "abstract": "Empirical methods to examine the capability of Large Language Models (LLMs) to use Automated Theorem Prover (ATP) reasoning strategies are studied. We evaluate the performance of State of the Art models from December 2023 and August 2024 on PRONTOQA steamroller reasoning problems. For that, we develop methods for assessing LLM response accuracy and correct answer correlation. Our results show that progress in improving LLM reasoning abilities has stalled over the nine month period. By tracking completion tokens, we show that almost all improvement in reasoning ability since GPT-4 was released can be attributed to either hidden system prompts or the training of models to automatically use generic Chain of Thought prompting strategies. Among the ATP reasoning strategies tried, we found that current frontier LLMs are best able to follow the bottom-up (also known as forward-chaining) strategy. A low positive correlation was found between an LLM response containing correct reasoning and arriving at the correct conclusion.", "authors": ["Lachlan McGinness", "Peter Baumgartner"], "categories": ["cs.AI"], "fields_of_study": ["Computer Science"], "published_date": "2025-05-26", "url": "https://arxiv.org/abs/2505.19676", "pdf_url": "https://arxiv.org/pdf/2505.19676v3", "arxiv_id": "2505.19676", "doi": "10.48550/arXiv.2505.19676", "citation_count": 1, "influential_citation_count": 0, "has_code": false, "code_url": null, "venue": "arXiv.org", "quality_score": 0.173}
{"id": "614bc24b04387db3ea76f2f43045cc05660298c44a2242a1be8771823b9d2734", "sources": ["arxiv", "semantic_scholar"], "title": "WebCoT: Enhancing Web Agent Reasoning by Reconstructing Chain-of-Thought in Reflection, Branching, and Rollback", "abstract": "Web agents powered by Large Language Models (LLMs) show promise for next-generation AI, but their limited reasoning in uncertain, dynamic web environments hinders robust deployment. In this paper, we identify key reasoning skills essential for effective web agents, i.e., reflection & lookahead, branching, and rollback, and curate trajectory data that exemplifies these abilities by reconstructing the agent's (inference-time) reasoning algorithms into chain-of-thought rationales. We conduct experiments in the agent self-improving benchmark, OpenWebVoyager, and demonstrate that distilling salient reasoning patterns into the backbone LLM via simple fine-tuning can substantially enhance its performance. Our approach yields significant improvements across multiple benchmarks, including WebVoyager, Mind2web-live, and SimpleQA (web search), highlighting the potential of targeted reasoning skill enhancement for web agents.", "authors": ["Minda Hu", "Tianqing Fang", "Jianshu Zhang", "Junyu Ma", "Zhisong Zhang", "Jingyan Zhou", "Hongming Zhang", "Haitao Mi", "Dong Yu", "Irwin King"], "categories": ["cs.CL"], "fields_of_study": ["Computer Science"], "published_date": "2025-05-26", "url": "https://arxiv.org/abs/2505.20013", "pdf_url": "https://arxiv.org/pdf/2505.20013v2", "arxiv_id": "2505.20013", "doi": "10.48550/arXiv.2505.20013", "citation_count": 12, "influential_citation_count": 0, "has_code": false, "code_url": null, "venue": "Conference on Empirical Methods in Natural Language Processing", "quality_score": 0.2785}
{"id": "d1a3a3701dd15785b82d9b73a878edb7303d593d64fd8fef7282efa93a1e126d", "sources": ["arxiv", "semantic_scholar"], "title": "LIMOPro: Reasoning Refinement for Efficient and Effective Test-time Scaling", "abstract": "Large language models (LLMs) have demonstrated remarkable reasoning capabilities through test-time scaling approaches, particularly when fine-tuned with chain-of-thought (CoT) data distilled from more powerful large reasoning models (LRMs). However, these reasoning chains often contain verbose elements that mirror human problem-solving, categorized as progressive reasoning (the essential solution development path) and functional elements (verification processes, alternative solution approaches, and error corrections). While progressive reasoning is crucial, the functional elements significantly increase computational demands during test-time inference. We introduce PIR (Perplexity-based Importance Refinement), a principled framework that quantitatively evaluates the importance of each reasoning step based on its impact on answer prediction confidence. PIR systematically identifies and selectively prunes only low-importance functional steps while preserving progressive reasoning components, creating optimized training data that maintains the integrity of the core solution path while reducing verbosity. Models fine-tuned on PIR-optimized data exhibit superior test-time scaling properties, generating more concise reasoning chains while achieving improved accuracy (+0.9\\% to +6.6\\%) with significantly reduced token usage (-3\\% to -41\\%) across challenging reasoning benchmarks (AIME, AMC, and GPQA Diamond). Our approach demonstrates strong generalizability across different model sizes, data sources, and token budgets, offering a practical solution for deploying reasoning-capable LLMs in scenarios where efficient test-time scaling, response time, and computational efficiency are valuable constraints.", "authors": ["Yang Xiao", "Jiashuo Wang", "Ruifeng Yuan", "Chunpu Xu", "Kaishuai Xu", "Wenjie Li", "Pengfei Liu"], "categories": ["cs.CL", "cs.AI"], "fields_of_study": ["Computer Science"], "published_date": "2025-05-25", "url": "https://arxiv.org/abs/2505.19187", "pdf_url": "https://arxiv.org/pdf/2505.19187v2", "arxiv_id": "2505.19187", "doi": "10.48550/arXiv.2505.19187", "citation_count": 15, "influential_citation_count": 0, "has_code": false, "code_url": null, "venue": "arXiv.org", "quality_score": 0.301}
{"id": "f537c78568513278ecbbf6326e47dbdfd3e3588f50d773771c3df0b125cb5a20", "sources": ["arxiv", "semantic_scholar"], "title": "LogicCat: A Chain-of-Thought Text-to-SQL Benchmark for Complex Reasoning", "abstract": "Text-to-SQL is a critical task in natural language processing that aims to transform natural language questions into accurate and executable SQL queries. In real-world scenarios, these reasoning tasks are often accompanied by complex mathematical computations, domain knowledge, and hypothetical reasoning scenarios. However, existing large-scale Text-to-SQL datasets typically focus on business logic and task logic, neglecting critical factors such as vertical domain knowledge, complex mathematical reasoning, and hypothetical reasoning, which are essential for realistically reflecting the reasoning demands in practical applications and completing data querying and analysis. To bridge this gap, we introduce LogicCat, the first Text-to-SQL benchmark dataset specifically designed for complex reasoning and chain-of-thought parsing, encompassing physics, arithmetic, commonsense, and hypothetical reasoning scenarios. LogicCat comprises 4,038 English questions paired 12,114 detailed chain-of-thought reasoning steps, spanning 45 databases across diverse domains, significantly surpassing existing datasets in complexity. Experimental results demonstrate that LogicCat substantially increases the task difficulty for current state-of-the-art models to at most 33.20% execution accuracy, indicating that this task remains exceptionally challenging. The advancement of LogicCat represents a crucial step toward developing systems suitable for real-world enterprise data analysis and autonomous query generation. We have released our dataset code at https://github.com/Ffunkytao/LogicCat.", "authors": ["Tao Liu", "Xutao Mao", "Hongying Zan", "Dixuan Zhang", "Yifan Li", "Haixin Liu", "Lulu Kong", "Jiaming Hou", "Rui Li", "YunLong Li", "aoze zheng", "Zhiqiang Zhang", "Luo Zhewei", "Kunli Zhang", "Min Peng"], "categories": ["cs.CL"], "fields_of_study": ["Computer Science"], "published_date": "2025-05-24", "url": "https://arxiv.org/abs/2505.18744", "pdf_url": "https://arxiv.org/pdf/2505.18744v3", "arxiv_id": "2505.18744", "doi": "10.1609/aaai.v40i36.40243", "citation_count": 3, "influential_citation_count": 1, "has_code": true, "code_url": "https://github.com/Ffunkytao/LogicCat", "venue": "AAAI Conference on Artificial Intelligence", "quality_score": 0.2639}
{"id": "933dbf1f62279ad9da6b0ce3b13ba99084f2937251d3823345993f493ee2fa7f", "sources": ["arxiv", "semantic_scholar"], "title": "Thought calibration: Efficient and confident test-time scaling", "abstract": "Reasoning large language models achieve impressive test-time scaling by thinking for longer, but this performance gain comes at significant compute cost. Directly limiting test-time budget hurts overall performance, but not all problems are equally difficult. We propose thought calibration to decide dynamically when thinking can be terminated. To calibrate our decision rule, we view a language model's growing body of thoughts as a nested sequence of reasoning trees, where the goal is to identify the point at which novel reasoning plateaus. We realize this framework through lightweight probes that operate on top of the language model's hidden representations, which are informative of both the reasoning structure and overall consistency of response. Based on three reasoning language models and four datasets, thought calibration preserves model performance with up to a 60% reduction in thinking tokens on in-distribution data, and up to 20% in out-of-distribution data.", "authors": ["Menghua Wu", "Cai Zhou", "Stephen Bates", "Tommi Jaakkola"], "categories": ["cs.LG", "cs.AI"], "fields_of_study": ["Computer Science"], "published_date": "2025-05-23", "url": "https://arxiv.org/abs/2505.18404", "pdf_url": "https://arxiv.org/pdf/2505.18404v1", "arxiv_id": "2505.18404", "doi": "10.48550/arXiv.2505.18404", "citation_count": 13, "influential_citation_count": 2, "has_code": false, "code_url": null, "venue": "Conference on Empirical Methods in Natural Language Processing", "quality_score": 0.2865}
{"id": "30f2870e3a04a1eeb867d769ad5f3cd993c55855ed68cc753dc1aaa7455f2e2c", "sources": ["arxiv", "semantic_scholar"], "title": "Stepwise Reasoning Checkpoint Analysis: A Test Time Scaling Method to Enhance LLMs' Reasoning", "abstract": "Mathematical reasoning through Chain-of-Thought (CoT) has emerged as a powerful capability of Large Language Models (LLMs), which can be further enhanced through Test-Time Scaling (TTS) methods like Beam Search and DVTS. However, these methods, despite improving accuracy by allocating more computational resources during inference, often suffer from path homogenization and inefficient use of intermediate results. To address these limitations, we propose Stepwise Reasoning Checkpoint Analysis (SRCA), a framework that introduces checkpoints between reasoning steps. It incorporates two key strategies: (1) Answer-Clustered Search, which groups reasoning paths by their intermediate checkpoint answers to maintain diversity while ensuring quality, and (2) Checkpoint Candidate Augmentation, which leverages all intermediate answers for final decision-making. Our approach effectively reduces path homogenization and creates a fault-tolerant mechanism by utilizing high-quality intermediate results. Experimental results show that SRCA improves reasoning accuracy compared to existing TTS methods across various mathematical datasets.", "authors": ["Zezhong Wang", "Xingshan Zeng", "Weiwen Liu", "Yufei Wang", "Liangyou Li", "Yasheng Wang", "Lifeng Shang", "Xin Jiang", "Qun Liu", "Kam-Fai Wong"], "categories": ["cs.CL"], "fields_of_study": ["Computer Science"], "published_date": "2025-05-23", "url": "https://arxiv.org/abs/2505.17829", "pdf_url": "https://arxiv.org/pdf/2505.17829v1", "arxiv_id": "2505.17829", "doi": "10.48550/arXiv.2505.17829", "citation_count": 1, "influential_citation_count": 0, "has_code": false, "code_url": null, "venue": "Conference on Empirical Methods in Natural Language Processing", "quality_score": 0.1696}
{"id": "26e50770e3c9de0701961ff7887dad7203e2663406665dc33847cd297bfc1bd7", "sources": ["arxiv", "semantic_scholar"], "title": "Reward Model Generalization for Compute-Aware Test-Time Reasoning", "abstract": "External test-time reasoning enhances large language models (LLMs) by decoupling generation and selection. At inference time, the model generates multiple reasoning paths, and an auxiliary process reward model (PRM) is used to score and select the best one. A central challenge in this setting is test-time compute optimality (TCO), i.e., how to maximize answer accuracy under a fixed inference budget. In this work, we establish a theoretical framework to analyze how the generalization error of the PRM affects compute efficiency and reasoning performance. Leveraging PAC-Bayes theory, we derive generalization bounds and show that a lower generalization error of PRM leads to fewer samples required to find correct answers. Motivated by this analysis, we propose Compute-Aware Tree Search (CATS), an actor-critic framework that dynamically controls search behavior. The actor outputs sampling hyperparameters based on reward distributions and sparsity statistics, while the critic estimates their utility to guide budget allocation. Experiments on the MATH and AIME benchmarks with various LLMs and PRMs demonstrate that CATS consistently outperforms other external TTS methods, validating our theoretical predictions.", "authors": ["Zeen Song", "Wenwen Qiang", "Siyu Zhao", "Changwen Zheng", "Gang Hua"], "categories": ["cs.LG"], "fields_of_study": ["Computer Science"], "published_date": "2025-05-23", "url": "https://arxiv.org/abs/2505.18065", "pdf_url": "https://arxiv.org/pdf/2505.18065v1", "arxiv_id": "2505.18065", "doi": "10.48550/arXiv.2505.18065", "citation_count": 0, "influential_citation_count": 0, "has_code": false, "code_url": null, "venue": "arXiv.org", "quality_score": 0.1696}
{"id": "f5e3e63c1b335b970ae5888db5b2519aeabd2cd092d9a0e966bef3e7249bddd7", "sources": ["arxiv", "semantic_scholar"], "title": "Don't Overthink it. Preferring Shorter Thinking Chains for Improved LLM Reasoning", "abstract": "Reasoning large language models (LLMs) heavily rely on scaling test-time compute to perform complex reasoning tasks by generating extensive \"thinking\" chains. While demonstrating impressive results, this approach incurs significant computational costs and inference time. In this work, we challenge the assumption that long thinking chains results in better reasoning capabilities. We first demonstrate that shorter reasoning chains within individual questions are significantly more likely to yield correct answers - up to 34.5% more accurate than the longest chain sampled for the same question. Based on these results, we suggest short-m@k, a novel reasoning LLM inference method. Our method executes k independent generations in parallel and halts computation once the first m thinking processes are done. The final answer is chosen using majority voting among these m chains. Basic short-1@k demonstrates similar or even superior performance over standard majority voting in low-compute settings - using up to 40% fewer thinking tokens. short-3@k, while slightly less efficient than short-1@k, consistently surpasses majority voting across all compute budgets, while still being substantially faster (up to 33% wall time reduction). To further validate our findings, we finetune LLMs using short, long, and randomly selected reasoning chains. We then observe that training on the shorter ones leads to better performance. Our findings suggest rethinking current methods of test-time compute in reasoning LLMs, emphasizing that longer \"thinking\" does not necessarily translate to improved performance and can, counter-intuitively, lead to degraded results.", "authors": ["Michael Hassid", "Gabriel Synnaeve", "Yossi Adi", "Roy Schwartz"], "categories": ["cs.CL", "cs.AI"], "fields_of_study": ["Computer Science"], "published_date": "2025-05-23", "url": "https://arxiv.org/abs/2505.17813", "pdf_url": "https://arxiv.org/pdf/2505.17813v2", "arxiv_id": "2505.17813", "doi": "10.48550/arXiv.2505.17813", "citation_count": 65, "influential_citation_count": 4, "has_code": false, "code_url": null, "venue": "arXiv.org", "quality_score": 0.4549}
{"id": "cee449e68b59bbecea1adc178156878f57944ea1dbab872a91241eb52be9a11f", "sources": ["arxiv", "semantic_scholar"], "title": "Does Chain-of-Thought Reasoning Really Reduce Harmfulness from Jailbreaking?", "abstract": "Jailbreak attacks have been observed to largely fail against recent reasoning models enhanced by Chain-of-Thought (CoT) reasoning. However, the underlying mechanism remains underexplored, and relying solely on reasoning capacity may raise security concerns. In this paper, we try to answer the question: Does CoT reasoning really reduce harmfulness from jailbreaking? Through rigorous theoretical analysis, we demonstrate that CoT reasoning has dual effects on jailbreaking harmfulness. Based on the theoretical insights, we propose a novel jailbreak method, FicDetail, whose practical performance validates our theoretical findings.", "authors": ["Chengda Lu", "Xiaoyu Fan", "Yu Huang", "Rongwu Xu", "Jijie Li", "Wei Xu"], "categories": ["cs.AI"], "fields_of_study": ["Computer Science"], "published_date": "2025-05-23", "url": "https://arxiv.org/abs/2505.17650", "pdf_url": "https://arxiv.org/pdf/2505.17650v1", "arxiv_id": "2505.17650", "doi": "10.48550/arXiv.2505.17650", "citation_count": 12, "influential_citation_count": 2, "has_code": false, "code_url": null, "venue": "Annual Meeting of the Association for Computational Linguistics", "quality_score": 0.2785}
{"id": "88b65ed7d6f8920dee26e7344353aadec5213b67606b1381750683fbd86a656a", "sources": ["arxiv", "semantic_scholar"], "title": "DEL-ToM: Inference-Time Scaling for Theory-of-Mind Reasoning via Dynamic Epistemic Logic", "abstract": "Theory-of-Mind (ToM) tasks pose a unique challenge for large language models (LLMs), which often lack the capability for dynamic logical reasoning. In this work, we propose DEL-ToM, a framework that improves verifiable ToM reasoning through inference-time scaling rather than architectural changes. Our approach decomposes ToM tasks into a sequence of belief updates grounded in Dynamic Epistemic Logic (DEL), enabling structured and verifiable dynamic logical reasoning. We use data generated automatically via a DEL simulator to train a verifier, which we call the Process Belief Model (PBM), to score each belief update step. During inference, the PBM evaluates candidate belief traces from the LLM and selects the highest-scoring one. This allows LLMs to allocate extra inference-time compute to yield more transparent reasoning. Experiments across model scales and benchmarks show that DEL-ToM consistently improves performance, demonstrating that verifiable belief supervision significantly enhances LLMs' ToM capabilities without retraining. Code is available at https://github.com/joel-wu/DEL-ToM.", "authors": ["Yuheng Wu", "Jianwen Xie", "Denghui Zhang", "Zhaozhuo Xu"], "categories": ["cs.AI", "cs.CL"], "fields_of_study": ["Computer Science"], "published_date": "2025-05-22", "url": "https://arxiv.org/abs/2505.17348", "pdf_url": "https://arxiv.org/pdf/2505.17348v2", "arxiv_id": "2505.17348", "doi": "10.48550/arXiv.2505.17348", "citation_count": 2, "influential_citation_count": 0, "has_code": true, "code_url": "https://github.com/joel-wu/DEL-ToM", "venue": "Conference on Empirical Methods in Natural Language Processing", "quality_score": 0.2603}
{"id": "349069dfb052ad9a73abfe799a300d9cf9d721101e3ac7e65008d678890d0a28", "sources": ["arxiv", "semantic_scholar"], "title": "Reasoning Beyond Language: A Comprehensive Survey on Latent Chain-of-Thought Reasoning", "abstract": "Large Language Models (LLMs) have shown impressive performance on complex tasks through Chain-of-Thought (CoT) reasoning. However, conventional CoT relies on explicitly verbalized intermediate steps, which constrains its broader applicability, particularly in abstract reasoning tasks beyond language. To address this, there has been growing research interest in \\textit{latent CoT reasoning}, where the reasoning process is embedded within latent spaces. By decoupling reasoning from explicit language generation, latent CoT offers the promise of richer cognitive representations and facilitates more flexible, faster inference. This paper aims to present a comprehensive overview of this emerging paradigm and establish a systematic taxonomy. We analyze recent advances in methods, categorizing them from token-wise horizontal approaches to layer-wise vertical strategies. We then provide in-depth discussions of these methods, highlighting their design principles, applications, and remaining challenges. We hope that our survey provides a structured foundation for advancing this promising direction in LLM reasoning. The relevant papers will be regularly updated at https://github.com/EIT-NLP/Awesome-Latent-CoT.", "authors": ["Xinghao Chen", "Anhao Zhao", "Heming Xia", "Xuan Lu", "Hanlin Wang", "Yanjun Chen", "Wei Zhang", "Jian Wang", "Wenjie Li", "Xiaoyu Shen"], "categories": ["cs.CL"], "fields_of_study": ["Computer Science"], "published_date": "2025-05-22", "url": "https://arxiv.org/abs/2505.16782", "pdf_url": "https://arxiv.org/pdf/2505.16782v2", "arxiv_id": "2505.16782", "doi": "10.48550/arXiv.2505.16782", "citation_count": 46, "influential_citation_count": 3, "has_code": true, "code_url": "https://github.com/EIT-NLP/Awesome-Latent-CoT", "venue": "arXiv.org", "quality_score": 0.418}
{"id": "6e87c9da1591596d2686e0a338c203f2aa40f0d7f06d98a806482cb35652347e", "sources": ["arxiv", "semantic_scholar"], "title": "IRONIC: Coherence-Aware Reasoning Chains for Multi-Modal Sarcasm Detection", "abstract": "Interpreting figurative language such as sarcasm across multi-modal inputs presents unique challenges, often requiring task-specific fine-tuning and extensive reasoning steps. However, current Chain-of-Thought approaches do not efficiently leverage the same cognitive processes that enable humans to identify sarcasm. We present IRONIC, an in-context learning framework that leverages Multi-modal Coherence Relations to analyze referential, analogical and pragmatic image-text linkages. Our experiments show that IRONIC achieves state-of-the-art performance on zero-shot Multi-modal Sarcasm Detection across different baselines. This demonstrates the need for incorporating linguistic and cognitive insights into the design of multi-modal reasoning strategies. Our code is available at: https://github.com/aashish2000/IRONIC", "authors": ["Aashish Anantha Ramakrishnan", "Aadarsh Anantha Ramakrishnan", "Dongwon Lee"], "categories": ["cs.CL", "cs.AI", "cs.CV"], "fields_of_study": ["Computer Science"], "published_date": "2025-05-22", "url": "https://arxiv.org/abs/2505.16258", "pdf_url": "https://arxiv.org/pdf/2505.16258v2", "arxiv_id": "2505.16258", "doi": "10.48550/arXiv.2505.16258", "citation_count": 5, "influential_citation_count": 0, "has_code": true, "code_url": "https://github.com/aashish2000/IRONIC", "venue": "arXiv.org", "quality_score": 0.2603}
{"id": "424556496c6ef18cd0f9453b9ee22e5c4215d5dd14f34916edd4a030f5fbe55f", "sources": ["arxiv", "semantic_scholar"], "title": "Multilingual Test-Time Scaling via Initial Thought Transfer", "abstract": "Test-time scaling has emerged as a widely adopted inference-time strategy for boosting reasoning performance. However, its effectiveness has been studied almost exclusively in English, leaving its behavior in other languages largely unexplored. We present the first systematic study of test-time scaling in multilingual settings, evaluating DeepSeek-R1-Distill-LLama-8B and DeepSeek-R1-Distill-Qwen-7B across both high- and low-resource Latin-script languages. Our findings reveal that the relative gains from test-time scaling vary significantly across languages. Additionally, models frequently switch to English mid-reasoning, even when operating under strictly monolingual prompts. We further show that low-resource languages not only produce initial reasoning thoughts that differ significantly from English but also have lower internal consistency across generations in their early reasoning. Building on our findings, we introduce MITT (Multilingual Initial Thought Transfer), an unsupervised and lightweight reasoning prefix-tuning approach that transfers high-resource reasoning prefixes to enhance test-time scaling across all languages, addressing inconsistencies in multilingual reasoning performance. MITT significantly boosts DeepSeek-R1-Distill-Qwen-7B's reasoning performance, especially for underrepresented languages.", "authors": ["Prasoon Bajpai", "Tanmoy Chakraborty"], "categories": ["cs.CL"], "fields_of_study": ["Computer Science"], "published_date": "2025-05-21", "url": "https://arxiv.org/abs/2505.15508", "pdf_url": "https://arxiv.org/pdf/2505.15508v1", "arxiv_id": "2505.15508", "doi": "10.48550/arXiv.2505.15508", "citation_count": 4, "influential_citation_count": 0, "has_code": false, "code_url": null, "venue": "arXiv.org", "quality_score": 0.1747}
{"id": "75535b563e3c3a2966a8c034231665a309fd700f0db646fd28bdbc09e9c752e2", "sources": ["arxiv", "semantic_scholar"], "title": "SSR: Speculative Parallel Scaling Reasoning in Test-time", "abstract": "Large language models (LLMs) have achieved impressive results on multi-step mathematical reasoning, yet at the cost of high computational overhead. This challenge is particularly acute for test-time scaling methods such as parallel decoding, which increase answer diversity but scale poorly in efficiency. To address this efficiency-accuracy trade-off, we propose SSR (Speculative Parallel Scaling Reasoning), a training-free framework that leverages a key insight: by introducing speculative decoding at the step level, we can accelerate reasoning without sacrificing correctness. SSR integrates two components: a Selective Parallel Module (SPM) that identifies a small set of promising reasoning strategies via model-internal scoring, and Step-level Speculative Decoding (SSD), which enables efficient draft-target collaboration for fine-grained reasoning acceleration. Experiments on three mathematical benchmarks-AIME 2024, MATH-500, and LiveMathBench - demonstrate that SSR achieves strong gains over baselines. For instance, on LiveMathBench, SSR improves pass@1 accuracy by 13.84% while reducing computation to 80.5% of the baseline FLOPs. On MATH-500, SSR reduces compute to only 30% with no loss in accuracy.", "authors": ["Yuanlin Chu", "Bo Wang", "Xiang Liu", "Hong Chen", "Aiwei Liu", "Xuming Hu"], "categories": ["cs.LG"], "fields_of_study": ["Computer Science"], "published_date": "2025-05-21", "url": "https://arxiv.org/abs/2505.15340", "pdf_url": "https://arxiv.org/pdf/2505.15340v2", "arxiv_id": "2505.15340", "doi": "10.48550/arXiv.2505.15340", "citation_count": 2, "influential_citation_count": 0, "has_code": false, "code_url": null, "venue": "arXiv.org", "quality_score": 0.1673}
{"id": "5cc0833beb46adf2633f285516afbac30c4ff6814620ad9fcc3bfce469305344", "sources": ["arxiv", "semantic_scholar"], "title": "Adaptive Chain-of-Focus Reasoning via Dynamic Visual Search and Zooming for Efficient VLMs", "abstract": "Vision language models (VLMs) have achieved impressive performance across a variety of computer vision tasks. However, the multimodal reasoning capability has not been fully explored in existing models. In this paper, we propose a Chain-of-Focus (CoF) method that allows VLMs to perform adaptive focusing and zooming in on key image regions based on obtained visual cues and the given questions, achieving efficient multimodal reasoning. To enable this CoF capability, we present a two-stage training pipeline, including supervised fine-tuning (SFT) and reinforcement learning (RL). In the SFT stage, we construct the MM-CoF dataset, comprising 3K samples derived from a visual agent designed to adaptively identify key regions to solve visual tasks with different image resolutions and questions. We use MM-CoF to fine-tune the Qwen2.5-VL model for cold start. In the RL stage, we leverage the outcome accuracies and formats as rewards to update the Qwen2.5-VL model, enabling further refining the search and reasoning strategy of models without human priors. Our model achieves significant improvements on multiple benchmarks. On the V* benchmark that requires strong visual reasoning capability, our model outperforms existing VLMs by 5% among 8 image resolutions ranging from 224 to 4K, demonstrating the effectiveness of the proposed CoF method and facilitating the more efficient deployment of VLMs in practical applications.", "authors": ["Xintong Zhang", "Zhi Gao", "Bofei Zhang", "Pengxiang Li", "Xiaowen Zhang", "Yang Liu", "Tao Yuan", "Yuwei Wu", "Yunde Jia", "Song-Chun Zhu", "Qing Li"], "categories": ["cs.CV"], "fields_of_study": ["Computer Science"], "published_date": "2025-05-21", "url": "https://arxiv.org/abs/2505.15436", "pdf_url": "https://arxiv.org/pdf/2505.15436v3", "arxiv_id": "2505.15436", "doi": null, "citation_count": 45, "influential_citation_count": 5, "has_code": true, "code_url": "https://github.com/xtong-zhang/Chain-of-Focus", "venue": null, "quality_score": 0.4157}
{"id": "97a43220464dc6e2a7e24d2f555e35ae3b83140ea560a009b3ea3362e2306426", "sources": ["arxiv", "semantic_scholar"], "title": "Learning to Reason via Mixture-of-Thought for Logical Reasoning", "abstract": "Human beings naturally utilize multiple reasoning modalities to learn and solve logical problems, i.e., different representational formats such as natural language, code, and symbolic logic. In contrast, most existing LLM-based approaches operate with a single reasoning modality during training, typically natural language. Although some methods explored modality selection or augmentation at inference time, the training process remains modality-blind, limiting synergy among modalities. To fill in this gap, we propose Mixture-of-Thought (MoT), a framework that enables LLMs to reason across three complementary modalities: natural language, code, and a newly introduced symbolic modality, truth-table, which systematically enumerates logical cases and partially mitigates key failure modes in natural language reasoning. MoT adopts a two-phase design: (1) self-evolving MoT training, which jointly learns from filtered, self-generated rationales across modalities; and (2) MoT inference, which fully leverages the synergy of three modalities to produce better predictions. Experiments on logical reasoning benchmarks including FOLIO and ProofWriter demonstrate that our MoT framework consistently and significantly outperforms strong LLM baselines with single-modality chain-of-thought approaches, achieving up to +11.7pp average accuracy gain. Further analyses show that our MoT framework benefits both training and inference stages; that it is particularly effective on harder logical reasoning problems; and that different modalities contribute complementary strengths, with truth-table reasoning helping to overcome key bottlenecks in natural language inference.", "authors": ["Tong Zheng", "Lichang Chen", "Simeng Han", "R. Thomas McCoy", "Heng Huang"], "categories": ["cs.CL"], "fields_of_study": ["Computer Science"], "published_date": "2025-05-21", "url": "https://arxiv.org/abs/2505.15817", "pdf_url": "https://arxiv.org/pdf/2505.15817v2", "arxiv_id": "2505.15817", "doi": "10.48550/arXiv.2505.15817", "citation_count": 23, "influential_citation_count": 0, "has_code": false, "code_url": null, "venue": "arXiv.org", "quality_score": 0.3451}
{"id": "fdcf5305638678598e07fc3c4214504813ea03c6dcc691782e83d73297bdb001", "sources": ["arxiv", "semantic_scholar"], "title": "RL of Thoughts: Navigating LLM Reasoning with Inference-time Reinforcement Learning", "abstract": "Despite rapid advancements in large language models (LLMs), the token-level autoregressive nature constrains their complex reasoning capabilities. To enhance LLM reasoning, inference-time techniques, including Chain/Tree/Graph-of-Thought(s), successfully improve the performance, as they are fairly cost-effective by guiding reasoning through sophisticated logical structures without modifying LLMs' parameters. However, these manually predefined, task-agnostic frameworks are applied uniformly across diverse tasks, lacking adaptability. To improve this, we propose RL-of-Thoughts (RLoT), where we train a lightweight navigator model with reinforcement learning (RL) to adaptively enhance LLM reasoning at inference time. Specifically, we design five basic logic blocks from the perspective of human cognition. During the reasoning process, the trained RL navigator dynamically selects the suitable logic blocks and combines them into task-specific logical structures according to problem characteristics. Experiments across multiple reasoning benchmarks (AIME, MATH, GPQA, etc.) with multiple LLMs (GPT, Llama, Qwen, and DeepSeek) illustrate that RLoT outperforms established inference-time techniques by up to 13.4%. Remarkably, with less than 3K parameters, our RL navigator is able to make sub-10B LLMs comparable to 100B-scale counterparts. Moreover, the RL navigator demonstrates strong transferability: a model trained on one specific LLM-task pair can effectively generalize to unseen LLMs and tasks. Our code is open-source at https://github.com/tsinghua-fib-lab/RL-LLM-Reasoning for reproducibility.", "authors": ["Qianyue Hao", "Sibo Li", "Jian Yuan", "Yong Li"], "categories": ["cs.AI"], "fields_of_study": ["Computer Science"], "published_date": "2025-05-20", "url": "https://arxiv.org/abs/2505.14140", "pdf_url": "https://arxiv.org/pdf/2505.14140v3", "arxiv_id": "2505.14140", "doi": "10.48550/arXiv.2505.14140", "citation_count": 16, "influential_citation_count": 1, "has_code": true, "code_url": "https://github.com/tsinghua-fib-lab/RL-LLM-Reasoning", "venue": "arXiv.org", "quality_score": 0.3076}
{"id": "41e16231c4ed0ffa39c4abc75aeb1d3fa142edd53fcbd0ca4acf921e99a98532", "sources": ["arxiv", "semantic_scholar"], "title": "The Energy Cost of Reasoning: Analyzing Energy Usage in LLMs with Test-time Compute", "abstract": "Scaling large language models (LLMs) has driven significant advancements, yet it faces diminishing returns and escalating energy demands. This work explores how test-time compute (TTC) can serve as an energy-efficient complement to conventional scaling strategies by allocating additional computational resources at inference time rather than during training. Specifically, we investigate whether employing TTC can achieve superior accuracy-energy trade-offs compared to simply increasing model size. Our empirical analysis reveals that TTC surpasses traditional model scaling in accuracy/energy efficiency, with notable gains in tasks demanding complex reasoning rather than mere factual recall. Further, we identify a critical interaction between TTC performance and output sequence length, demonstrating that strategically adjusting compute resources at inference time according to query complexity can substantially enhance efficiency. Our findings advocate for TTC as a promising direction, enabling more sustainable, accurate, and adaptable deployment of future language models.", "authors": ["Yunho Jin", "Gu-Yeon Wei", "David Brooks"], "categories": ["cs.LG", "cs.AI"], "fields_of_study": ["Computer Science"], "published_date": "2025-05-20", "url": "https://arxiv.org/abs/2505.14733", "pdf_url": "https://arxiv.org/pdf/2505.14733v2", "arxiv_id": "2505.14733", "doi": "10.48550/arXiv.2505.14733", "citation_count": 14, "influential_citation_count": 1, "has_code": false, "code_url": null, "venue": "arXiv.org", "quality_score": 0.294}
{"id": "42b1df209d5188586c18be0ba653f6c09385f6dadd8210866371a31eb533e5c2", "sources": ["arxiv", "semantic_scholar"], "title": "Output Scaling: YingLong-Delayed Chain of Thought in a Large Pretrained Time Series Forecasting Model", "abstract": "We present a joint forecasting framework for time series prediction that contrasts with traditional direct or recursive methods. This framework achieves state-of-the-art performance for our designed foundation model, YingLong, and reveals a novel scaling effect: longer outputs significantly enhance model accuracy due to delayed chain-of-thought reasoning in our non-causal approach. YingLong is a non-causal, bidirectional attention encoder-only transformer trained through masked token recovery, aligning more effectively with language understanding tasks than with generation tasks. Additionally, we boost performance by tackling output variance with a multi-input ensemble. We release four foundation models ranging from 6M to 300M parameters, demonstrating superior results in zero-shot tasks on the ETT and Weather datasets. YingLong achieves more than 60% best performance. To ensure generalizability, we assessed the models using the GIFT-Eval benchmark, which comprises 23 time series datasets across 7 domains. Yinglong significantly outperformed the best time-series foundation models, end-to-end trained models by 14% and 44% in rank respectively.The pretrained 300M model is available at https://huggingface.co/qcw1314/YingLong_300m", "authors": ["Xue Wang", "Tian Zhou", "Jinyang Gao", "Bolin Ding", "Jingren Zhou"], "categories": ["cs.LG", "cs.AI"], "fields_of_study": ["Computer Science"], "published_date": "2025-05-20", "url": "https://arxiv.org/abs/2506.11029", "pdf_url": "https://arxiv.org/pdf/2506.11029v1", "arxiv_id": "2506.11029", "doi": "10.48550/arXiv.2506.11029", "citation_count": 14, "influential_citation_count": 6, "has_code": false, "code_url": null, "venue": "arXiv.org", "quality_score": 0.4225}
{"id": "1ae393c5f24e4b0aa3cabb3abff8b2f8d51e2427ef67ba0287e2b82ad6fd40d7", "sources": ["arxiv", "semantic_scholar"], "title": "Log-Augmented Generation: Scaling Test-Time Reasoning with Reusable Computation", "abstract": "While humans naturally learn and adapt from past experiences, large language models (LLMs) and their agentic counterparts struggle to retain reasoning from previous tasks and apply them in future contexts. To address this limitation, we propose a novel framework, log-augmented generation (LAG) that directly reuses prior computation and reasoning from past logs at test time to enhance model's ability to learn from previous tasks and perform better on new, unseen challenges, all while keeping the system efficient and scalable. Specifically, our system represents task logs using key-value (KV) caches, encoding the full reasoning context of prior tasks while storing KV caches for only a selected subset of tokens. When a new task arises, LAG retrieves the KV values from relevant logs to augment generation. Our approach differs from reflection-based memory mechanisms by directly reusing prior reasoning and computations without requiring additional steps for knowledge extraction or distillation. Our method also goes beyond existing KV caching techniques, which primarily target efficiency gains rather than improving accuracy. Experiments on knowledge- and reasoning-intensive datasets demonstrate that our method significantly outperforms standard agentic systems that do not utilize logs, as well as existing solutions based on reflection and KV cache techniques.", "authors": ["Peter Baile Chen", "Yi Zhang", "Dan Roth", "Samuel Madden", "Jacob Andreas", "Michael Cafarella"], "categories": ["cs.CL", "cs.AI", "cs.LG"], "fields_of_study": ["Computer Science"], "published_date": "2025-05-20", "url": "https://arxiv.org/abs/2505.14398", "pdf_url": "https://arxiv.org/pdf/2505.14398v1", "arxiv_id": "2505.14398", "doi": "10.48550/arXiv.2505.14398", "citation_count": 4, "influential_citation_count": 1, "has_code": false, "code_url": null, "venue": "arXiv.org", "quality_score": 0.1747}
{"id": "55f04237ff34c13d9e6cf301fe0085f78c56ce50dca1712d86920f5bd2ee1f21", "sources": ["arxiv", "semantic_scholar"], "title": "Fractured Chain-of-Thought Reasoning", "abstract": "Inference-time scaling techniques have significantly bolstered the reasoning capabilities of large language models (LLMs) by harnessing additional computational effort at inference without retraining. Similarly, Chain-of-Thought (CoT) prompting and its extension, Long CoT, improve accuracy by generating rich intermediate reasoning trajectories, but these approaches incur substantial token costs that impede their deployment in latency-sensitive settings. In this work, we first show that truncated CoT, which stops reasoning before completion and directly generates the final answer, often matches the full CoT sampling while using dramatically fewer tokens. Building on this insight, we introduce Fractured Sampling, a unified inference-time strategy that interpolates between full CoT and solution-only sampling along three orthogonal axes: (1) the number of reasoning trajectories, (2) the number of final solutions per trajectory, and (3) the depth at which reasoning traces are truncated. Through extensive experiments on five diverse reasoning benchmarks and several model scales, we demonstrate that Fractured Sampling consistently achieves superior accuracy-cost trade-offs, yielding steep log-linear scaling gains in Pass@k versus token budget. Our analysis reveals how to allocate computation across these dimensions to maximize performance, paving the way for more efficient and scalable LLM reasoning. Code is available at https://github.com/BaohaoLiao/frac-cot.", "authors": ["Baohao Liao", "Hanze Dong", "Yuhui Xu", "Doyen Sahoo", "Christof Monz", "Junnan Li", "Caiming Xiong"], "categories": ["cs.LG", "cs.AI", "cs.CL", "stat.ML"], "fields_of_study": ["Computer Science", "Mathematics"], "published_date": "2025-05-19", "url": "https://arxiv.org/abs/2505.12992", "pdf_url": "https://arxiv.org/pdf/2505.12992v4", "arxiv_id": "2505.12992", "doi": "10.48550/arXiv.2505.12992", "citation_count": 7, "influential_citation_count": 0, "has_code": true, "code_url": "https://github.com/BaohaoLiao/frac-cot", "venue": "arXiv.org", "quality_score": 0.255}
{"id": "3768b71f5b0202339c9a04c01624a1b9c32096da297d23e565b4506fd0aadd78", "sources": ["arxiv", "semantic_scholar"], "title": "Reasoning by Superposition: A Theoretical Perspective on Chain of Continuous Thought", "abstract": "Large Language Models (LLMs) have demonstrated remarkable performance in many applications, including challenging reasoning problems via chain-of-thoughts (CoTs) techniques that generate ``thinking tokens'' before answering the questions. While existing theoretical works demonstrate that CoTs with discrete tokens boost the capability of LLMs, recent work on continuous CoTs lacks a theoretical understanding of why it outperforms discrete counterparts in various reasoning tasks such as directed graph reachability, a fundamental graph reasoning problem that includes many practical domain applications as special cases. In this paper, we prove that a two-layer transformer with $D$ steps of continuous CoTs can solve the directed graph reachability problem, where $D$ is the diameter of the graph, while the best known result of constant-depth transformers with discrete CoTs requires $O(n^2)$ decoding steps where $n$ is the number of vertices ($D<n$). In our construction, each continuous thought vector is a superposition state that encodes multiple search frontiers simultaneously (i.e., parallel breadth-first search (BFS)), while discrete CoTs must choose a single path sampled from the superposition state, which leads to sequential search that requires many more steps and may be trapped into local solutions. We also performed extensive experiments to verify that our theoretical construction aligns well with the empirical solution obtained via training dynamics. Notably, encoding of multiple search frontiers as a superposition state automatically emerges in training continuous CoTs, without explicit supervision to guide the model to explore multiple paths simultaneously.", "authors": ["Hanlin Zhu", "Shibo Hao", "Zhiting Hu", "Jiantao Jiao", "Stuart Russell", "Yuandong Tian"], "categories": ["cs.LG"], "fields_of_study": ["Computer Science"], "published_date": "2025-05-18", "url": "https://arxiv.org/abs/2505.12514", "pdf_url": "https://arxiv.org/pdf/2505.12514v3", "arxiv_id": "2505.12514", "doi": "10.48550/arXiv.2505.12514", "citation_count": 61, "influential_citation_count": 5, "has_code": false, "code_url": null, "venue": "arXiv.org", "quality_score": 0.4481}
{"id": "4dbca6bcf76d4a52abec3fa639e2ad1ddfa03f1ef23f314202630804f69c4e9b", "sources": ["arxiv", "semantic_scholar"], "title": "SoftCoT++: Test-Time Scaling with Soft Chain-of-Thought Reasoning", "abstract": "Test-Time Scaling (TTS) refers to approaches that improve reasoning performance by allocating extra computation during inference, without altering the model's parameters. While existing TTS methods operate in a discrete token space by generating more intermediate steps, recent studies in Coconut and SoftCoT have demonstrated that thinking in the continuous latent space can further enhance the reasoning performance. Such latent thoughts encode informative thinking without the information loss associated with autoregressive token generation, sparking increased interest in continuous-space reasoning. Unlike discrete decoding, where repeated sampling enables exploring diverse reasoning paths, latent representations in continuous space are fixed for a given input, which limits diverse exploration, as all decoded paths originate from the same latent thought. To overcome this limitation, we introduce SoftCoT++ to extend SoftCoT to the Test-Time Scaling paradigm by enabling diverse exploration of thinking paths. Specifically, we perturb latent thoughts via multiple specialized initial tokens and apply contrastive learning to promote diversity among soft thought representations. Experiments across five reasoning benchmarks and two distinct LLM architectures demonstrate that SoftCoT++ significantly boosts SoftCoT and also outperforms SoftCoT with self-consistency scaling. Moreover, it shows strong compatibility with conventional scaling techniques such as self-consistency. Source code is available at https://github.com/xuyige/SoftCoT.", "authors": ["Yige Xu", "Xu Guo", "Zhiwei Zeng", "Chunyan Miao"], "categories": ["cs.CL"], "fields_of_study": ["Computer Science"], "published_date": "2025-05-16", "url": "https://arxiv.org/abs/2505.11484", "pdf_url": "https://arxiv.org/pdf/2505.11484v2", "arxiv_id": "2505.11484", "doi": "10.48550/arXiv.2505.11484", "citation_count": 24, "influential_citation_count": 5, "has_code": true, "code_url": "https://github.com/xuyige/SoftCoT", "venue": "arXiv.org", "quality_score": 0.3891}
{"id": "1a0d88023f7e3559b84132916bb1a2ee59cae09faa33caab0785361a89e1edae", "sources": ["arxiv", "semantic_scholar"], "title": "Roman Observations Time Allocation Committee: Final Report and Recommendations", "abstract": "The Nancy Grace Roman Space Telescope is poised to revolutionize our scientific understanding of exoplanets, dark matter, dark energy, and general astrophysics, including through an innovative community approach to defining and executing sky surveys. The Roman Observations Time Allocation Committee (ROTAC) was convened to recommend time allocations for the three Core Community Surveys (CCS) using the Wide Field Instrument (WFI): the High Latitude Wide Area Survey, the High Latitude Time Domain Survey, and the Galactic Bulge Time Domain Survey, as well as balance the time allocation for the General Astrophysics Surveys. Each CCS had a corresponding Definition Committee that collected community input and designed proposals for a nominal (in-guide) survey, as well as underguide and overguide options with smaller and larger time allocations, respectively. These options explored different ways of fulfilling the mission science requirements while maximizing general astrophysics science goals enabled by the surveys. In this report, the ROTAC lays out its recommendations for the three CCS observing designs and the WFI time allotment for CCS (74.5%) and the General Astrophysics Surveys (25.5%).", "authors": ["Roman Observations Time Allocation Committee", "Core Community Survey Definition Committees"], "categories": ["astro-ph.IM", "astro-ph.CO", "astro-ph.EP", "astro-ph.GA", "astro-ph.HE", "astro-ph.SR"], "fields_of_study": ["Physics"], "published_date": "2025-05-13", "url": "https://arxiv.org/abs/2505.10574", "pdf_url": "https://arxiv.org/pdf/2505.10574v2", "arxiv_id": "2505.10574", "doi": null, "citation_count": 3, "influential_citation_count": 1, "has_code": false, "code_url": null, "venue": null, "quality_score": 0.1505}
{"id": "c92da9ca7dc77f18140a29fbe5c659abe3854077e9683db694cb82f79fbb13a9", "sources": ["arxiv", "semantic_scholar"], "title": "Shots and Boosters: Exploring the Use of Combined Prebunking Interventions to Raise Critical Thinking and Create Long-Term Protection Against Misinformation", "abstract": "The problem of how to effectively mitigate the flow of misinformation remains a significant challenge. The classical approach to this is public disapproval of claims or \"debunking.\" The approach is still widely used on social media, but it has some severe limitations in terms of applicability and efficiency. An alternative strategy is to enhance individuals' critical thinking through educational interventions. Instead of merely disproving misinformation, these approaches aim to strengthen users' reasoning skills, enabling them to evaluate and reject false information independently. In this position paper, we explore a combination of intervention methods designed to improve critical thinking in the context of online media consumption. We highlight the role of AI in supporting different stages of these interventions and present a design concept that integrates AI-driven strategies to foster critical reasoning and media literacy.", "authors": ["Huiyun Tang", "Anastasia Sergeeva"], "categories": ["cs.HC"], "fields_of_study": ["Computer Science"], "published_date": "2025-05-12", "url": "https://arxiv.org/abs/2505.07486", "pdf_url": "https://arxiv.org/pdf/2505.07486v1", "arxiv_id": "2505.07486", "doi": "10.48550/arXiv.2505.07486", "citation_count": 4, "influential_citation_count": 0, "has_code": false, "code_url": null, "venue": "arXiv.org", "quality_score": 0.1747}
{"id": "6e0bb360f084842872b59fdcef282ed7ee63923944efc965b643f9449f1c8cd2", "sources": ["arxiv", "semantic_scholar"], "title": "Scalable Chain of Thoughts via Elastic Reasoning", "abstract": "Large reasoning models (LRMs) have achieved remarkable progress on complex tasks by generating extended chains of thought (CoT). However, their uncontrolled output lengths pose significant challenges for real-world deployment, where inference-time budgets on tokens, latency, or compute are strictly constrained. We propose Elastic Reasoning, a novel framework for scalable chain of thoughts that explicitly separates reasoning into two phases--thinking and solution--with independently allocated budgets. At test time, Elastic Reasoning prioritizes the completeness of solution segments, significantly improving reliability under tight resource constraints. To train models that are robust to truncated thinking, we introduce a lightweight budget-constrained rollout strategy, integrated into GRPO, which teaches the model to reason adaptively when the thinking process is cut short and generalizes effectively to unseen budget constraints without additional training. Empirical results on mathematical (AIME, MATH500) and programming (LiveCodeBench, Codeforces) benchmarks demonstrate that Elastic Reasoning performs robustly under strict budget constraints, while incurring significantly lower training cost than baseline methods. Remarkably, our approach also produces more concise and efficient reasoning even in unconstrained settings. Our code has been made available at https://github.com/SalesforceAIResearch/Elastic-Reasoning.", "authors": ["Yuhui Xu", "Hanze Dong", "Lei Wang", "Doyen Sahoo", "Junnan Li", "Caiming Xiong"], "categories": ["cs.LG", "cs.AI", "cs.CL"], "fields_of_study": ["Computer Science"], "published_date": "2025-05-08", "url": "https://arxiv.org/abs/2505.05315", "pdf_url": "https://arxiv.org/pdf/2505.05315v2", "arxiv_id": "2505.05315", "doi": "10.48550/arXiv.2505.05315", "citation_count": 40, "influential_citation_count": 5, "has_code": true, "code_url": "https://github.com/SalesforceAIResearch/Elastic-Reasoning", "venue": "arXiv.org", "quality_score": 0.4032}
{"id": "9987bf470a2df25f463c9390172c1b46612ddd4d6cc93e47342930a837df8b55", "sources": ["arxiv", "semantic_scholar"], "title": "Crosslingual Reasoning through Test-Time Scaling", "abstract": "Reasoning capabilities of large language models are primarily studied for English, even when pretrained models are multilingual. In this work, we investigate to what extent English reasoning finetuning with long chain-of-thoughts (CoTs) can generalize across languages. First, we find that scaling up inference compute for English-centric reasoning language models (RLMs) improves multilingual mathematical reasoning across many languages including low-resource languages, to an extent where they outperform models twice their size. Second, we reveal that while English-centric RLM's CoTs are naturally predominantly English, they consistently follow a quote-and-think pattern to reason about quoted non-English inputs. Third, we discover an effective strategy to control the language of long CoT reasoning, and we observe that models reason better and more efficiently in high-resource languages. Finally, we observe poor out-of-domain reasoning generalization, in particular from STEM to cultural commonsense knowledge, even for English. Overall, we demonstrate the potentials, study the mechanisms and outline the limitations of crosslingual generalization of English reasoning test-time scaling. We conclude that practitioners should let English-centric RLMs reason in high-resource languages, while further work is needed to improve reasoning in low-resource languages and out-of-domain contexts.", "authors": ["Zheng-Xin Yong", "M. Farid Adilazuarda", "Jonibek Mansurov", "Ruochen Zhang", "Niklas Muennighoff", "Carsten Eickhoff", "Genta Indra Winata", "Julia Kreutzer", "Stephen H. Bach", "Alham Fikri Aji"], "categories": ["cs.CL", "cs.AI", "cs.LG"], "fields_of_study": ["Computer Science"], "published_date": "2025-05-08", "url": "https://arxiv.org/abs/2505.05408", "pdf_url": "https://arxiv.org/pdf/2505.05408v1", "arxiv_id": "2505.05408", "doi": "10.48550/arXiv.2505.05408", "citation_count": 43, "influential_citation_count": 4, "has_code": false, "code_url": null, "venue": "arXiv.org", "quality_score": 0.4109}
{"id": "20bbdca5353204c957957cff30bb114b775566c2f0bd7ef64e099afe62430474", "sources": ["arxiv", "semantic_scholar"], "title": "Putting the Value Back in RL: Better Test-Time Scaling by Unifying LLM Reasoners With Verifiers", "abstract": "Prevalent reinforcement learning~(RL) methods for fine-tuning LLM reasoners, such as GRPO or Leave-one-out PPO, abandon the learned value function in favor of empirically estimated returns. This hinders test-time compute scaling that relies on using the value-function for verification. Yet if parallel test-time compute is already part of the deployment plan, training should be designed to support it. In this work, we propose RL$^V$ that augments any ``value-free'' RL method by jointly training the LLM as both a reasoner and a generative verifier using RL-generated data, adding verification capabilities without significant overhead. Empirically, RL$^V$ boosts MATH accuracy by over 20\\% with parallel sampling and enables $8-32\\times$ efficient test-time compute scaling compared to the base RL method. RL$^V$ also exhibits strong generalization capabilities for both easy-to-hard and out-of-domain tasks. Furthermore, RL$^V$ achieves $1.2-1.6\\times$ higher performance when jointly scaling parallel and sequential test-time compute with a long reasoning R1 model. More broadly, RL$^V$ instantiates the principle of co-training for test-time scaling: jointly optimizing for task performance and a capability useful at inference, using data that RL training already produces.", "authors": ["Kusha Sareen", "Morgane M Moss", "Alessandro Sordoni", "Rishabh Agarwal", "Arian Hosseini"], "categories": ["cs.LG", "cs.AI"], "fields_of_study": ["Computer Science"], "published_date": "2025-05-07", "url": "https://arxiv.org/abs/2505.04842", "pdf_url": "https://arxiv.org/pdf/2505.04842v2", "arxiv_id": "2505.04842", "doi": "10.48550/arXiv.2505.04842", "citation_count": 15, "influential_citation_count": 1, "has_code": false, "code_url": null, "venue": "arXiv.org", "quality_score": 0.301}
{"id": "d407ada93657c045d53def12e293a96f149efb9cbd14f4194364e5d98251cfd5", "sources": ["arxiv", "semantic_scholar"], "title": "Long-Short Chain-of-Thought Mixture Supervised Fine-Tuning Eliciting Efficient Reasoning in Large Language Models", "abstract": "Recent advances in large language models have demonstrated that Supervised Fine-Tuning (SFT) with Chain-of-Thought (CoT) reasoning data distilled from large reasoning models (e.g., DeepSeek R1) can effectively transfer reasoning capabilities to non-reasoning models. However, models fine-tuned with this approach inherit the \"overthinking\" problem from teacher models, producing verbose and redundant reasoning chains during inference. To address this challenge, we propose Long-Short Chain-of-Thought Mixture Supervised Fine-Tuning (LS-Mixture SFT), which combines long CoT reasoning dataset with their short counterparts obtained through structure-preserved rewriting. Our experiments demonstrate that models trained using the LS-Mixture SFT method, compared to those trained with direct SFT, achieved an average accuracy improvement of 2.3% across various benchmarks while substantially reducing model response length by approximately 47.61%. This work offers an approach to endow non-reasoning models with reasoning capabilities through supervised fine-tuning while avoiding the inherent overthinking problems inherited from teacher models, thereby enabling efficient reasoning in the fine-tuned models.", "authors": ["Bin Yu", "Hang Yuan", "Haotian Li", "Xueyin Xu", "Yuliang Wei", "Bailing Wang", "Weizhen Qi", "Kai Chen"], "categories": ["cs.CL"], "fields_of_study": ["Computer Science"], "published_date": "2025-05-06", "url": "https://arxiv.org/abs/2505.03469", "pdf_url": "https://arxiv.org/pdf/2505.03469v2", "arxiv_id": "2505.03469", "doi": "10.48550/arXiv.2505.03469", "citation_count": 31, "influential_citation_count": 3, "has_code": false, "code_url": null, "venue": "arXiv.org", "quality_score": 0.3763}
{"id": "1c10f410a2ed35d333e907197434d236291fb0e8390c644a477d614eb69e5e7a", "sources": ["arxiv", "semantic_scholar"], "title": "When Reasoning Beats Scale: A 1.5B Reasoning Model Outranks 13B LLMs as Discriminator", "abstract": "Large Language Models (LLM) with reasoning capabilities offer a promising path for improving candidate evaluation in planning frameworks, but their relative performance against traditional non-reasoning models remains largely underexplored. In this study, we benchmark a distilled 1.5B parameter reasoning model (DeepSeek-R1) against several state-of-the-art non-reasoning LLMs within a generator-discriminator LLM planning framework for the text-to-SQL task. For this, we introduce a novel method for extracting soft scores from the chain-of-thought (CoT) outputs from reasoning that enables fine-grained ranking of candidates. Our central hypothesis is that reasoning models are more effective discriminators than non-reasoning LLMs. Our results show that distilled DeepSeek-R1-1.5B achieves up to $87\\%$ higher F1 and $3.7\\%$ better discrimination accuracy than CodeLlama-7B, as well as $3.7\\%$ higher execution accuracy than CodeLlama-13B, despite having significantly fewer parameters. Furthermore, we find that there is a limit to the logical capabilities of reasoning models, and only providing more context or allowing more compute budget for reasoning is not enough to improve their discrimination performance. Finally, we demonstrate that, unlike non-reasoning LLMs, reasoning models find generation more challenging than discrimination and may underperform as generators compared to smaller non-reasoning LLMs. Our work highlights the potential of reasoning models as discriminators in agentic frameworks, far outweighing their capabilities as generators, offering insights into their optimal role within LLM planning infrastructures.", "authors": ["Md Fahim Anjum"], "categories": ["cs.LG", "cs.CL"], "fields_of_study": ["Computer Science"], "published_date": "2025-04-30", "url": "https://arxiv.org/abs/2505.03786", "pdf_url": "https://arxiv.org/pdf/2505.03786v1", "arxiv_id": "2505.03786", "doi": "10.48550/arXiv.2505.03786", "citation_count": 1, "influential_citation_count": 0, "has_code": true, "code_url": "https://github.com/MDFahimAnjum/llm-planning-with-reasoning", "venue": "arXiv.org", "quality_score": 0.2214}
{"id": "a754294bef46e990c790d7e0f3ef9138fde493d7c37912656d5ee4932897cd69", "sources": ["arxiv", "semantic_scholar"], "title": "Chain-of-Defensive-Thought: Structured Reasoning Elicits Robustness in Large Language Models against Reference Corruption", "abstract": "Chain-of-thought prompting has demonstrated great success in facilitating the reasoning abilities of large language models. In this work, we explore how these enhanced reasoning abilities can be exploited to improve the robustness of large language models in tasks that are not necessarily reasoning-focused. In particular, we show how a wide range of large language models exhibit significantly improved robustness against reference corruption using a simple method called chain-of-defensive-thought, where only a few exemplars with structured and defensive reasoning are provided as demonstrations. Empirically, the improvements can be astounding, especially given the simplicity and applicability of the method. For example, in the Natural Questions task, the accuracy of GPT-4o degrades from 60% to as low as 3% with standard prompting when 1 out of 10 references provided is corrupted with prompt injection attacks. In contrast, GPT-4o using chain-of-defensive-thought prompting maintains an accuracy of 50%.", "authors": ["Wenxiao Wang", "Parsa Hosseini", "Soheil Feizi"], "categories": ["cs.CL", "cs.AI", "cs.LG"], "fields_of_study": ["Computer Science"], "published_date": "2025-04-29", "url": "https://arxiv.org/abs/2504.20769", "pdf_url": "https://arxiv.org/pdf/2504.20769v1", "arxiv_id": "2504.20769", "doi": "10.48550/arXiv.2504.20769", "citation_count": 4, "influential_citation_count": 0, "has_code": false, "code_url": null, "venue": "arXiv.org", "quality_score": 0.1747}
{"id": "502e3bca9e13eab52e444124589cf7edb25b3c039acbbb8700d93515cabe3a1e", "sources": ["arxiv", "semantic_scholar"], "title": "Efficient Reasoning for LLMs through Speculative Chain-of-Thought", "abstract": "Large reasoning language models such as OpenAI-o1 and Deepseek-R1 have recently attracted widespread attention due to their impressive task-solving abilities. However, the enormous model size and the generation of lengthy thought chains introduce significant reasoning costs and response latency. Existing methods for efficient reasoning mainly focus on reducing the number of model parameters or shortening the chain-of-thought length. In this paper, we introduce Speculative Chain-of-Thought (SCoT), which reduces reasoning latency from another perspective by accelerated average reasoning speed through large and small model collaboration. SCoT conducts thought-level drafting using a lightweight draft model. Then it selects the best CoT draft and corrects the error cases with the target model. The proposed thinking behavior alignment improves the efficiency of drafting and the draft selection strategy maintains the prediction accuracy of the target model for complex tasks. Experimental results on GSM8K, MATH, GaoKao, CollegeMath and Olympiad datasets show that SCoT reduces reasoning latency by 48\\%$\\sim$66\\% and 21\\%$\\sim$49\\% for Deepseek-R1-Distill-Qwen-32B and Deepseek-R1-Distill-Llama-70B while achieving near-target-model-level performance. Our code is available at https://github.com/Jikai0Wang/Speculative_CoT.", "authors": ["Jikai Wang", "Juntao Li", "Jianye Hou", "Bowen Yan", "Lijun Wu", "Min Zhang"], "categories": ["cs.CL"], "fields_of_study": ["Computer Science"], "published_date": "2025-04-27", "url": "https://arxiv.org/abs/2504.19095", "pdf_url": "https://arxiv.org/pdf/2504.19095v2", "arxiv_id": "2504.19095", "doi": "10.48550/arXiv.2504.19095", "citation_count": 21, "influential_citation_count": 3, "has_code": true, "code_url": "https://github.com/Jikai0Wang/Speculative_CoT", "venue": "arXiv.org", "quality_score": 0.3356}
{"id": "58fc1a7ce9a9622cf9982e09ae7f3a1ebd3cf4fd24f767b16629a36c16bb40a6", "sources": ["arxiv", "semantic_scholar"], "title": "Cognitive Dissonance Artificial Intelligence (CD-AI): The Mind at War with Itself. Harnessing Discomfort to Sharpen Critical Thinking", "abstract": "AI-augmented systems are traditionally designed to streamline human decision-making by minimizing cognitive load, clarifying arguments, and optimizing efficiency. However, in a world where algorithmic certainty risks becoming an Orwellian tool of epistemic control, true intellectual growth demands not passive acceptance but active struggle. Drawing on the dystopian visions of George Orwell and Philip K. Dick - where reality is unstable, perception malleable, and truth contested - this paper introduces Cognitive Dissonance AI (CD-AI): a novel framework that deliberately sustains uncertainty rather than resolving it. CD-AI does not offer closure, but compels users to navigate contradictions, challenge biases, and wrestle with competing truths. By delaying resolution and promoting dialectical engagement, CD-AI enhances reflective reasoning, epistemic humility, critical thinking, and adaptability in complex decision-making. This paper examines the theoretical foundations of the approach, presents an implementation model, explores its application in domains such as ethics, law, politics, and science, and addresses key ethical concerns - including decision paralysis, erosion of user autonomy, cognitive manipulation, and bias in AI reasoning. In reimagining AI as an engine of doubt rather than a deliverer of certainty, CD-AI challenges dominant paradigms of AI-augmented reasoning and offers a new vision - one in which AI sharpens the mind not by resolving conflict, but by sustaining it. Rather than reinforcing Huxleyan complacency or pacifying the user into intellectual conformity, CD-AI echoes Nietzsche's vision of the Uebermensch - urging users to transcend passive cognition through active epistemic struggle.", "authors": ["Delia Deliu"], "categories": ["cs.HC", "cs.CY"], "fields_of_study": ["Computer Science"], "published_date": "2025-04-23", "url": "https://arxiv.org/abs/2507.08804", "pdf_url": "https://arxiv.org/pdf/2507.08804v1", "arxiv_id": "2507.08804", "doi": "10.48550/arXiv.2507.08804", "citation_count": 1, "influential_citation_count": 0, "has_code": false, "code_url": null, "venue": "arXiv.org", "quality_score": 0.1352}
{"id": "611a27e414bff6cae8df4d1b1451e02848e2588b42361d2be3bf11e7593761f7", "sources": ["arxiv", "semantic_scholar"], "title": "Enhancing Critical Thinking with AI: A Tailored Warning System for RAG Models", "abstract": "Retrieval-Augmented Generation (RAG) systems offer a powerful approach to enhancing large language model (LLM) outputs by incorporating fact-checked, contextually relevant information. However, fairness and reliability concerns persist, as hallucinations can emerge at both the retrieval and generation stages, affecting users' reasoning and decision-making. Our research explores how tailored warning messages -- whose content depends on the specific context of hallucination -- shape user reasoning and actions in an educational quiz setting. Preliminary findings suggest that while warnings improve accuracy and awareness of high-level hallucinations, they may also introduce cognitive friction, leading to confusion and diminished trust in the system. By examining these interactions, this work contributes to the broader goal of AI-augmented reasoning: developing systems that actively support human reflection, critical thinking, and informed decision-making rather than passive information consumption.", "authors": ["Xuyang Zhu", "Sejoon Chang", "Andrew Kuik"], "categories": ["cs.HC"], "fields_of_study": ["Computer Science"], "published_date": "2025-04-23", "url": "https://arxiv.org/abs/2504.16883", "pdf_url": "https://arxiv.org/pdf/2504.16883v1", "arxiv_id": "2504.16883", "doi": "10.48550/arXiv.2504.16883", "citation_count": 1, "influential_citation_count": 0, "has_code": false, "code_url": null, "venue": "arXiv.org", "quality_score": 0.1352}
{"id": "9cb4fb726ece338b8a49e92f3264cb634723c2dd650e9292a691b6d80f95ca8d", "sources": ["arxiv", "semantic_scholar"], "title": "Navigating the State of Cognitive Flow: Context-Aware AI Interventions for Effective Reasoning Support", "abstract": "Flow theory describes an optimal cognitive state where individuals experience deep focus and intrinsic motivation when a task's difficulty aligns with their skill level. In AI-augmented reasoning, interventions that disrupt the state of cognitive flow can hinder rather than enhance decision-making. This paper proposes a context-aware cognitive augmentation framework that adapts interventions based on three key contextual factors: type, timing, and scale. By leveraging multimodal behavioral cues (e.g., gaze behavior, typing hesitation, interaction speed), AI can dynamically adjust cognitive support to maintain or restore flow. We introduce the concept of cognitive flow, an extension of flow theory in AI-augmented reasoning, where interventions are personalized, adaptive, and minimally intrusive. By shifting from static interventions to context-aware augmentation, our approach ensures that AI systems support deep engagement in complex decision-making and reasoning without disrupting cognitive immersion.", "authors": ["Dinithi Dissanayake", "Suranga Nanayakkara"], "categories": ["cs.HC", "cs.AI"], "fields_of_study": ["Computer Science"], "published_date": "2025-04-22", "url": "https://arxiv.org/abs/2504.16021", "pdf_url": "https://arxiv.org/pdf/2504.16021v1", "arxiv_id": "2504.16021", "doi": "10.48550/arXiv.2504.16021", "citation_count": 1, "influential_citation_count": 0, "has_code": false, "code_url": null, "venue": "arXiv.org", "quality_score": 0.1341}
{"id": "1f74e082556d30cc8206ae36cb4e2a571fd01c0e9f35536e375920938d46a858", "sources": ["arxiv", "semantic_scholar"], "title": "Distributed Cognition for AI-supported Remote Operations: Challenges and Research Directions", "abstract": "This paper investigates the impact of artificial intelligence integration on remote operations, emphasising its influence on both distributed and team cognition. As remote operations increasingly rely on digital interfaces, sensors, and networked communication, AI-driven systems transform decision-making processes across domains such as air traffic control, industrial automation, and intelligent ports. However, the integration of AI introduces significant challenges, including the reconfiguration of human-AI team cognition, the need for adaptive AI memory that aligns with human distributed cognition, and the design of AI fallback operators to maintain continuity during communication disruptions. Drawing on theories of distributed and team cognition, we analyse how cognitive overload, loss of situational awareness, and impaired team coordination may arise in AI-supported environments. Based on real-world intelligent port scenarios, we propose research directions that aim to safeguard human reasoning and enhance collaborative decision-making in AI-augmented remote operations.", "authors": ["Rune Møberg Jacobsen", "Joel Wester", "Helena Bøjer Djernæs", "Niels van Berkel"], "categories": ["cs.HC"], "fields_of_study": ["Computer Science"], "published_date": "2025-04-21", "url": "https://arxiv.org/abs/2504.14996", "pdf_url": "https://arxiv.org/pdf/2504.14996v1", "arxiv_id": "2504.14996", "doi": "10.48550/arXiv.2504.14996", "citation_count": 5, "influential_citation_count": 0, "has_code": false, "code_url": null, "venue": "arXiv.org", "quality_score": 0.1945}
{"id": "f055ba594009042ae4c64998019b8912bf2e54b411cc3336d127610e8b9303e6", "sources": ["arxiv", "semantic_scholar"], "title": "Learning Adaptive Parallel Reasoning with Language Models", "abstract": "Scaling inference-time computation has substantially improved the reasoning capabilities of language models. However, existing methods have significant limitations: serialized chain-of-thought approaches generate overly long outputs, leading to increased latency and exhausted context windows, while parallel methods such as self-consistency suffer from insufficient coordination, resulting in redundant computations and limited performance gains. To address these shortcomings, we propose Adaptive Parallel Reasoning (APR), a novel reasoning framework that enables language models to orchestrate both serialized and parallel computations end-to-end. APR generalizes existing reasoning methods by enabling adaptive multi-threaded inference using spawn() and join() operations. A key innovation is our end-to-end reinforcement learning strategy, optimizing both parent and child inference threads to enhance task success rate without requiring predefined reasoning structures. Experiments on the Countdown reasoning task demonstrate significant benefits of APR: (1) higher performance within the same context window (83.4% vs. 60.0% at 4k context); (2) superior scalability with increased computation (80.1% vs. 66.6% at 20k total tokens); (3) improved accuracy at equivalent latency (75.2% vs. 57.3% at approximately 5,000ms). APR represents a step towards enabling language models to autonomously optimize their reasoning processes through adaptive allocation of computation.", "authors": ["Jiayi Pan", "Xiuyu Li", "Long Lian", "Charlie Snell", "Yifei Zhou", "Adam Yala", "Trevor Darrell", "Kurt Keutzer", "Alane Suhr"], "categories": ["cs.AI", "cs.CL"], "fields_of_study": ["Computer Science"], "published_date": "2025-04-21", "url": "https://arxiv.org/abs/2504.15466", "pdf_url": "https://arxiv.org/pdf/2504.15466v2", "arxiv_id": "2504.15466", "doi": "10.48550/arXiv.2504.15466", "citation_count": 60, "influential_citation_count": 4, "has_code": true, "code_url": "https://github.com/Parallel-Reasoning/APR", "venue": "arXiv.org", "quality_score": 0.4463}
{"id": "0c3419613ea2ebcb60c0d165907aac38f5fa2d8f4dade6325088224c6552e185", "sources": ["arxiv", "semantic_scholar"], "title": "Intelligent Interaction Strategies for Context-Aware Cognitive Augmentation", "abstract": "Human cognition is constrained by processing limitations, leading to cognitive overload and inefficiencies in knowledge synthesis and decision-making. Large Language Models (LLMs) present an opportunity for cognitive augmentation, but their current reactive nature limits their real-world applicability. This position paper explores the potential of context-aware cognitive augmentation, where LLMs dynamically adapt to users' cognitive states and task environments to provide appropriate support. Through a think-aloud study in an exhibition setting, we examine how individuals interact with multi-modal information and identify key cognitive challenges in structuring, retrieving, and applying knowledge. Our findings highlight the need for AI-driven cognitive support systems that integrate real-time contextual awareness, personalized reasoning assistance, and socially adaptive interactions. We propose a framework for AI augmentation that seamlessly transitions between real-time cognitive support and post-experience knowledge organization, contributing to the design of more effective human-centered AI systems.", "authors": [" Xiangrong", " Zhu", "Yuan Xu", "Tianjian Liu", "Jingwei Sun", "Yu Zhang", "Xin Tong"], "categories": ["cs.HC"], "fields_of_study": ["Computer Science"], "published_date": "2025-04-18", "url": "https://arxiv.org/abs/2504.13684", "pdf_url": "https://arxiv.org/pdf/2504.13684v1", "arxiv_id": "2504.13684", "doi": "10.48550/arXiv.2504.13684", "citation_count": 1, "influential_citation_count": 0, "has_code": false, "code_url": null, "venue": "arXiv.org", "quality_score": 0.1295}
{"id": "eac64937f09ea47969b3fccce96076f6feca58a6e401e216f748319ecf1b0d69", "sources": ["arxiv", "semantic_scholar"], "title": "CoT-RAG: Integrating Chain of Thought and Retrieval-Augmented Generation to Enhance Reasoning in Large Language Models", "abstract": "Chain-of-thought (CoT) reasoning boosts large language models' (LLMs) performance on complex tasks but faces two key limitations: a lack of reliability when solely relying on LLM-generated reasoning chains and lower reasoning performance from natural language prompts compared with code prompts. To address these issues, we propose CoT-RAG, a novel reasoning framework with three key designs: (i) Knowledge Graph-driven CoT Generation, featuring knowledge graphs to modulate reasoning chain generation of LLMs, thereby enhancing reasoning credibility; (ii) Learnable Knowledge Case-aware RAG, which incorporates retrieval-augmented generation (RAG) into knowledge graphs to retrieve relevant sub-cases and sub-descriptions, providing LLMs with learnable information; (iii) Pseudo Program Prompting Execution, which promotes greater logical rigor by guiding LLMs to execute reasoning tasks as pseudo-programs. Evaluations on nine public datasets spanning three reasoning tasks reveal significant accuracy gains-ranging from 4.0% to 44.3%-over state-of-the-art methods. Furthermore, tests on four domain-specific datasets demonstrate exceptional accuracy and efficient execution, underscoring its practical applicability and scalability. Our code and data are available at https: //github.com/hustlfy123/CoT-RAG.", "authors": ["Feiyang Li", "Peng Fang", "Zhan Shi", "Arijit Khan", "Fang Wang", "Weihao Wang", "Xin Zhang", "Yongjian Cui"], "categories": ["cs.CL", "cs.AI"], "fields_of_study": ["Computer Science"], "published_date": "2025-04-18", "url": "https://arxiv.org/abs/2504.13534", "pdf_url": "https://arxiv.org/pdf/2504.13534v3", "arxiv_id": "2504.13534", "doi": "10.48550/arXiv.2504.13534", "citation_count": 12, "influential_citation_count": 0, "has_code": true, "code_url": null, "venue": "Conference on Empirical Methods in Natural Language Processing", "quality_score": 0.2785}
{"id": "965f89df035d31c546d3e43fbb93369250d2cf6e6a6f04ced1eca59ea6525548", "sources": ["arxiv", "semantic_scholar"], "title": "Thought Manipulation: External Thought Can Be Efficient for Large Reasoning Models", "abstract": "Recent advancements in large reasoning models (LRMs) have demonstrated the effectiveness of scaling test-time computation to enhance reasoning capabilities on various tasks. However, LRMs often suffer from an ``overthinking'' problem, where the model generates excessively redundant reasoning steps with limited performance gains. In this work, we empirically reveal an important characteristic of LRM behaviors that placing external CoTs generated by smaller models between the thinking token (\\texttt{<think>} and \\texttt{</think>}) can effectively manipulate the model to generate fewer thoughts. Building on this finding, we propose a simple yet efficient pipeline, \\Method, to enable LRMs to bypass unnecessary intermediate steps, thereby significantly reducing computational costs. We conduct extensive experiments to evaluate the utility and efficiency of \\Method. For instance, when applied to QwQ-32B on the LiveBench/Code dataset, \\Method keeps the original performance while reducing output token counts by approximately 30\\%, with minimal overhead introduced by the CoT generator. Furthermore, we identify two suboptimal modes, blindly following flawed external thoughts and unnecessary rethinking, and show that simple mitigations, such as difficulty-aware fallbacks, can further improve performance. Overall, \\Method offers a practical, general, and efficient way to optimize LRM inference, making powerful reasoning models more accessible and scalable for real-world applications.", "authors": ["Yule Liu", "Jingyi Zheng", "Zhen Sun", "Zifan Peng", "Wenhan Dong", "Zeyang Sha", "Shiwen Cui", "Weiqiang Wang", "Xinlei He"], "categories": ["cs.CL", "cs.AI"], "fields_of_study": ["Computer Science"], "published_date": "2025-04-18", "url": "https://arxiv.org/abs/2504.13626", "pdf_url": "https://arxiv.org/pdf/2504.13626v2", "arxiv_id": "2504.13626", "doi": "10.48550/arXiv.2504.13626", "citation_count": 26, "influential_citation_count": 3, "has_code": false, "code_url": null, "venue": "arXiv.org", "quality_score": 0.3578}
{"id": "c70da377dc3461d656b3a6a34a078b073feec15b3a9b91f9c3d9f55aff6d7ac1", "sources": ["arxiv", "semantic_scholar"], "title": "Sleep-time Compute: Beyond Inference Scaling at Test-time", "abstract": "Scaling test-time compute has emerged as a key ingredient for enabling large language models (LLMs) to solve difficult problems, but comes with high latency and inference cost. We introduce sleep-time compute, which allows models to \"think\" offline about contexts before queries are presented: by anticipating what queries users might ask and pre-computing useful quantities, we can significantly reduce the compute requirements at test-time. To demonstrate the efficacy of our method, we create modified versions of two reasoning tasks - Stateful GSM-Symbolic and Stateful AIME. We find that sleep-time compute can reduce the amount of test-time compute needed to achieve the same accuracy by ~ 5x on Stateful GSM-Symbolic and Stateful AIME and that by scaling sleep-time compute we can further increase accuracy by up to 13% on Stateful GSM-Symbolic and 18% on Stateful AIME. Furthermore, we introduce Multi-Query GSM-Symbolic, which extends GSM-Symbolic by including multiple related queries per context. By amortizing sleep-time compute across related queries about the same context using Multi-Query GSM-Symbolic, we can decrease the average cost per query by 2.5x. We then conduct additional analysis to understand when sleep-time compute is most effective, finding the predictability of the user query to be well correlated with the efficacy of sleep-time compute. Finally, we conduct a case-study of applying sleep-time compute to a realistic agentic SWE task.", "authors": ["Kevin Lin", "Charlie Snell", "Yu Wang", "Charles Packer", "Sarah Wooders", "Ion Stoica", "Joseph E. Gonzalez"], "categories": ["cs.AI", "cs.CL"], "fields_of_study": ["Computer Science"], "published_date": "2025-04-17", "url": "https://arxiv.org/abs/2504.13171", "pdf_url": "https://arxiv.org/pdf/2504.13171v1", "arxiv_id": "2504.13171", "doi": "10.48550/arXiv.2504.13171", "citation_count": 31, "influential_citation_count": 6, "has_code": true, "code_url": "https://github.com/letta-ai/sleep-time-compute", "venue": "arXiv.org", "quality_score": 0.4225}
{"id": "d16cc43c25272539b8b1deecf0d922697f8e80c0cca60fd802cd1a2658f4cf17", "sources": ["arxiv", "semantic_scholar"], "title": "M1: Towards Scalable Test-Time Compute with Mamba Reasoning Models", "abstract": "Effective reasoning is crucial to solving complex mathematical problems. Recent large language models (LLMs) have boosted performance by scaling test-time computation through long chain-of-thought reasoning. However, transformer-based models are inherently limited in extending context length due to their quadratic computational complexity and linear memory requirements. In this paper, we introduce a novel hybrid linear RNN reasoning model, M1, built on the Mamba architecture, which allows memory-efficient inference. Our approach leverages a distillation process from existing reasoning models and is further enhanced through RL training. Experimental results on the AIME and MATH benchmarks show that M1 not only outperforms previous linear RNN models but also matches the performance of state-of-the-art Deepseek R1 distilled reasoning models at a similar scale. We also compare our generation speed with a highly performant general purpose inference engine, vLLM, and observe more than a 3x speedup compared to a same size transformer. With throughput speedup, we are able to achieve higher accuracy compared to DeepSeek R1 distilled transformer reasoning models under a fixed generation time budget using self-consistency voting. Overall, we introduce a hybrid Mamba reasoning model and provide a more effective approach to scaling test-time generation using self-consistency or long chain of thought reasoning.", "authors": ["Junxiong Wang", "Wen-Ding Li", "Daniele Paliotta", "Daniel Ritter", "Alexander M. Rush", "Tri Dao"], "categories": ["cs.LG"], "fields_of_study": ["Computer Science"], "published_date": "2025-04-14", "url": "https://arxiv.org/abs/2504.10449", "pdf_url": "https://arxiv.org/pdf/2504.10449v3", "arxiv_id": "2504.10449", "doi": "10.48550/arXiv.2504.10449", "citation_count": 20, "influential_citation_count": 1, "has_code": true, "code_url": "https://github.com/jxiw/M1", "venue": "arXiv.org", "quality_score": 0.3306}
{"id": "9e532367aedc38e4a3f8721a43afe3b965b0e7f11075d5de5e681db56275a5f2", "sources": ["arxiv", "semantic_scholar"], "title": "Beyond Chains of Thought: Benchmarking Latent-Space Reasoning Abilities in Large Language Models", "abstract": "Large language models (LLMs) can perform reasoning computations both internally within their latent space and externally by generating explicit token sequences like chains of thought. Significant progress in enhancing reasoning abilities has been made by scaling test-time compute. However, understanding and quantifying model-internal reasoning abilities - the inferential \"leaps\" models make between individual token predictions - remains crucial. This study introduces a benchmark (n = 4,000 items) designed to quantify model-internal reasoning in different domains. We achieve this by having LLMs indicate the correct solution to reasoning problems not through descriptive text, but by selecting a specific language of their initial response token that is different from English, the benchmark language. This not only requires models to reason beyond their context window, but also to overrise their default tendency to respond in the same language as the prompt, thereby posing an additional cognitive strain. We evaluate a set of 18 LLMs, showing significant performance variations, with GPT-4.5 achieving the highest accuracy (74.7%), outperforming models like Grok-2 (67.2%), and Llama 3.1 405B (65.6%). Control experiments and difficulty scaling analyses suggest that while LLMs engage in internal reasoning, we cannot rule out heuristic exploitations under certain conditions, marking an area for future investigation. Our experiments demonstrate that LLMs can \"think\" via latent-space computations, revealing model-internal inference strategies that need further understanding, especially regarding safety-related concerns such as covert planning, goal-seeking, or deception emerging without explicit token traces.", "authors": ["Thilo Hagendorff", "Sarah Fabi"], "categories": ["cs.CL"], "fields_of_study": ["Computer Science"], "published_date": "2025-04-14", "url": "https://arxiv.org/abs/2504.10615", "pdf_url": "https://arxiv.org/pdf/2504.10615v1", "arxiv_id": "2504.10615", "doi": "10.48550/arXiv.2504.10615", "citation_count": 1, "influential_citation_count": 0, "has_code": false, "code_url": null, "venue": "arXiv.org", "quality_score": 0.1249}
{"id": "1792d10cbdf711c47d677c52532ffdf6e7b1c0f551972561156d287ecedf90b7", "sources": ["arxiv", "semantic_scholar"], "title": "SpecReason: Fast and Accurate Inference-Time Compute via Speculative Reasoning", "abstract": "Recent advances in inference-time compute have significantly improved performance on complex tasks by generating long chains of thought (CoTs) using Large Reasoning Models (LRMs). However, this improved accuracy comes at the cost of high inference latency due to the length of generated reasoning sequences and the autoregressive nature of decoding. Our key insight in tackling these overheads is that LRM inference, and the reasoning that it embeds, is highly tolerant of approximations: complex tasks are typically broken down into simpler steps, each of which brings utility based on the semantic insight it provides for downstream steps rather than the exact tokens it generates. Accordingly, we introduce SpecReason, a system that automatically accelerates LRM inference by using a lightweight model to (speculatively) carry out simpler intermediate reasoning steps and reserving the costly base model only to assess (and potentially correct) the speculated outputs. Importantly, SpecReason's focus on exploiting the semantic flexibility of thinking tokens in preserving final-answer accuracy is complementary to prior speculation techniques, most notably speculative decoding, which demands token-level equivalence at each step. Across a variety of reasoning benchmarks, SpecReason achieves $1.4-3.0\\times$ speedup over vanilla LRM inference while improving accuracy by $0.4-9.0\\%$. Compared to speculative decoding without SpecReason, their combination yields an additional $8.8-58.0\\%$ latency reduction. We open-source SpecReason at https://github.com/ruipeterpan/specreason.", "authors": ["Rui Pan", "Yinwei Dai", "Zhihao Zhang", "Gabriele Oliaro", "Zhihao Jia", "Ravi Netravali"], "categories": ["cs.LG", "cs.AI"], "fields_of_study": ["Computer Science"], "published_date": "2025-04-10", "url": "https://arxiv.org/abs/2504.07891", "pdf_url": "https://arxiv.org/pdf/2504.07891v2", "arxiv_id": "2504.07891", "doi": "10.48550/arXiv.2504.07891", "citation_count": 57, "influential_citation_count": 12, "has_code": true, "code_url": "https://github.com/ruipeterpan/specreason", "venue": "arXiv.org", "quality_score": 0.557}
{"id": "b21c6ebca22813cb440aadf691fdf1aa37ceadda3ae308e231d194081d63159a", "sources": ["arxiv", "semantic_scholar"], "title": "VCR-Bench: A Comprehensive Evaluation Framework for Video Chain-of-Thought Reasoning", "abstract": "The advancement of Chain-of-Thought (CoT) reasoning has significantly enhanced the capabilities of large language models (LLMs) and large vision-language models (LVLMs). However, a rigorous evaluation framework for video CoT reasoning remains absent. Current video benchmarks fail to adequately assess the reasoning process and expose whether failures stem from deficiencies in perception or reasoning capabilities. Therefore, we introduce VCR-Bench, a novel benchmark designed to comprehensively evaluate LVLMs' Video Chain-of-Thought Reasoning capabilities. VCR-Bench comprises 859 videos spanning a variety of video content and durations, along with 1,034 high-quality question-answer pairs. Each pair is manually annotated with a stepwise CoT rationale, where every step is tagged to indicate its association with the perception or reasoning capabilities. Furthermore, we design seven distinct task dimensions and propose the CoT score to assess the entire CoT process based on the stepwise tagged CoT rationals. Extensive experiments on VCR-Bench highlight substantial limitations in current LVLMs. Even the top-performing model, o1, only achieves a 62.8% CoT score and an 56.7% accuracy, while most models score below 40%. Experiments show most models score lower on perception than reasoning steps, revealing LVLMs' key bottleneck in temporal-spatial information processing for complex video reasoning. A robust positive correlation between the CoT score and accuracy confirms the validity of our evaluation framework and underscores the critical role of CoT reasoning in solving complex video reasoning tasks. We hope VCR-Bench to serve as a standardized evaluation framework and expose the actual drawbacks in complex video reasoning task.", "authors": ["Yukun Qi", "Yiming Zhao", "Yu Zeng", "Xikun Bao", "Wenxuan Huang", "Lin Chen", "Zehui Chen", "Jie Zhao", "Zhongang Qi", "Feng Zhao"], "categories": ["cs.CV", "cs.AI", "cs.CL"], "fields_of_study": ["Computer Science"], "published_date": "2025-04-10", "url": "https://arxiv.org/abs/2504.07956", "pdf_url": "https://arxiv.org/pdf/2504.07956v1", "arxiv_id": "2504.07956", "doi": "10.48550/arXiv.2504.07956", "citation_count": 35, "influential_citation_count": 3, "has_code": false, "code_url": null, "venue": "arXiv.org", "quality_score": 0.3891}
{"id": "c7a01a1f272326b4d54664b06929a82f9db3fc0a97b919d38e955407be8ff328", "sources": ["arxiv", "semantic_scholar"], "title": "Skywork R1V: Pioneering Multimodal Reasoning with Chain-of-Thought", "abstract": "We introduce Skywork R1V, a multimodal reasoning model extending the an R1-series Large language models (LLM) to visual modalities via an efficient multimodal transfer method. Leveraging a lightweight visual projector, Skywork R1V facilitates seamless multimodal adaptation without necessitating retraining of either the foundational language model or the vision encoder. To strengthen visual-text alignment, we propose a hybrid optimization strategy that combines Iterative Supervised Fine-Tuning (SFT) with Group Relative Policy Optimization (GRPO), significantly enhancing cross-modal integration efficiency. Additionally, we introduce an adaptive-length Chain-of-Thought distillation approach for reasoning data generation. This approach dynamically optimizes reasoning chain lengths, thereby enhancing inference efficiency and preventing excessive reasoning overthinking. Empirical evaluations demonstrate that Skywork R1V, with only 38B parameters, delivers competitive performance, achieving a score of 69.0 on the MMMU benchmark and 67.5 on MathVista. Meanwhile, it maintains robust textual reasoning performance, evidenced by impressive scores of 72.0 on AIME and 94.0 on MATH500. The Skywork R1V model weights have been publicly released to promote openness and reproducibility.", "authors": ["Yi Peng", "Peiyu Wang", "Xiaokun Wang", "Yichen Wei", "Jiangbo Pei", "Weijie Qiu", "Ai Jian", "Yunzhuo Hao", "Jiachun Pan", "Tianyidan Xie", "Li Ge", "Rongxian Zhuang", "Xuchen Song", "Yang Liu", "Yahui Zhou"], "categories": ["cs.CV", "cs.CL"], "fields_of_study": ["Computer Science"], "published_date": "2025-04-08", "url": "https://arxiv.org/abs/2504.05599", "pdf_url": "https://arxiv.org/pdf/2504.05599v2", "arxiv_id": "2504.05599", "doi": "10.48550/arXiv.2504.05599", "citation_count": 67, "influential_citation_count": 2, "has_code": false, "code_url": null, "venue": "arXiv.org", "quality_score": 0.4581}
{"id": "fe044815989bbe1ff3b2ecc2305693fb21c7d5f52eb90ed1d8945e7c07acc800", "sources": ["arxiv", "semantic_scholar"], "title": "Adaptive Rectification Sampling for Test-Time Compute Scaling", "abstract": "The newly released OpenAI-o1 and DeepSeek-R1 have demonstrated that test-time scaling can significantly improve model performance, especially in complex tasks such as logical reasoning. Common test-time scaling methods involve generating more chains of thought (CoTs) or longer CoTs with self-correction. However, while self-correction can improve performance, it may lead to significant token waste and reduce readability of the CoT if the reasoning steps are already correct. To demonstrate that large language models (LLMs) can rectify errors at a more fine-grained level, we propose Adaptive Rectification Sampling (AR-Sampling), which can guide the LLMs to self-correction at the appropriate step. AR-Sampling leverages a process-supervised reward model (PRM) as a verifier and constructed trigger sentences to guide the model in adaptive step-level rethinking. Through the experiments on GSM8K and MATH500, it indicates that our approach enables the models to rethink in more fine-grained level, improving the accuracy of solutions, while generating a reasonable number of additional tokens.", "authors": ["Zhendong Tan", "Xingjun Zhang", "Chaoyi Hu", "Yancheng Pan", "Shaoxun Wang"], "categories": ["cs.CL", "cs.AI"], "fields_of_study": ["Computer Science"], "published_date": "2025-04-02", "url": "https://arxiv.org/abs/2504.01317", "pdf_url": "https://arxiv.org/pdf/2504.01317v2", "arxiv_id": "2504.01317", "doi": "10.48550/arXiv.2504.01317", "citation_count": 4, "influential_citation_count": 0, "has_code": false, "code_url": null, "venue": "arXiv.org", "quality_score": 0.1747}
{"id": "c8ff94f105052a4f2937bee2c223f4e44d4c8855c4f8a059e2251d176653662d", "sources": ["arxiv", "semantic_scholar"], "title": "GenPRM: Scaling Test-Time Compute of Process Reward Models via Generative Reasoning", "abstract": "Recent advancements in Large Language Models (LLMs) have shown that it is promising to utilize Process Reward Models (PRMs) as verifiers to enhance the performance of LLMs. However, current PRMs face three key challenges: (1) limited process supervision and generalization capabilities, (2) dependence on scalar value prediction without leveraging the generative abilities of LLMs, and (3) inability to scale the test-time compute of PRMs. In this work, we introduce GenPRM, a generative process reward model that performs explicit Chain-of-Thought (CoT) reasoning with code verification before providing judgment for each reasoning step. To obtain high-quality process supervision labels and rationale data, we propose Relative Progress Estimation (RPE) and a rationale synthesis framework that incorporates code verification. Experimental results on ProcessBench and several mathematical reasoning tasks show that GenPRM significantly outperforms prior PRMs with only 23K training data from MATH dataset. Through test-time scaling, a 1.5B GenPRM outperforms GPT-4o, and a 7B GenPRM surpasses Qwen2.5-Math-PRM-72B on ProcessBench. Additionally, GenPRM demonstrates strong abilities to serve as a critic model for policy model refinement. This work establishes a new paradigm for process supervision that bridges the gap between PRMs and critic models in LLMs. Our code, model, and data will be available in https://ryanliu112.github.io/GenPRM.", "authors": ["Jian Zhao", "Runze Liu", "Kaiyan Zhang", "Zhimu Zhou", "Junqi Gao", "Dong Li", "Jiafei Lyu", "Zhouyi Qian", "Biqing Qi", "Xiu Li", "Bowen Zhou"], "categories": ["cs.CL"], "fields_of_study": ["Computer Science"], "published_date": "2025-04-01", "url": "https://arxiv.org/abs/2504.00891", "pdf_url": "https://arxiv.org/pdf/2504.00891v2", "arxiv_id": "2504.00891", "doi": "10.48550/arXiv.2504.00891", "citation_count": 79, "influential_citation_count": 9, "has_code": false, "code_url": null, "venue": "AAAI Conference on Artificial Intelligence", "quality_score": 0.5}
{"id": "e5c95f30c3856ac99b616387f30b625f6b20bf34716e1f27cd957bd56352c1d7", "sources": ["arxiv", "semantic_scholar"], "title": "m1: Unleash the Potential of Test-Time Scaling for Medical Reasoning with Large Language Models", "abstract": "Test-time scaling has emerged as a powerful technique for enhancing the reasoning capabilities of large language models. However, its effectiveness in medical reasoning remains uncertain, as the medical domain fundamentally differs from mathematical tasks in terms of knowledge representation and decision-making processes. In this paper, we provide the first comprehensive investigation of test-time scaling for medical reasoning and present m1, a simple yet effective approach that increases a model's medical reasoning capability at inference. Our evaluation across diverse medical tasks demonstrates that test-time scaling consistently enhances medical reasoning, enabling lightweight fine-tuned models under 10B parameters to establish new state-of-the-art performance, while our 32B model rivals previous 70B-scale medical LLMs. However, we identify an optimal reasoning token budget of approximately 4K, beyond which performance may degrade due to overthinking. Budget forcing, which extends test-time computation through iterative prompts, helps models double-check answers but does not necessarily improve the overall medical QA performance and, in some cases, even introduces errors into previously correct responses. Our case-by-case analysis identifies insufficient medical knowledge as a key bottleneck that prevents further performance gains through test-time scaling. We find that increasing data scale, improving data quality, and expanding model capacity consistently enhance medical knowledge grounding, enabling continued performance improvements, particularly on challenging medical benchmarks where smaller models reach saturation. These findings underscore fundamental differences between medical and mathematical reasoning in LLMs, highlighting that enriched medical knowledge, other than increased reasoning depth alone, is essential for realizing the benefits of test-time scaling.", "authors": ["Xiaoke Huang", "Juncheng Wu", "Hui Liu", "Xianfeng Tang", "Yuyin Zhou"], "categories": ["cs.CL", "cs.AI"], "fields_of_study": ["Computer Science"], "published_date": "2025-04-01", "url": "https://arxiv.org/abs/2504.00869", "pdf_url": "https://arxiv.org/pdf/2504.00869v2", "arxiv_id": "2504.00869", "doi": "10.48550/arXiv.2504.00869", "citation_count": 37, "influential_citation_count": 10, "has_code": true, "code_url": "https://github.com/UCSC-VLAA/m1", "venue": "arXiv.org", "quality_score": 0.5207}
{"id": "317983ee0259f81c1813f71755efab595b7dbd870d367b9aaed0d7b37f8457f0", "sources": ["arxiv", "semantic_scholar"], "title": "Harnessing the Reasoning Economy: A Survey of Efficient Reasoning for Large Language Models", "abstract": "Recent advancements in Large Language Models (LLMs) have significantly enhanced their ability to perform complex reasoning tasks, transitioning from fast and intuitive thinking (System 1) to slow and deep reasoning (System 2). While System 2 reasoning improves task accuracy, it often incurs substantial computational costs due to its slow thinking nature and inefficient or unnecessary reasoning behaviors. In contrast, System 1 reasoning is computationally efficient but leads to suboptimal performance. Consequently, it is critical to balance the trade-off between performance (benefits) and computational costs (budgets), giving rise to the concept of reasoning economy. In this survey, we provide a comprehensive analysis of reasoning economy in both the post-training and test-time inference stages of LLMs, encompassing i) the cause of reasoning inefficiency, ii) behavior analysis of different reasoning patterns, and iii) potential solutions to achieve reasoning economy. By offering actionable insights and highlighting open challenges, we aim to shed light on strategies for improving the reasoning economy of LLMs, thereby serving as a valuable resource for advancing research in this evolving area. We also provide a public repository to continually track developments in this fast-evolving field.", "authors": ["Rui Wang", "Hongru Wang", "Boyang Xue", "Jianhui Pang", "Shudong Liu", "Yi Chen", "Jiahao Qiu", "Derek Fai Wong", "Heng Ji", "Kam-Fai Wong"], "categories": ["cs.CL", "cs.AI"], "fields_of_study": ["Computer Science"], "published_date": "2025-03-31", "url": "https://arxiv.org/abs/2503.24377", "pdf_url": "https://arxiv.org/pdf/2503.24377v1", "arxiv_id": "2503.24377", "doi": "10.48550/arXiv.2503.24377", "citation_count": 26, "influential_citation_count": 0, "has_code": true, "code_url": "https://github.com/DevoAllen/Awesome-Reasoning-Economy-Papers", "venue": "arXiv.org", "quality_score": 0.3578}
{"id": "f59dd54caf907cda3d13ebd5f5e42958aca801d2c26166c009c772198b6f26c6", "sources": ["arxiv", "semantic_scholar"], "title": "Inference-Time Scaling for Flow Models via Stochastic Generation and Rollover Budget Forcing", "abstract": "We propose an inference-time scaling approach for pretrained flow models. Recently, inference-time scaling has gained significant attention in LLMs and diffusion models, improving sample quality or better aligning outputs with user preferences by leveraging additional computation. For diffusion models, particle sampling has allowed more efficient scaling due to the stochasticity at intermediate denoising steps. On the contrary, while flow models have gained popularity as an alternative to diffusion models--offering faster generation and high-quality outputs in state-of-the-art image and video generative models--efficient inference-time scaling methods used for diffusion models cannot be directly applied due to their deterministic generative process. To enable efficient inference-time scaling for flow models, we propose three key ideas: 1) SDE-based generation, enabling particle sampling in flow models, 2) Interpolant conversion, broadening the search space and enhancing sample diversity, and 3) Rollover Budget Forcing (RBF), an adaptive allocation of computational resources across timesteps to maximize budget utilization. Our experiments show that SDE-based generation, particularly variance-preserving (VP) interpolant-based generation, improves the performance of particle sampling methods for inference-time scaling in flow models. Additionally, we demonstrate that RBF with VP-SDE achieves the best performance, outperforming all previous inference-time scaling approaches.", "authors": ["Jaihoon Kim", "Taehoon Yoon", "Jisung Hwang", "Minhyuk Sung"], "categories": ["cs.CV", "cs.LG"], "fields_of_study": ["Computer Science"], "published_date": "2025-03-25", "url": "https://arxiv.org/abs/2503.19385", "pdf_url": "https://arxiv.org/pdf/2503.19385v5", "arxiv_id": "2503.19385", "doi": "10.48550/arXiv.2503.19385", "citation_count": 37, "influential_citation_count": 7, "has_code": false, "code_url": null, "venue": "arXiv.org", "quality_score": 0.4515}
{"id": "407e2449ed8ecc93c127fdce0061fb5fdfd653d8313170797796b84bf58e674c", "sources": ["arxiv", "semantic_scholar"], "title": "Scaling Evaluation-time Compute with Reasoning Models as Evaluators", "abstract": "As language model (LM) outputs get more and more natural, it is becoming more difficult than ever to evaluate their quality. Simultaneously, increasing LMs' \"thinking\" time through scaling test-time compute has proven an effective technique to solve challenging problems in domains such as math and code. This raises a natural question: can an LM's evaluation capability also be improved by spending more test-time compute? To answer this, we investigate employing reasoning models-LMs that natively generate long chain-of-thought reasoning-as evaluators. Specifically, we examine methods to leverage more test-time compute by (1) using reasoning models, and (2) prompting these models to evaluate not only the response as a whole (i.e., outcome evaluation) but also assess each step in the response separately (i.e., process evaluation). In experiments, we observe that the evaluator's performance improves monotonically when generating more reasoning tokens, similar to the trends observed in LM-based generation. Furthermore, we use these more accurate evaluators to rerank multiple generations, and demonstrate that spending more compute at evaluation time can be as effective as using more compute at generation time in improving an LM's problem-solving capability.", "authors": ["Seungone Kim", "Ian Wu", "Jinu Lee", "Xiang Yue", "Seongyun Lee", "Mingyeong Moon", "Carolin Lawrence", "Kiril Gashteovski", "Julia Hockenmaier", "Graham Neubig", "Sean Welleck"], "categories": ["cs.CL"], "fields_of_study": ["Computer Science"], "published_date": "2025-03-25", "url": "https://arxiv.org/abs/2503.19877", "pdf_url": "https://arxiv.org/pdf/2503.19877v2", "arxiv_id": "2503.19877", "doi": null, "citation_count": 0, "influential_citation_count": 0, "has_code": false, "code_url": null, "venue": null, "quality_score": 0.0649}
{"id": "de885a5e7100d7ab608a2f306978d17d8d15d7f2a827e6d7745ffaac0283e63d", "sources": ["arxiv", "semantic_scholar"], "title": "ImageGen-CoT: Enhancing Text-to-Image In-context Learning with Chain-of-Thought Reasoning", "abstract": "In this work, we study the problem of Text-to-Image In-Context Learning (T2I-ICL). While Unified Multimodal LLMs (MLLMs) have advanced rapidly in recent years, they struggle with contextual reasoning in T2I-ICL scenarios. To address this limitation, we propose a novel framework that incorporates a thought process called ImageGen-CoT prior to image generation. To avoid generating unstructured ineffective reasoning steps, we develop an automatic pipeline to curate a high-quality ImageGen-CoT dataset. We then fine-tune MLLMs using this dataset to enhance their contextual reasoning capabilities. To further enhance performance, we explore test-time scale-up strategies and propose a novel hybrid scaling approach. This approach first generates multiple ImageGen-CoT chains and then produces multiple images for each chain via sampling. Extensive experiments demonstrate the effectiveness of our proposed method. Notably, fine-tuning with the ImageGen-CoT dataset leads to a substantial 80\\% performance gain for SEED-X on T2I-ICL tasks. See our project page at https://ImageGen-CoT.github.io/. Code and model weights will be open-sourced.", "authors": ["Jiaqi Liao", "Zhengyuan Yang", "Linjie Li", "Dianqi Li", "Kevin Lin", "Yu Cheng", "Lijuan Wang"], "categories": ["cs.CV"], "fields_of_study": ["Computer Science"], "published_date": "2025-03-25", "url": "https://arxiv.org/abs/2503.19312", "pdf_url": "https://arxiv.org/pdf/2503.19312v1", "arxiv_id": "2503.19312", "doi": "10.1109/ICCV51701.2025.01599", "citation_count": 35, "influential_citation_count": 2, "has_code": true, "code_url": null, "venue": "IEEE International Conference on Computer Vision", "quality_score": 0.3891}
{"id": "8fd48692ec80325187b3064c552d8f351dc5591e75b1f1eb0d0b6e18c94e2ef5", "sources": ["arxiv", "semantic_scholar"], "title": "Reasoning to Learn from Latent Thoughts", "abstract": "Compute scaling for language model (LM) pretraining has outpaced the growth of human-written texts, leading to concerns that data will become the bottleneck to LM scaling. To continue scaling pretraining in this data-constrained regime, we propose that explicitly modeling and inferring the \\emph{latent thoughts} that underlie the text generation process can significantly improve pretraining data efficiency. Intuitively, our approach views web text as the compressed final outcome of a verbose human thought process and that the latent thoughts contain important contextual knowledge and reasoning steps that are critical to data-efficient learning. We empirically demonstrate the effectiveness of our approach through data-constrained continued pretraining for math. We first show that synthetic data approaches to inferring latent thoughts significantly improve data efficiency over training on the same amount of raw data. Furthermore, we demonstrate latent thought inference without a strong teacher, where an LM \\emph{bootstraps its own performance} by using an EM algorithm to iteratively improve the capability of the trained LM and the quality of thought-augmented pretraining data. We show that a 1B LM can bootstrap its performance across at least three iterations and significantly outperform baselines trained on raw data, with increasing gains from additional inference compute when performing the E-step. The gains from inference scaling and EM iterations suggest new opportunities for scaling data-constrained pretraining.", "authors": ["Yangjun Ruan", "Neil Band", "Chris J. Maddison", "Tatsunori Hashimoto"], "categories": ["cs.LG", "cs.AI", "cs.CL"], "fields_of_study": ["Computer Science"], "published_date": "2025-03-24", "url": "https://arxiv.org/abs/2503.18866", "pdf_url": "https://arxiv.org/pdf/2503.18866v2", "arxiv_id": "2503.18866", "doi": "10.48550/arXiv.2503.18866", "citation_count": 36, "influential_citation_count": 3, "has_code": false, "code_url": null, "venue": "arXiv.org", "quality_score": 0.3921}
{"id": "f22c6f0a44ec948c460e95287e85d47dfad96e909dae6c9e9b4eb638e0ea84fc", "sources": ["arxiv", "semantic_scholar"], "title": "Evaluating Test-Time Scaling LLMs for Legal Reasoning: OpenAI o1, DeepSeek-R1, and Beyond", "abstract": "Recent advances in test-time scaling of large language models (LLMs), exemplified by DeepSeek-R1 and OpenAI's o1, show that extending the chain of thought during inference can significantly improve general reasoning performance. However, the impact of this paradigm on legal reasoning remains insufficiently explored. To address this gap, we present the first systematic evaluation of 12 LLMs, including both reasoning-focused and general-purpose models, across 17 Chinese and English legal tasks spanning statutory and case-law traditions. In addition, we curate a bilingual chain-of-thought dataset for legal reasoning through distillation from DeepSeek-R1 and develop Legal-R1, an open-source model specialized for the legal domain. Experimental results show that Legal-R1 delivers competitive performance across diverse tasks. DeepSeek-R1 exhibits clear advantages in Chinese legal reasoning, while OpenAI's o1 achieves comparable results on English tasks. We further conduct a detailed error analysis, which reveals recurring issues such as outdated legal knowledge, limited capacity for legal interpretation, and susceptibility to factual hallucinations. These findings delineate the main obstacles confronting legal-domain LLMs and suggest promising directions for future research.", "authors": ["Yinghao Hu", "Yaoyao Yu", "Leilei Gan", "Bin Wei", "Kun Kuang", "Fei Wu"], "categories": ["cs.CL"], "fields_of_study": ["Computer Science"], "published_date": "2025-03-20", "url": "https://arxiv.org/abs/2503.16040", "pdf_url": "https://arxiv.org/pdf/2503.16040v2", "arxiv_id": "2503.16040", "doi": "10.18653/v1/2025.findings-emnlp.742", "citation_count": 13, "influential_citation_count": 0, "has_code": true, "code_url": null, "venue": "Conference on Empirical Methods in Natural Language Processing", "quality_score": 0.2865}
{"id": "968e244dcac126d1cdf1eaf94876666345fe3e0d873a2f0782a79f949a4d3ff2", "sources": ["arxiv", "semantic_scholar"], "title": "Stop Overthinking: A Survey on Efficient Reasoning for Large Language Models", "abstract": "Large Language Models (LLMs) have demonstrated remarkable capabilities in complex tasks. Recent advancements in Large Reasoning Models (LRMs), such as OpenAI o1 and DeepSeek-R1, have further improved performance in System-2 reasoning domains like mathematics and programming by harnessing supervised fine-tuning (SFT) and reinforcement learning (RL) techniques to enhance the Chain-of-Thought (CoT) reasoning. However, while longer CoT reasoning sequences improve performance, they also introduce significant computational overhead due to verbose and redundant outputs, known as the \"overthinking phenomenon\". In this paper, we provide the first structured survey to systematically investigate and explore the current progress toward achieving efficient reasoning in LLMs. Overall, relying on the inherent mechanism of LLMs, we categorize existing works into several key directions: (1) model-based efficient reasoning, which considers optimizing full-length reasoning models into more concise reasoning models or directly training efficient reasoning models; (2) reasoning output-based efficient reasoning, which aims to dynamically reduce reasoning steps and length during inference; (3) input prompts-based efficient reasoning, which seeks to enhance reasoning efficiency based on input prompt properties such as difficulty or length control. Additionally, we introduce the use of efficient data for training reasoning models, explore the reasoning capabilities of small language models, and discuss evaluation methods and benchmarking. Project website: https://github.com/Eclipsess/Awesome-Efficient-Reasoning-LLMs", "authors": ["Yang Sui", "Yu-Neng Chuang", "Guanchu Wang", "Jiamu Zhang", "Tianyi Zhang", "Jiayi Yuan", "Hongyi Liu", "Andrew Wen", "Shaochen Zhong", "Na Zou", "Hanjie Chen", "Xia Hu"], "categories": ["cs.CL"], "fields_of_study": ["Computer Science"], "published_date": "2025-03-20", "url": "https://arxiv.org/abs/2503.16419", "pdf_url": "https://arxiv.org/pdf/2503.16419v4", "arxiv_id": "2503.16419", "doi": "10.48550/arXiv.2503.16419", "citation_count": 400, "influential_citation_count": 22, "has_code": true, "code_url": "https://github.com/Eclipsess/Awesome-Efficient-Reasoning-LLMs", "venue": null, "quality_score": 0.6809}
{"id": "158b1da1402f87d99cddf0e271f6f7e17869210a025892b3334921a009aaf4c9", "sources": ["arxiv", "semantic_scholar"], "title": "Deconstructing Long Chain-of-Thought: A Structured Reasoning Optimization Framework for Long CoT Distillation", "abstract": "Recent advancements in large language models (LLMs) have demonstrated remarkable reasoning capabilities through long chain-of-thought (CoT) reasoning. The R1 distillation scheme has emerged as a promising approach for training cost-effective models with enhanced reasoning abilities. However, the underlying mechanisms driving its effectiveness remain unclear. This study examines the universality of distillation data and identifies key components that enable the efficient transfer of long-chain reasoning capabilities in LLM distillation. Our findings reveal that the effectiveness of long CoT reasoning distillation from teacher models like Qwen-QwQ degrades significantly on nonhomologous models, challenging the assumed universality of current distillation methods. To gain deeper insights into the structure and patterns of long CoT reasoning, we propose DLCoT (Deconstructing Long Chain-of-Thought), a distillation data enhancement framework. DLCoT consists of three key steps: (1) data segmentation to decompose complex long CoT structures, (2) simplification by eliminating unsolvable and redundant solutions, and (3) optimization of intermediate error states. Our approach significantly improves model performance and token efficiency, facilitating the development of high-performance LLMs.", "authors": ["Yijia Luo", "Yulin Song", "Xingyao Zhang", "Jiaheng Liu", "Weixun Wang", "GengRu Chen", "Wenbo Su", "Bo Zheng"], "categories": ["cs.AI"], "fields_of_study": ["Computer Science"], "published_date": "2025-03-20", "url": "https://arxiv.org/abs/2503.16385", "pdf_url": "https://arxiv.org/pdf/2503.16385v1", "arxiv_id": "2503.16385", "doi": "10.48550/arXiv.2503.16385", "citation_count": 36, "influential_citation_count": 1, "has_code": false, "code_url": null, "venue": "arXiv.org", "quality_score": 0.3921}
{"id": "643c5c0218622e87e90bf8226011bb97b4af15d57b51c4867b1240cd2ef67b99", "sources": ["arxiv", "semantic_scholar"], "title": "Reasoning Effort and Problem Complexity: A Scaling Analysis in LLMs", "abstract": "Large Language Models (LLMs) have demonstrated remarkable text generation capabilities, and recent advances in training paradigms have led to breakthroughs in their reasoning performance. In this work, we investigate how the reasoning effort of such models scales with problem complexity. We use the infinitely scalable Tents puzzle, which has a known linear-time solution, to analyze this scaling behavior. Our results show that reasoning effort scales with problem size, but only up to a critical problem complexity. Beyond this threshold, the reasoning effort does not continue to increase, and may even decrease. This observation highlights a critical limitation in the logical coherence of current LLMs as problem complexity increases, and underscores the need for strategies to improve reasoning scalability. Furthermore, our results reveal significant performance differences between current state-of-the-art reasoning models when faced with increasingly complex logical puzzles.", "authors": ["Benjamin Estermann", "Roger Wattenhofer"], "categories": ["cs.AI"], "fields_of_study": ["Computer Science"], "published_date": "2025-03-19", "url": "https://arxiv.org/abs/2503.15113", "pdf_url": "https://arxiv.org/pdf/2503.15113v1", "arxiv_id": "2503.15113", "doi": "10.48550/arXiv.2503.15113", "citation_count": 2, "influential_citation_count": 0, "has_code": false, "code_url": null, "venue": "arXiv.org", "quality_score": 0.1193}
{"id": "8f0c3c51dd9cac7a8dd53df1ccf97267ada67471adf1d75906cd24ce32736c55", "sources": ["arxiv", "semantic_scholar"], "title": "MetaScale: Test-Time Scaling with Evolving Meta-Thoughts", "abstract": "One critical challenge for large language models (LLMs) for making complex reasoning is their reliance on matching reasoning patterns from training data, instead of proactively selecting the most appropriate cognitive strategy to solve a given task. Existing approaches impose fixed cognitive structures that enhance performance in specific tasks but lack adaptability across diverse scenarios. To address this limitation, we introduce METASCALE, a test-time scaling framework based on meta-thoughts -- adaptive thinking strategies tailored to each task. METASCALE initializes a pool of candidate meta-thoughts, then iteratively selects and evaluates them using a multi-armed bandit algorithm with upper confidence bound selection, guided by a reward model. To further enhance adaptability, a genetic algorithm evolves high-reward meta-thoughts, refining and extending the strategy pool over time. By dynamically proposing and optimizing meta-thoughts at inference time, METASCALE improves both accuracy and generalization across a wide range of tasks. Experimental results demonstrate that MetaScale consistently outperforms standard inference approaches, achieving an 11% performance gain in win rate on Arena-Hard for GPT-4o, surpassing o1-mini by 0.9% under style control. Notably, METASCALE scales more effectively with increasing sampling budgets and produces more structured, expert-level responses.", "authors": ["Qin Liu", "Wenxuan Zhou", "Nan Xu", "James Y. Huang", "Fei Wang", "Sheng Zhang", "Hoifung Poon", "Muhao Chen"], "categories": ["cs.CL", "cs.AI", "cs.LG"], "fields_of_study": ["Computer Science"], "published_date": "2025-03-17", "url": "https://arxiv.org/abs/2503.13447", "pdf_url": "https://arxiv.org/pdf/2503.13447v1", "arxiv_id": "2503.13447", "doi": "10.48550/arXiv.2503.13447", "citation_count": 10, "influential_citation_count": 0, "has_code": false, "code_url": null, "venue": "arXiv.org", "quality_score": 0.2603}
{"id": "2f3d67bd0ca8cc8fdb975d7c014d9c367172cb908dd215190eac34ea9c89eccf", "sources": ["arxiv", "semantic_scholar"], "title": "Multimodal Chain-of-Thought Reasoning: A Comprehensive Survey", "abstract": "By extending the advantage of chain-of-thought (CoT) reasoning in human-like step-by-step processes to multimodal contexts, multimodal CoT (MCoT) reasoning has recently garnered significant research attention, especially in the integration with multimodal large language models (MLLMs). Existing MCoT studies design various methodologies and innovative reasoning paradigms to address the unique challenges of image, video, speech, audio, 3D, and structured data across different modalities, achieving extensive success in applications such as robotics, healthcare, autonomous driving, and multimodal generation. However, MCoT still presents distinct challenges and opportunities that require further focus to ensure consistent thriving in this field, where, unfortunately, an up-to-date review of this domain is lacking. To bridge this gap, we present the first systematic survey of MCoT reasoning, elucidating the relevant foundational concepts and definitions. We offer a comprehensive taxonomy and an in-depth analysis of current methodologies from diverse perspectives across various application scenarios. Furthermore, we provide insights into existing challenges and future research directions, aiming to foster innovation toward multimodal AGI.", "authors": ["Yaoting Wang", "Shengqiong Wu", "Yuecheng Zhang", "Shuicheng Yan", "Ziwei Liu", "Jiebo Luo", "Hao Fei"], "categories": ["cs.CV"], "fields_of_study": ["Computer Science"], "published_date": "2025-03-16", "url": "https://arxiv.org/abs/2503.12605", "pdf_url": "https://arxiv.org/pdf/2503.12605v2", "arxiv_id": "2503.12605", "doi": "10.48550/arXiv.2503.12605", "citation_count": 184, "influential_citation_count": 6, "has_code": true, "code_url": "https://github.com/yaotingwangofficial/Awesome-MCoT", "venue": "arXiv.org", "quality_score": 0.5668}
{"id": "26eb20dd99cfaf8717f626a7bf0c95f03ee6d583cb1586c04bebceae78a5652f", "sources": ["arxiv", "semantic_scholar"], "title": "V-STaR: Benchmarking Video-LLMs on Video Spatio-Temporal Reasoning", "abstract": "Human processes video reasoning in a sequential spatio-temporal reasoning logic, we first identify the relevant frames (\"when\") and then analyse the spatial relationships (\"where\") between key objects, and finally leverage these relationships to draw inferences (\"what\"). However, can Video Large Language Models (Video-LLMs) also \"reason through a sequential spatio-temporal logic\" in videos? Existing Video-LLM benchmarks primarily focus on assessing object presence, neglecting relational reasoning. Consequently, it is difficult to measure whether a model truly comprehends object interactions (actions/events) in videos or merely relies on pre-trained \"memory\" of co-occurrences as biases in generating answers. In this work, we introduce a Video Spatio-Temporal Reasoning (V-STaR) benchmark to address these shortcomings. The key idea is to decompose video understanding into a Reverse Spatio-Temporal Reasoning (RSTR) task that simultaneously evaluates what objects are present, when events occur, and where they are located while capturing the underlying Chain-of-thought (CoT) logic. To support this evaluation, we construct a dataset to elicit the spatial-temporal reasoning process of Video-LLMs. It contains coarse-to-fine CoT questions generated by a semi-automated GPT-4-powered pipeline, embedding explicit reasoning chains to mimic human cognition. Experiments from 14 Video-LLMs on our V-STaR reveal significant gaps between current Video-LLMs and the needs for robust and consistent spatio-temporal reasoning.", "authors": ["Zixu Cheng", "Jian Hu", "Ziquan Liu", "Chenyang Si", "Wei Li", "Shaogang Gong"], "categories": ["cs.CV"], "fields_of_study": ["Computer Science"], "published_date": "2025-03-14", "url": "https://arxiv.org/abs/2503.11495", "pdf_url": "https://arxiv.org/pdf/2503.11495v1", "arxiv_id": "2503.11495", "doi": "10.48550/arXiv.2503.11495", "citation_count": 57, "influential_citation_count": 5, "has_code": false, "code_url": null, "venue": "arXiv.org", "quality_score": 0.4409}
{"id": "bc25d21dd318c1b10444512f06a6a26ff2f0b8173e1529cc06a21fa03a09c84b", "sources": ["arxiv", "semantic_scholar"], "title": "Towards Reasoning Era: A Survey of Long Chain-of-Thought for Reasoning Large Language Models", "abstract": "Recent advancements in reasoning with large language models (RLLMs), such as OpenAI-O1 and DeepSeek-R1, have demonstrated their impressive capabilities in complex domains like mathematics and coding. A central factor in their success lies in the application of long chain-of-thought (Long CoT) characteristics, which enhance reasoning abilities and enable the solution of intricate problems. However, despite these developments, a comprehensive survey on Long CoT is still lacking, limiting our understanding of its distinctions from traditional short chain-of-thought (Short CoT) and complicating ongoing debates on issues like \"overthinking\" and \"inference-time scaling.\" This survey seeks to fill this gap by offering a unified perspective on Long CoT. (1) We first distinguish Long CoT from Short CoT and introduce a novel taxonomy to categorize current reasoning paradigms. (2) Next, we explore the key characteristics of Long CoT: deep reasoning, extensive exploration, and feasible reflection, which enable models to handle more complex tasks and produce more efficient, coherent outcomes compared to the shallower Short CoT. (3) We then investigate key phenomena such as the emergence of Long CoT with these characteristics, including overthinking, and inference-time scaling, offering insights into how these processes manifest in practice. (4) Finally, we identify significant research gaps and highlight promising future directions, including the integration of multi-modal reasoning, efficiency improvements, and enhanced knowledge frameworks. By providing a structured overview, this survey aims to inspire future research and further the development of logical reasoning in artificial intelligence.", "authors": ["Qiguang Chen", "Libo Qin", "Jinhao Liu", "Dengyun Peng", "Jiannan Guan", "Peng Wang", "Mengkang Hu", "Yuhang Zhou", "Te Gao", "Wanxiang Che"], "categories": ["cs.AI", "cs.CL"], "fields_of_study": ["Computer Science"], "published_date": "2025-03-12", "url": "https://arxiv.org/abs/2503.09567", "pdf_url": "https://arxiv.org/pdf/2503.09567v5", "arxiv_id": "2503.09567", "doi": "10.48550/arXiv.2503.09567", "citation_count": 322, "influential_citation_count": 18, "has_code": true, "code_url": "https://github.com/LightChen233/Awesome-Long-Chain-of-Thought-Reasoning", "venue": "arXiv.org", "quality_score": 0.6394}
{"id": "656cdeb5d282779c22783b2cc66988e1b26209c39c3489f8261cc19765265254", "sources": ["arxiv", "semantic_scholar"], "title": "Investigating the Effectiveness of a Socratic Chain-of-Thoughts Reasoning Method for Task Planning in Robotics, A Case Study", "abstract": "Large language models (LLMs) have demonstrated unprecedented capability in reasoning with natural language. Coupled with this development is the emergence of embodied AI in robotics. Despite showing promise for verbal and written reasoning tasks, it remains unknown whether LLMs are capable of navigating complex spatial tasks with physical actions in the real world. To this end, it is of interest to investigate applying LLMs to robotics in zero-shot learning scenarios, and in the absence of fine-tuning - a feat which could significantly improve human-robot interaction, alleviate compute cost, and eliminate low-level programming tasks associated with robot tasks. To explore this question, we apply GPT-4(Omni) with a simulated Tiago robot in Webots engine for an object search task. We evaluate the effectiveness of three reasoning strategies based on Chain-of-Thought (CoT) sub-task list generation with the Socratic method (SocraCoT) (in order of increasing rigor): (1) Non-CoT/Non-SocraCoT, (2) CoT only, and (3) SocraCoT. Performance was measured in terms of the proportion of tasks successfully completed and execution time (N = 20). Our preliminary results show that when combined with chain-of-thought reasoning, the Socratic method can be used for code generation for robotic tasks that require spatial awareness. In extension of this finding, we propose EVINCE-LoC; a modified EVINCE method that could further enhance performance in highly complex and or dynamic testing scenarios.", "authors": ["Veronica Bot", "Zheyuan Xu"], "categories": ["cs.RO", "cs.AI"], "fields_of_study": ["Computer Science"], "published_date": "2025-03-11", "url": "https://arxiv.org/abs/2503.08174", "pdf_url": "https://arxiv.org/pdf/2503.08174v1", "arxiv_id": "2503.08174", "doi": "10.48550/arXiv.2503.08174", "citation_count": 0, "influential_citation_count": 0, "has_code": false, "code_url": null, "venue": "arXiv.org", "quality_score": 0.0859}
{"id": "fe4543872eb5b316f85216c66d17a043eb04f1c0782b9fb19cffa7a7fe66d5cc", "sources": ["arxiv", "semantic_scholar"], "title": "A Theory of Learning with Autoregressive Chain of Thought", "abstract": "For a given base class of sequence-to-next-token generators, we consider learning prompt-to-answer mappings obtained by iterating a fixed, time-invariant generator for multiple steps, thus generating a chain-of-thought, and then taking the final token as the answer. We formalize the learning problems both when the chain-of-thought is observed and when training only on prompt-answer pairs, with the chain-of-thought latent. We analyze the sample and computational complexity both in terms of general properties of the base class (e.g. its VC dimension) and for specific base classes such as linear thresholds. We present a simple base class that allows for universal representability and computationally tractable chain-of-thought learning. Central to our development is that time invariance allows for sample complexity that is independent of the length of the chain-of-thought. Attention arises naturally in our construction.", "authors": ["Nirmit Joshi", "Gal Vardi", "Adam Block", "Surbhi Goel", "Zhiyuan Li", "Theodor Misiakiewicz", "Nathan Srebro"], "categories": ["stat.ML", "cs.AI", "cs.CC", "cs.LG"], "fields_of_study": ["Mathematics", "Computer Science"], "published_date": "2025-03-11", "url": "https://arxiv.org/abs/2503.07932", "pdf_url": "https://arxiv.org/pdf/2503.07932v2", "arxiv_id": "2503.07932", "doi": "10.48550/arXiv.2503.07932", "citation_count": 19, "influential_citation_count": 5, "has_code": false, "code_url": null, "venue": "Annual Conference Computational Learning Theory", "quality_score": 0.3891}
{"id": "55493c73006ed795ac11777552d06f06da71de232c36046fd3e2dd3cd3241e9a", "sources": ["arxiv", "semantic_scholar"], "title": "Think Before You Segment: High-Quality Reasoning Segmentation with GPT Chain of Thoughts", "abstract": "Reasoning segmentation is a challenging vision-language task that aims to output the segmentation mask with respect to a complex, implicit, and even non-visual query text. Previous works incorporated multimodal Large Language Models (MLLMs) with segmentation models to approach the difficult problem. However, their segmentation quality often falls short in complex cases, particularly when dealing with out-of-domain objects with intricate structures, blurry boundaries, occlusions, or high similarity with surroundings. In this paper, we introduce ThinkFirst, a training-free reasoning segmentation framework that leverages GPT's chain of thought to address these challenging cases. Our approach allows GPT-4o or other powerful MLLMs to generate a detailed, chain-of-thought description of an image. This summarized description is then passed to a language-instructed segmentation assistant to aid the segmentation process. Our framework allows users to easily interact with the segmentation agent using multimodal inputs, such as easy text and image scribbles, for successive refinement or communication. We evaluate the performance of ThinkFirst on diverse objects. Extensive experiments show that, this zero-shot-CoT approach significantly improves the vanilla reasoning segmentation agent, both qualitatively and quantitatively, while being less sensitive or critical to user-supplied prompts after Thinking First.", "authors": ["Shiu-hong Kao", "Yu-Wing Tai", "Chi-Keung Tang"], "categories": ["cs.CV"], "fields_of_study": ["Computer Science"], "published_date": "2025-03-10", "url": "https://arxiv.org/abs/2503.07503", "pdf_url": "https://arxiv.org/pdf/2503.07503v4", "arxiv_id": "2503.07503", "doi": "10.48550/arXiv.2503.07503", "citation_count": 4, "influential_citation_count": 0, "has_code": false, "code_url": null, "venue": "arXiv.org", "quality_score": 0.1747}
{"id": "307308cff374e6ca229b01c611d3f8bf0c09e1064b15b86b73641ca3de00bc0d", "sources": ["arxiv", "semantic_scholar"], "title": "Compositional Reasoning with Transformers, RNNs, and Chain of Thought", "abstract": "It is well understood that different neural network architectures are suited to different tasks, but is there always a single best architecture for a given task? We compare the expressive power of transformers, RNNs, and transformers with chain of thought tokens on a simple and natural class of tasks we term Compositional Reasoning Questions (CRQ). This family captures multi-step problems with tree-like compositional structure, such as evaluating Boolean formulas. We prove that under standard hardness assumptions, \\emph{none} of these three architectures is capable of solving CRQs unless some hyperparameter (depth, embedding dimension, and number of chain of thought tokens, respectively) grows with the size of the input. We then provide constructions for solving CRQs with each architecture. For transformers, our construction uses depth that is logarithmic in the problem size. For RNNs, logarithmic embedding dimension is necessary and sufficient, so long as the inputs are provided in a certain order. For transformers with chain of thought, our construction uses $n$ CoT tokens for input size $n$. These results show that, while CRQs are inherently hard, there are several different ways for language models to overcome this hardness. Even for a single class of problems, each architecture has strengths and weaknesses, and none is strictly better than the others.", "authors": ["Gilad Yehudai", "Noah Amsel", "Joan Bruna"], "categories": ["cs.LG", "cs.AI", "cs.CL"], "fields_of_study": ["Computer Science"], "published_date": "2025-03-03", "url": "https://arxiv.org/abs/2503.01544", "pdf_url": "https://arxiv.org/pdf/2503.01544v2", "arxiv_id": "2503.01544", "doi": "10.48550/arXiv.2503.01544", "citation_count": 4, "influential_citation_count": 0, "has_code": false, "code_url": null, "venue": "arXiv.org", "quality_score": 0.1747}
{"id": "79f0303a1311bfb5a16a64f34bedfbed951c2f68dbedb5da41876ec05358f9b3", "sources": ["arxiv", "semantic_scholar"], "title": "ARIES: Autonomous Reasoning with LLMs on Interactive Thought Graph Environments", "abstract": "Recent research has shown that LLM performance on reasoning tasks can be enhanced by scaling test-time compute. One promising approach, particularly with decomposable problems, involves arranging intermediate solutions as a graph on which transformations are performed to explore the solution space. However, prior works rely on pre-determined, task-specific transformation schedules which are subject to a set of searched hyperparameters. In this work, we view thought graph transformations as actions in a Markov decision process, and implement policy agents to drive effective action policies for the underlying reasoning LLM agent. In particular, we investigate the ability for another LLM to act as a policy agent on thought graph environments and introduce ARIES, a multi-agent architecture for reasoning with LLMs. In ARIES, reasoning LLM agents solve decomposed subproblems, while policy LLM agents maintain visibility of the thought graph states, and dynamically adapt the problem-solving strategy. Through extensive experiments, we observe that using off-the-shelf LLMs as policy agents with no supervised fine-tuning (SFT) can yield up to $29\\%$ higher accuracy on HumanEval relative to static transformation schedules, as well as reducing inference costs by $35\\%$ and avoid any search requirements. We also conduct a thorough analysis of observed failure modes, highlighting that limitations on LLM sizes and the depth of problem decomposition can be seen as challenges to scaling LLM-guided reasoning.", "authors": ["Pedro Gimenes", "Zeyu Cao", "Jeffrey Wong", "Yiren Zhao"], "categories": ["cs.AI", "cs.LG"], "fields_of_study": ["Computer Science"], "published_date": "2025-02-28", "url": "https://arxiv.org/abs/2502.21208", "pdf_url": "https://arxiv.org/pdf/2502.21208v1", "arxiv_id": "2502.21208", "doi": "10.48550/arXiv.2502.21208", "citation_count": 0, "influential_citation_count": 0, "has_code": false, "code_url": null, "venue": "arXiv.org", "quality_score": 0.0733}
{"id": "79f7ce9bfc65264f3ef55ea9c00694bce81252a4bbcc7513c5f59eb19d62dccb", "sources": ["arxiv", "semantic_scholar"], "title": "CoTMR: Chain-of-Thought Multi-Scale Reasoning for Training-Free Zero-Shot Composed Image Retrieval", "abstract": "Zero-Shot Composed Image Retrieval (ZS-CIR) aims to retrieve target images by integrating information from a composed query (reference image and modification text) without training samples. Existing methods primarily combine caption models and large language models (LLMs) to generate target captions based on composed queries but face various issues such as incompatibility, visual information loss, and insufficient reasoning. In this work, we propose CoTMR, a training-free framework crafted for ZS-CIR with novel Chain-of-thought (CoT) and Multi-scale Reasoning. Instead of relying on caption models for modality transformation, CoTMR employs the Large Vision-Language Model (LVLM) to achieve unified understanding and reasoning for composed queries. To enhance the reasoning reliability, we devise CIRCoT, which guides the LVLM through a step-by-step inference process using predefined subtasks. Considering that existing approaches focus solely on global-level reasoning, our CoTMR incorporates multi-scale reasoning to achieve more comprehensive inference via fine-grained predictions about the presence or absence of key elements at the object scale. Further, we design a Multi-Grained Scoring (MGS) mechanism, which integrates CLIP similarity scores of the above reasoning outputs with candidate images to realize precise retrieval. Extensive experiments demonstrate that our CoTMR not only drastically outperforms previous methods across four prominent benchmarks but also offers appealing interpretability.", "authors": ["Zelong Sun", "Dong Jing", "Zhiwu Lu"], "categories": ["cs.CV", "cs.IR"], "fields_of_study": ["Computer Science"], "published_date": "2025-02-28", "url": "https://arxiv.org/abs/2502.20826", "pdf_url": "https://arxiv.org/pdf/2502.20826v1", "arxiv_id": "2502.20826", "doi": "10.1109/ICCV51701.2025.02105", "citation_count": 10, "influential_citation_count": 0, "has_code": false, "code_url": null, "venue": "IEEE International Conference on Computer Vision", "quality_score": 0.2603}
{"id": "cbb882266bebf1ee0abafbc4899a2dccf3fbe176addb3f2f61f6446d581a08f4", "sources": ["arxiv", "semantic_scholar"], "title": "Thinking Slow, Fast: Scaling Inference Compute with Distilled Reasoners", "abstract": "Recent advancements have demonstrated that the performance of large language models (LLMs) can be significantly enhanced by scaling computational resources at test time. A common strategy involves generating multiple Chain-of-Thought (CoT) trajectories and aggregating their outputs through various selection mechanisms. This raises a fundamental question: can models with lower complexity leverage their superior generation throughput to outperform similarly sized Transformers for a fixed computational budget? To address this question and overcome the lack of strong subquadratic reasoners, we distill pure and hybrid Mamba models from pretrained Transformers. Trained on only 8 billion tokens, our distilled models show strong performance and scaling on mathematical reasoning datasets while being much faster at inference for large batches and long sequences. Despite the zero-shot performance hit due to distillation, both pure and hybrid Mamba models can scale their coverage and accuracy performance past their Transformer teacher models under fixed time budgets, opening a new direction for scaling inference compute.", "authors": ["Daniele Paliotta", "Junxiong Wang", "Matteo Pagliardini", "Kevin Y. Li", "Aviv Bick", "J. Zico Kolter", "Albert Gu", "François Fleuret", "Tri Dao"], "categories": ["cs.CL", "cs.AI"], "fields_of_study": ["Computer Science"], "published_date": "2025-02-27", "url": "https://arxiv.org/abs/2502.20339", "pdf_url": "https://arxiv.org/pdf/2502.20339v1", "arxiv_id": "2502.20339", "doi": "10.48550/arXiv.2502.20339", "citation_count": 23, "influential_citation_count": 1, "has_code": false, "code_url": null, "venue": "arXiv.org", "quality_score": 0.3451}
{"id": "83fa466e68540cf6c925034ae67a6d892060d10b871c173bdc7f7f26742a32ec", "sources": ["arxiv", "semantic_scholar"], "title": "Meta-Reasoner: Dynamic Guidance for Optimized Inference-time Reasoning in Large Language Models", "abstract": "Large Language Models (LLMs) often struggle with computational efficiency and error propagation in multi-step reasoning tasks. While recent advancements on prompting and post-training have enabled LLMs to perform step-wise reasoning, they still tend to explore unproductive solution paths without effective backtracking or strategy adjustment. In this paper, we propose Meta-Reasoner, a new framework that empowers LLMs to \"think about how to think\". It optimizes the inference process by dynamically adapting reasoning strategies in real-time. Our approach employs contextual multi-armed bandits (CMABs) to learn an adaptive policy. It learns to evaluate the current state of LLM's reasoning and determine optimal strategy that is most likely to lead to a successful outcome during inference, like whether to backtrack, switch to a new approach, or restart the problem-solving process. This meta-guidance helps avoid unproductive paths exploration during inference and hence improves computational efficiency. We evaluate Meta-Reasoner on math problems (e.g., Game-of-24, TheoremQA) and scientific tasks (e.g., SciBench). Results show that our method outperform previous SOTA methods by 9-12% in accuracy, while reducing inference time by 28-35% under the same compute budget. Additional experiments on creative writing demonstrate the generalizability of our approach to diverse reasoning-intensive tasks.", "authors": ["Yuan Sui", "Yufei He", "Tri Cao", "Simeng Han", "Yulin Chen", "Bryan Hooi"], "categories": ["cs.AI", "cs.LG"], "fields_of_study": ["Computer Science"], "published_date": "2025-02-27", "url": "https://arxiv.org/abs/2502.19918", "pdf_url": "https://arxiv.org/pdf/2502.19918v6", "arxiv_id": "2502.19918", "doi": "10.48550/arXiv.2502.19918", "citation_count": 28, "influential_citation_count": 1, "has_code": false, "code_url": null, "venue": "arXiv.org", "quality_score": 0.3656}
{"id": "b702d8c1eaad7c4e98f671f88e699d333d2f1f302020db3b4451f3d341b34bd1", "sources": ["arxiv", "semantic_scholar"], "title": "Random Forest-of-Thoughts: Uncertainty-aware Reasoning for Computational Social Science", "abstract": "Social surveys in computational social science are well-designed by elaborate domain theories that can effectively reflect the interviewee's deep thoughts without concealing their true feelings. The candidate questionnaire options highly depend on the interviewee's previous answer, which results in the complexity of social survey analysis, the time, and the expertise required. The ability of large language models (LLMs) to perform complex reasoning is well-enhanced by prompting learning such as Chain-of-thought (CoT) but still confined to left-to-right decision-making processes or limited paths during inference. This means they can fall short in problems that require exploration and uncertainty searching. In response, a novel large language model prompting method, called Random Forest of Thoughts (RFoT), is proposed for generating uncertainty reasoning to fit the area of computational social science. The RFoT allows LLMs to perform deliberate decision-making by generating diverse thought space and randomly selecting the sub-thoughts to build the forest of thoughts. It can extend the exploration and prediction of overall performance, benefiting from the extensive research space of response. The method is applied to optimize computational social science analysis on two datasets covering a spectrum of social survey analysis problems. Our experiments show that RFoT significantly enhances language models' abilities on two novel social survey analysis problems requiring non-trivial reasoning.", "authors": ["Xiaohua Wu", "Xiaohui Tao", "Wenjie Wu", "Yuefeng Li", "Lin Li"], "categories": ["cs.CL"], "fields_of_study": ["Computer Science"], "published_date": "2025-02-26", "url": "https://arxiv.org/abs/2502.18729", "pdf_url": "https://arxiv.org/pdf/2502.18729v1", "arxiv_id": "2502.18729", "doi": "10.48550/arXiv.2502.18729", "citation_count": 0, "influential_citation_count": 0, "has_code": false, "code_url": null, "venue": "arXiv.org", "quality_score": 0.071}
{"id": "d6a638e2675476d58e814957b45fe9e342e34e87d16a8cac1220e4a954a15b5a", "sources": ["arxiv", "semantic_scholar"], "title": "Towards Thinking-Optimal Scaling of Test-Time Compute for LLM Reasoning", "abstract": "Recent studies have shown that making a model spend more time thinking through longer Chain of Thoughts (CoTs) enables it to gain significant improvements in complex reasoning tasks. While current researches continue to explore the benefits of increasing test-time compute by extending the CoT lengths of Large Language Models (LLMs), we are concerned about a potential issue hidden behind the current pursuit of test-time scaling: Would excessively scaling the CoT length actually bring adverse effects to a model's reasoning performance? Our explorations on mathematical reasoning tasks reveal an unexpected finding that scaling with longer CoTs can indeed impair the reasoning performance of LLMs in certain domains. Moreover, we discover that there exists an optimal scaled length distribution that differs across different domains. Based on these insights, we propose a Thinking-Optimal Scaling strategy. Our method first uses a small set of seed data with varying response length distributions to teach the model to adopt different reasoning efforts for deep thinking. Then, the model selects its shortest correct response under different reasoning efforts on additional problems for self-improvement. Our self-improved models built upon Qwen2.5-32B-Instruct outperform other distillation-based 32B o1-like models across various math benchmarks, and achieve performance on par with the teacher model QwQ-32B-Preview that produces the seed data.", "authors": ["Wenkai Yang", "Shuming Ma", "Yankai Lin", "Furu Wei"], "categories": ["cs.CL", "cs.AI"], "fields_of_study": ["Computer Science"], "published_date": "2025-02-25", "url": "https://arxiv.org/abs/2502.18080", "pdf_url": "https://arxiv.org/pdf/2502.18080v2", "arxiv_id": "2502.18080", "doi": "10.48550/arXiv.2502.18080", "citation_count": 127, "influential_citation_count": 4, "has_code": false, "code_url": null, "venue": "arXiv.org", "quality_score": 0.5268}
{"id": "b7abb301d5c97a5a0eb540b1def7179f25c656fea1c288e78d30bbc881c8e30b", "sources": ["arxiv", "semantic_scholar"], "title": "Linguistic Generalizability of Test-Time Scaling in Mathematical Reasoning", "abstract": "Scaling pre-training compute has proven effective for achieving mulitlinguality, but does the same hold for test-time scaling? In this work, we introduce MCLM, a multilingual math benchmark featuring competition-level problems in 55 languages. We test three test-time scaling methods-Outcome Reward Modeling (ORM), Process Reward Modeling (ORM), and Budget Forcing (BF)-on both Qwen2.5-1.5B Math and MR1-1.5B, a multilingual LLM we trained for extended reasoning. Our experiments show that using Qwen2.5-1.5B Math with ORM achieves a score of 35.8 on MCLM, while BF on MR1-1.5B attains 35.2. Although \"thinking LLMs\" have recently garnered significant attention, we find that their performance is comparable to traditional scaling methods like best-of-N once constrained to similar levels of inference FLOPs. Moreover, while BF yields a 20-point improvement on English AIME, it provides only a 1.94-point average gain across other languages-a pattern consistent across the other test-time scaling methods we studied-higlighting that test-time scaling may not generalize as effectively to multilingual tasks. To foster further research, we release MCLM, MR1-1.5B, and evaluation results.", "authors": ["Guijin Son", "Jiwoo Hong", "Hyunwoo Ko", "James Thorne"], "categories": ["cs.CL"], "fields_of_study": ["Computer Science"], "published_date": "2025-02-24", "url": "https://arxiv.org/abs/2502.17407", "pdf_url": "https://arxiv.org/pdf/2502.17407v2", "arxiv_id": "2502.17407", "doi": "10.48550/arXiv.2502.17407", "citation_count": 38, "influential_citation_count": 2, "has_code": false, "code_url": null, "venue": "Annual Meeting of the Association for Computational Linguistics", "quality_score": 0.3978}
{"id": "85402373d21b0335aa3108dd707cb48419f96f0c1bda7a745319e537a9c13d05", "sources": ["arxiv", "semantic_scholar"], "title": "Measuring Chain of Thought Faithfulness by Unlearning Reasoning Steps", "abstract": "When prompted to think step-by-step, language models (LMs) produce a chain of thought (CoT), a sequence of reasoning steps that the model supposedly used to produce its prediction. Despite much work on CoT prompting, it is unclear if reasoning verbalized in a CoT is faithful to the models' parametric beliefs. We introduce a framework for measuring parametric faithfulness of generated reasoning, and propose Faithfulness by Unlearning Reasoning steps (FUR), an instance of this framework. FUR erases information contained in reasoning steps from model parameters, and measures faithfulness as the resulting effect on the model's prediction. Our experiments with four LMs and five multi-hop multi-choice question answering (MCQA) datasets show that FUR is frequently able to precisely change the underlying models' prediction for a given instance by unlearning key steps, indicating when a CoT is parametrically faithful. Further analysis shows that CoTs generated by models post-unlearning support different answers, hinting at a deeper effect of unlearning.", "authors": ["Martin Tutek", "Fateme Hashemi Chaleshtori", "Ana Marasović", "Yonatan Belinkov"], "categories": ["cs.CL"], "fields_of_study": ["Computer Science"], "published_date": "2025-02-20", "url": "https://arxiv.org/abs/2502.14829", "pdf_url": "https://arxiv.org/pdf/2502.14829v4", "arxiv_id": "2502.14829", "doi": "10.18653/v1/2025.emnlp-main.504", "citation_count": 47, "influential_citation_count": 4, "has_code": false, "code_url": null, "venue": "Conference on Empirical Methods in Natural Language Processing", "quality_score": 0.4203}
{"id": "e352b9f71892131c08b6c4a184ea2ecf4a37a5ea07a754aa2a64deb3f2eb0e8e", "sources": ["arxiv", "semantic_scholar"], "title": "MM-Verify: Enhancing Multimodal Reasoning with Chain-of-Thought Verification", "abstract": "According to the Test-Time Scaling, the integration of External Slow-Thinking with the Verify mechanism has been demonstrated to enhance multi-round reasoning in large language models (LLMs). However, in the multimodal (MM) domain, there is still a lack of a strong MM-Verifier. In this paper, we introduce MM-Verifier and MM-Reasoner to enhance multimodal reasoning through longer inference and more robust verification. First, we propose a two-step MM verification data synthesis method, which combines a simulation-based tree search with verification and uses rejection sampling to generate high-quality Chain-of-Thought (COT) data. This data is then used to fine-tune the verification model, MM-Verifier. Additionally, we present a more efficient method for synthesizing MMCOT data, bridging the gap between text-based and multimodal reasoning. The synthesized data is used to fine-tune MM-Reasoner. Our MM-Verifier outperforms all larger models on the MathCheck, MathVista, and MathVerse benchmarks. Moreover, MM-Reasoner demonstrates strong effectiveness and scalability, with performance improving as data size increases. Finally, our approach achieves strong performance when combining MM-Reasoner and MM-Verifier, reaching an accuracy of 65.3 on MathVista, surpassing GPT-4o (63.8) with 12 rollouts.", "authors": ["Linzhuang Sun", "Hao Liang", "Jingxuan Wei", "Bihui Yu", "Tianpeng Li", "Fan Yang", "Zenan Zhou", "Wentao Zhang"], "categories": ["cs.CL", "cs.CV", "cs.LG"], "fields_of_study": ["Computer Science"], "published_date": "2025-02-19", "url": "https://arxiv.org/abs/2502.13383", "pdf_url": "https://arxiv.org/pdf/2502.13383v1", "arxiv_id": "2502.13383", "doi": "10.48550/arXiv.2502.13383", "citation_count": 44, "influential_citation_count": 1, "has_code": false, "code_url": null, "venue": "Annual Meeting of the Association for Computational Linguistics", "quality_score": 0.4133}
{"id": "e2775293da7c5a2f75f03adb86ee15bd6e95f24223c8f48c6a21519a1334215f", "sources": ["arxiv", "semantic_scholar"], "title": "Inference-Time Computations for LLM Reasoning and Planning: A Benchmark and Insights", "abstract": "We examine the reasoning and planning capabilities of large language models (LLMs) in solving complex tasks. Recent advances in inference-time techniques demonstrate the potential to enhance LLM reasoning without additional training by exploring intermediate steps during inference. Notably, OpenAI's o1 model shows promising performance through its novel use of multi-step reasoning and verification. Here, we explore how scaling inference-time techniques can improve reasoning and planning, focusing on understanding the tradeoff between computational cost and performance. To this end, we construct a comprehensive benchmark, known as Sys2Bench, and perform extensive experiments evaluating existing inference-time techniques on eleven diverse tasks across five categories, including arithmetic reasoning, logical reasoning, common sense reasoning, algorithmic reasoning, and planning. Our findings indicate that simply scaling inference-time computation has limitations, as no single inference-time technique consistently performs well across all reasoning and planning tasks.", "authors": ["Shubham Parashar", "Blake Olson", "Sambhav Khurana", "Eric Li", "Hongyi Ling", "James Caverlee", "Shuiwang Ji"], "categories": ["cs.AI", "cs.LG"], "fields_of_study": ["Computer Science"], "published_date": "2025-02-18", "url": "https://arxiv.org/abs/2502.12521", "pdf_url": "https://arxiv.org/pdf/2502.12521v1", "arxiv_id": "2502.12521", "doi": "10.48550/arXiv.2502.12521", "citation_count": 31, "influential_citation_count": 2, "has_code": false, "code_url": null, "venue": "arXiv.org", "quality_score": 0.3763}
{"id": "8c2265a3c98fe81f48be0e2575e917eab416e91365dc91e20338cd1605cf3871", "sources": ["arxiv", "semantic_scholar"], "title": "Investigating Inference-time Scaling for Chain of Multi-modal Thought: A Preliminary Study", "abstract": "Recently, inference-time scaling of chain-of-thought (CoT) has been demonstrated as a promising approach for addressing multi-modal reasoning tasks. While existing studies have predominantly centered on text-based thinking, the integration of both visual and textual modalities within the reasoning process remains unexplored. In this study, we pioneer the exploration of inference-time scaling with multi-modal thought, aiming to bridge this gap. To provide a comprehensive analysis, we systematically investigate popular sampling-based and tree search-based inference-time scaling methods on 10 challenging tasks spanning various domains. Besides, we uniformly adopt a consistency-enhanced verifier to ensure effective guidance for both methods across different thought paradigms. Results show that multi-modal thought promotes better performance against conventional text-only thought, and blending the two types of thought fosters more diverse thinking. Despite these advantages, multi-modal thoughts necessitate higher token consumption for processing richer visual inputs, which raises concerns in practical applications. We hope that our findings on the merits and drawbacks of this research line will inspire future works in the field.", "authors": ["Yujie Lin", "Ante Wang", "Moye Chen", "Jingyao Liu", "Hao Liu", "Jinsong Su", "Xinyan Xiao"], "categories": ["cs.CL"], "fields_of_study": ["Computer Science"], "published_date": "2025-02-17", "url": "https://arxiv.org/abs/2502.11514", "pdf_url": "https://arxiv.org/pdf/2502.11514v2", "arxiv_id": "2502.11514", "doi": "10.48550/arXiv.2502.11514", "citation_count": 14, "influential_citation_count": 0, "has_code": false, "code_url": null, "venue": "Annual Meeting of the Association for Computational Linguistics", "quality_score": 0.294}
{"id": "739c94033bf507e09cd33ad5ac9bd0e1f74955e28171e8edb107bd55e911d940", "sources": ["arxiv", "semantic_scholar"], "title": "Atom of Thoughts for Markov LLM Test-Time Scaling", "abstract": "Large Language Models (LLMs) have achieved significant performance gains through test-time scaling methods. However, existing approaches often incur redundant computations due to the accumulation of historical dependency information during inference. To address this challenge, we leverage the memoryless property of Markov processes to minimize reliance on historical context and propose a Markovian reasoning process. This foundational Markov chain structure enables seamless integration with various test-time scaling methods, thereby improving their scaling efficiency. By further scaling up the Markovian reasoning chain through integration with techniques such as tree search and reflective refinement, we uncover an emergent atomic reasoning structure, where reasoning trajectories are decomposed into a series of self-contained, low-complexity atomic units. We name this design Atom of Thoughts (\\our). Extensive experiments demonstrate that \\our consistently outperforms existing baselines as computational budgets increase. Importantly, \\our integrates seamlessly with existing reasoning frameworks and different LLMs (both reasoning and non-reasoning), facilitating scalable, high-performance inference.We submit our code alongside this paper and will make it publicly available to facilitate reproducibility and future research.", "authors": ["Fengwei Teng", "Quan Shi", "Zhaoyang Yu", "Jiayi Zhang", "Yuyu Luo", "Chenglin Wu", "Zhijiang Guo"], "categories": ["cs.CL", "cs.AI", "cs.LG"], "fields_of_study": ["Computer Science"], "published_date": "2025-02-17", "url": "https://arxiv.org/abs/2502.12018", "pdf_url": "https://arxiv.org/pdf/2502.12018v4", "arxiv_id": "2502.12018", "doi": "10.48550/arXiv.2502.12018", "citation_count": 63, "influential_citation_count": 8, "has_code": false, "code_url": null, "venue": "arXiv.org", "quality_score": 0.4771}
{"id": "59beda5706153349cb3c7e750a9bb695d46279ba3c9efe099ae2dc6cdbf72f75", "sources": ["arxiv", "semantic_scholar"], "title": "SoftCoT: Soft Chain-of-Thought for Efficient Reasoning with LLMs", "abstract": "Chain-of-Thought (CoT) reasoning enables Large Language Models (LLMs) to solve complex reasoning tasks by generating intermediate reasoning steps. However, most existing approaches focus on hard token decoding, which constrains reasoning within the discrete vocabulary space and may not always be optimal. While recent efforts explore continuous-space reasoning, they often require full-model fine-tuning and suffer from catastrophic forgetting, limiting their applicability to state-of-the-art LLMs that already perform well in zero-shot settings with a proper instruction. To address this challenge, we propose a novel approach for continuous-space reasoning that does not require modifying the LLM. Specifically, we employ a lightweight fixed assistant model to speculatively generate instance-specific soft thought tokens as the initial chain of thoughts, which are then mapped into the LLM's representation space via a trainable projection module. Experimental results on five reasoning benchmarks demonstrate that our method enhances LLM reasoning performance through supervised, parameter-efficient fine-tuning. Source code is available at https://github.com/xuyige/SoftCoT.", "authors": ["Yige Xu", "Xu Guo", "Zhiwei Zeng", "Chunyan Miao"], "categories": ["cs.CL"], "fields_of_study": ["Computer Science"], "published_date": "2025-02-17", "url": "https://arxiv.org/abs/2502.12134", "pdf_url": "https://arxiv.org/pdf/2502.12134v2", "arxiv_id": "2502.12134", "doi": "10.48550/arXiv.2502.12134", "citation_count": 130, "influential_citation_count": 8, "has_code": true, "code_url": "https://github.com/xuyige/SoftCoT", "venue": "Annual Meeting of the Association for Computational Linguistics", "quality_score": 0.5293}
{"id": "201598d8acc1721aac1ee8cc57d852a3a963ca19ad7b2f0815230c1fb9f82d34", "sources": ["arxiv", "semantic_scholar"], "title": "Scaling Test-Time Compute Without Verification or RL is Suboptimal", "abstract": "Despite substantial advances in scaling test-time compute, an ongoing debate in the community is how it should be scaled up to enable continued and efficient improvements with scaling. There are largely two approaches: first, distilling successful search or thinking traces; and second, using verification (e.g., 0/1 outcome rewards, reward models, or verifiers) to guide reinforcement learning (RL) and search algorithms. In this paper, we prove that finetuning LLMs with verifier-based (VB) methods based on RL or search is far superior to verifier-free (VF) approaches based on distilling or cloning search traces, given a fixed amount of compute/data budget. Further, we show that as we scale test-time compute (measured as the output token length) and training data, suboptimality of VF methods scales poorly compared to VB when the base pre-trained LLM presents a heterogeneous distribution over correct solution traces (e.g., different lengths, styles, etc.) and admits a non-sharp distribution over rewards on traces sampled from it. We formalize this condition using anti-concentration [Erdős, 1945]. This implies a stronger result that VB methods scale better asymptotically, with the performance gap between VB and VF methods widening as test-time budget grows. We corroborate our theory empirically on both didactic and math reasoning problems with 3/8/32B-sized pre-trained LLMs, where we find verification is crucial for scaling test-time compute.", "authors": ["Amrith Setlur", "Nived Rajaraman", "Sergey Levine", "Aviral Kumar"], "categories": ["cs.LG", "cs.CL"], "fields_of_study": ["Computer Science"], "published_date": "2025-02-17", "url": "https://arxiv.org/abs/2502.12118", "pdf_url": "https://arxiv.org/pdf/2502.12118v2", "arxiv_id": "2502.12118", "doi": "10.48550/arXiv.2502.12118", "citation_count": 86, "influential_citation_count": 3, "has_code": false, "code_url": null, "venue": "International Conference on Machine Learning", "quality_score": 0.4849}
{"id": "fbfeec1182ee95ae0a4c2727106d4ed31ed4ce835f3cc338ccf7c876a9f0c54e", "sources": ["arxiv", "semantic_scholar"], "title": "SafeChain: Safety of Language Models with Long Chain-of-Thought Reasoning Capabilities", "abstract": "Emerging large reasoning models (LRMs), such as DeepSeek-R1 models, leverage long chain-of-thought (CoT) reasoning to generate structured intermediate steps, enhancing their reasoning capabilities. However, long CoT does not inherently guarantee safe outputs, potentially leading to harmful consequences such as the introduction of security vulnerabilities in code or the spread of misinformation. Current research on large language model (LLM) safety usually focuses on short-answer responses, overlooking the long CoT style outputs of LRMs. To bridge this gap, we conduct a systematic study of LRM safety. First, we investigate safety evaluators calibrated against human annotations. Using our newly developed metrics, we thoroughly assess the safety of 12 state-of-the-art LRMs on StrongReject and WildJailbreak datasets. Our results show that LRMs are not safe compared to their reasoning advance. Further, we perform a fine-grained analysis of the reasoning trace and final answer. We find that three decoding strategies-ZeroThink, LessThink, and MoreThink-can improve model safety without additional training. However, these strategies either use constrained reasoning traces or incur high inference costs. To better strengthen LRM safety, we introduce SafeChain, the first-of-its-kind safety training dataset in CoT style. We fine-tune two LRMs with SafeChain, showing that it not only enhances model safety but also preserves performance across 6 reasoning benchmarks.", "authors": ["Fengqing Jiang", "Zhangchen Xu", "Yuetai Li", "Luyao Niu", "Zhen Xiang", "Bo Li", "Bill Yuchen Lin", "Radha Poovendran"], "categories": ["cs.AI", "cs.CL"], "fields_of_study": ["Computer Science"], "published_date": "2025-02-17", "url": "https://arxiv.org/abs/2502.12025", "pdf_url": "https://arxiv.org/pdf/2502.12025v1", "arxiv_id": "2502.12025", "doi": "10.48550/arXiv.2502.12025", "citation_count": 119, "influential_citation_count": 27, "has_code": false, "code_url": null, "venue": "Annual Meeting of the Association for Computational Linguistics", "quality_score": 0.7236}
{"id": "48ec7b95833fa4371731844f8926fdcc58cc3e84949dda6c516ef8f653365a83", "sources": ["arxiv", "semantic_scholar"], "title": "Enhancing Auto-regressive Chain-of-Thought through Loop-Aligned Reasoning", "abstract": "Chain-of-Thought (CoT) prompting has emerged as a powerful technique for enhancing language model's reasoning capabilities. However, generating long and correct CoT trajectories is challenging. Recent studies have demonstrated that Looped Transformers possess remarkable length generalization capabilities, but their limited generality and adaptability prevent them from serving as an alternative to auto-regressive solutions. To better leverage the strengths of Looped Transformers, we propose RELAY (REasoning through Loop Alignment iterativelY). Specifically, we align the steps of Chain-of-Thought (CoT) reasoning with loop iterations and apply intermediate supervision during the training of Looped Transformers. This additional iteration-wise supervision not only preserves the Looped Transformer's ability for length generalization but also enables it to predict CoT reasoning steps for unseen data. Therefore, we leverage this Looped Transformer to generate accurate reasoning chains for complex problems that exceed the training length, which will then be used to fine-tune an auto-regressive model. We conduct extensive experiments, and the results demonstrate the effectiveness of our approach, with significant improvements in the performance of the auto-regressive model. Code will be released at https://github.com/qifanyu/RELAY.", "authors": ["Qifan Yu", "Zhenyu He", "Sijie Li", "Xun Zhou", "Jun Zhang", "Jingjing Xu", "Di He"], "categories": ["cs.CL", "cs.AI", "cs.LG"], "fields_of_study": ["Computer Science"], "published_date": "2025-02-12", "url": "https://arxiv.org/abs/2502.08482", "pdf_url": "https://arxiv.org/pdf/2502.08482v1", "arxiv_id": "2502.08482", "doi": "10.48550/arXiv.2502.08482", "citation_count": 18, "influential_citation_count": 2, "has_code": true, "code_url": "https://github.com/qifanyu/RELAY", "venue": "Conference of the European Chapter of the Association for Computational Linguistics", "quality_score": 0.3197}
{"id": "ee81148a737011178e5e82a0a001ed5bd0c59e7ae24fb168b83b74526c7371f6", "sources": ["arxiv", "semantic_scholar"], "title": "Rethinking Fine-Tuning when Scaling Test-Time Compute: Limiting Confidence Improves Mathematical Reasoning", "abstract": "Recent progress in large language models (LLMs) highlights the power of scaling test-time compute to achieve strong performance on complex tasks, such as mathematical reasoning and code generation. This raises a critical question: how should model training be modified to optimize performance under a subsequent test-time compute strategy and budget? To explore this, we focus on pass@N, a simple test-time strategy that searches for a correct answer in $N$ independent samples. We show, surprisingly, that training with cross-entropy (CE) loss can be ${\\it misaligned}$ with pass@N in that pass@N accuracy ${\\it decreases}$ with longer training. We explain the origins of this misalignment in terms of model overconfidence induced by CE, and experimentally verify our prediction of overconfidence as an impediment to scaling test-time compute via pass@N. Furthermore we suggest a principled, modified training loss that is better aligned to pass@N by limiting model confidence and rescuing pass@N test performance. Our algorithm demonstrates improved mathematical reasoning on MATH and MiniF2F benchmarks under several scenarios: (1) providing answers to math questions; and (2) proving theorems by searching over proof trees of varying shapes. Overall our work underscores the importance of co-designing two traditionally separate phases of LLM development: training-time protocols and test-time search and reasoning strategies.", "authors": ["Feng Chen", "Allan Raventos", "Nan Cheng", "Surya Ganguli", "Shaul Druckmann"], "categories": ["cs.LG", "cs.AI"], "fields_of_study": ["Computer Science"], "published_date": "2025-02-11", "url": "https://arxiv.org/abs/2502.07154", "pdf_url": "https://arxiv.org/pdf/2502.07154v4", "arxiv_id": "2502.07154", "doi": "10.48550/arXiv.2502.07154", "citation_count": 26, "influential_citation_count": 2, "has_code": false, "code_url": null, "venue": "arXiv.org", "quality_score": 0.3578}
{"id": "0d72a753f7fcb9376c345e3fc81458605d0db9a30a103ada6194ebbe088e0fec", "sources": ["arxiv", "semantic_scholar"], "title": "Bag of Tricks for Inference-time Computation of LLM Reasoning", "abstract": "With the advancement of large language models (LLMs), solving complex reasoning tasks has gained increasing attention. Inference-time computation methods (e.g., Best-of-N, beam search, et al.) are particularly valuable as they can enhance reasoning performance without modifying model parameters or requiring additional training. However, these techniques come with implementation challenges, and most existing methods remain at the proof-of-concept stage with limited practical adoption due to their computational complexity and varying effectiveness across different tasks. In this paper, we investigate and benchmark diverse inference-time computation strategies across reasoning tasks of varying complexity. Since most current methods rely on a proposer-verifier pipeline that first generates candidate solutions (e.g., reasoning solutions) and then selects the best one based on reward signals (e.g., RLHF rewards, process rewards), our research focuses on optimizing both candidate solution generation (e.g., instructing prompts, hyperparameters such as temperature and top-p) and reward mechanisms (e.g., self-evaluation, reward types). Through extensive experiments (more than 20,000 A100-80G GPU hours with over 1,000 experiments) across a variety of models (e.g., Llama, Qwen, and Mistral families) of various sizes, our ablation studies reveal that previously overlooked strategies can significantly enhance performance (e.g., tuning temperature can improve reasoning task performance by up to 5%). Furthermore, we establish a standardized benchmark for inference-time computation by systematically evaluating six representative methods across eight reasoning tasks. These findings provide a stronger foundation for future research. The code is available at https://github.com/usail-hkust/benchmark_inference_time_computation_LLM", "authors": ["Fan Liu", "Wenshuo Chao", "Naiqiang Tan", "Hao Liu"], "categories": ["cs.AI"], "fields_of_study": ["Computer Science"], "published_date": "2025-02-11", "url": "https://arxiv.org/abs/2502.07191", "pdf_url": "https://arxiv.org/pdf/2502.07191v4", "arxiv_id": "2502.07191", "doi": "10.48550/arXiv.2502.07191", "citation_count": 15, "influential_citation_count": 2, "has_code": true, "code_url": "https://github.com/usail-hkust/benchmark_inference_time_computation_LLM", "venue": "arXiv.org", "quality_score": 0.301}
{"id": "592f82f18873f62d8b02a00ccc3f83f29fd5352f0669f16eb8d50a1640992a80", "sources": ["arxiv", "semantic_scholar"], "title": "ReasonFlux: Hierarchical LLM Reasoning via Scaling Thought Templates", "abstract": "We present that hierarchical LLM reasoning via scaling thought templates can effectively optimize the reasoning search space and outperform the mathematical reasoning capabilities of powerful LLMs like OpenAI o1-preview and DeepSeek V3. We train our ReasonFlux-32B model with only 8 GPUs and introduces three innovations: (i) a structured and generic thought template library, containing around 500 high-level thought templates capable of generalizing to similar or relevant reasoning problems; (ii) performing hierarchical reinforcement learning on a sequence of thought templates instead of long CoTs, optimizing a base LLM to plan out an optimal template trajectory for gradually handling complex problems; (iii) a brand new inference scaling system that enables hierarchical LLM reasoning by adaptively scaling thought templates at inference time. With a template trajectory containing more explainable reasoning structures than DeepSeek-R1 and o3-mini, our ReasonFlux-32B significantly advances math reasoning capabilities to state-of-the-art levels. Notably, on the MATH benchmark, it achieves an accuracy of 91.2% and surpasses o1-preview by 6.7%. On the USA Math Olympiad (AIME) benchmark, ReasonFlux-32B solves an average of 56.7% of problems, surpassing o1-preview and DeepSeek-V3 by 27% and 45%, respectively. Code: https://github.com/Gen-Verse/ReasonFlux", "authors": ["Ling Yang", "Zhaochen Yu", "Bin Cui", "Mengdi Wang"], "categories": ["cs.CL", "cs.AI", "cs.LG"], "fields_of_study": ["Computer Science"], "published_date": "2025-02-10", "url": "https://arxiv.org/abs/2502.06772", "pdf_url": "https://arxiv.org/pdf/2502.06772v2", "arxiv_id": "2502.06772", "doi": "10.48550/arXiv.2502.06772", "citation_count": 61, "influential_citation_count": 2, "has_code": true, "code_url": "https://github.com/Gen-Verse/ReasonFlux", "venue": "arXiv.org", "quality_score": 0.4481}
{"id": "5fe2953fa45f7e0d70f768dde799dc1e2a2c85199a47504e6ea77bdf16a89a57", "sources": ["arxiv", "semantic_scholar"], "title": "Adaptive Graph of Thoughts: Test-Time Adaptive Reasoning Unifying Chain, Tree, and Graph Structures", "abstract": "Large Language Models (LLMs) have demonstrated impressive reasoning capabilities, yet their performance is highly dependent on the prompting strategy and model scale. While reinforcement learning and fine-tuning have been deployed to boost reasoning, these approaches incur substantial computational and data overhead. In this work, we introduce Adaptive Graph of Thoughts (AGoT), a dynamic, graph-based inference framework that enhances LLM reasoning solely at test time. Rather than relying on fixed-step methods like Chain of Thought (CoT) or Tree of Thoughts (ToT), AGoT recursively decomposes complex queries into structured subproblems, forming an dynamic directed acyclic graph (DAG) of interdependent reasoning steps. By selectively expanding only those subproblems that require further analysis, AGoT unifies the strengths of chain, tree, and graph paradigms into a cohesive framework that allocates computation where it is most needed. We validate our approach on diverse benchmarks spanning multi-hop retrieval, scientific reasoning, and mathematical problem-solving, achieving up to 46.2% improvement on scientific reasoning tasks (GPQA) - comparable to gains achieved through computationally intensive reinforcement learning approaches and outperforming state-of-the-art iterative approaches. These results suggest that dynamic decomposition and structured recursion offer a scalable, cost-effective alternative to post-training modifications, paving the way for more robust, general-purpose reasoning in LLMs.", "authors": ["Tushar Pandey", "Ara Ghukasyan", "Oktay Goktas", "Santosh Kumar Radha"], "categories": ["cs.AI", "cs.CL"], "fields_of_study": ["Computer Science"], "published_date": "2025-02-07", "url": "https://arxiv.org/abs/2502.05078", "pdf_url": "https://arxiv.org/pdf/2502.05078v1", "arxiv_id": "2502.05078", "doi": "10.48550/arXiv.2502.05078", "citation_count": 16, "influential_citation_count": 2, "has_code": false, "code_url": null, "venue": "arXiv.org", "quality_score": 0.3076}
{"id": "e5732a92832ad0e3972596d8e5d88a9322814890b8898277b70d45e4be520023", "sources": ["arxiv", "semantic_scholar"], "title": "Scaling up Test-Time Compute with Latent Reasoning: A Recurrent Depth Approach", "abstract": "We study a novel language model architecture that is capable of scaling test-time computation by implicitly reasoning in latent space. Our model works by iterating a recurrent block, thereby unrolling to arbitrary depth at test-time. This stands in contrast to mainstream reasoning models that scale up compute by producing more tokens. Unlike approaches based on chain-of-thought, our approach does not require any specialized training data, can work with small context windows, and can capture types of reasoning that are not easily represented in words. We scale a proof-of-concept model to 3.5 billion parameters and 800 billion tokens. We show that the resulting model can improve its performance on reasoning benchmarks, sometimes dramatically, up to a computation load equivalent to 50 billion parameters.", "authors": ["Jonas Geiping", "Sean McLeish", "Neel Jain", "John Kirchenbauer", "Siddharth Singh", "Brian R. Bartoldson", "Bhavya Kailkhura", "Abhinav Bhatele", "Tom Goldstein"], "categories": ["cs.LG", "cs.CL"], "fields_of_study": ["Computer Science"], "published_date": "2025-02-07", "url": "https://arxiv.org/abs/2502.05171", "pdf_url": "https://arxiv.org/pdf/2502.05171v2", "arxiv_id": "2502.05171", "doi": "10.48550/arXiv.2502.05171", "citation_count": 253, "influential_citation_count": 49, "has_code": true, "code_url": "https://github.com/seal-rg/recurrent-pretraining", "venue": "arXiv.org", "quality_score": 0.8495}
{"id": "cda167d70fa4a73492ccb2f7fb0810b9b0456169e73d97454568871dd36f7ddd", "sources": ["arxiv", "semantic_scholar"], "title": "Dynamic Chain-of-Thought: Towards Adaptive Deep Reasoning", "abstract": "To reduce the cost and consumption of computing resources caused by computational redundancy and delayed reward assignment in long CoT, this research proposes the dynamic chain-of-thought (D-CoT) with adaptive reasoning time and steps. The researcher used simulation experiment to simulate the integration of D-CoT through Python 3.13 IDLE combined with a Python simulator based on GPTs. At the same time, the researcher used DeepSeek R1 as a control group to test and compare the performance of the D-CoT simulator in processing MIT OpenCourseWare's linear algebra exam questions. Experimental results show that D-CoT is better than DeepSeek R1 based on long CoT in three indicators: reasoning time, CoT length (reasoning steps) and token count, which achieves a significant reduction in computing resource consumption. In addition, this research has potential value in deep reasoning optimization that is used as a reference for future dynamic deep reasoning frameworks.", "authors": ["Libo Wang"], "categories": ["cs.AI", "cs.LG"], "fields_of_study": ["Computer Science"], "published_date": "2025-02-07", "url": "https://arxiv.org/abs/2502.10428", "pdf_url": "https://arxiv.org/pdf/2502.10428v4", "arxiv_id": "2502.10428", "doi": "10.48550/arXiv.2502.10428", "citation_count": 5, "influential_citation_count": 0, "has_code": true, "code_url": "https://github.com/brucewang123456789/GeniusTrail/tree/main/Dynamic%20CoT", "venue": "arXiv.org", "quality_score": 0.1945}
{"id": "0a36003ca8a005e6772934c7ec3529859a1b4d461631baf7b2bfb610f7ccdb94", "sources": ["arxiv", "semantic_scholar"], "title": "Llasa: Scaling Train-Time and Inference-Time Compute for Llama-based Speech Synthesis", "abstract": "Recent advances in text-based large language models (LLMs), particularly in the GPT series and the o1 model, have demonstrated the effectiveness of scaling both training-time and inference-time compute. However, current state-of-the-art TTS systems leveraging LLMs are often multi-stage, requiring separate models (e.g., diffusion models after LLM), complicating the decision of whether to scale a particular model during training or testing. This work makes the following contributions: First, we explore the scaling of train-time and inference-time compute for speech synthesis. Second, we propose a simple framework Llasa for speech synthesis that employs a single-layer vector quantizer (VQ) codec and a single Transformer architecture to fully align with standard LLMs such as Llama. Our experiments reveal that scaling train-time compute for Llasa consistently improves the naturalness of synthesized speech and enables the generation of more complex and accurate prosody patterns. Furthermore, from the perspective of scaling inference-time compute, we employ speech understanding models as verifiers during the search, finding that scaling inference-time compute shifts the sampling modes toward the preferences of specific verifiers, thereby improving emotional expressiveness, timbre consistency, and content accuracy. In addition, we released the checkpoint and training code for our TTS model (1B, 3B, 8B) and codec model publicly available.", "authors": ["Zhen Ye", "Xinfa Zhu", "Chi-Min Chan", "Xinsheng Wang", "Xu Tan", "Jiahe Lei", "Yi Peng", "Haohe Liu", "Yizhu Jin", "Zheqi Dai", "Hongzhan Lin", "Jianyi Chen", "Xingjian Du", "Liumeng Xue", "Yunlin Chen", "Zhifei Li", "Lei Xie", "Qiuqiang Kong", "Yike Guo", "Wei Xue"], "categories": ["eess.AS", "cs.AI", "cs.CL", "cs.MM", "cs.SD"], "fields_of_study": ["Engineering", "Computer Science"], "published_date": "2025-02-06", "url": "https://arxiv.org/abs/2502.04128", "pdf_url": "https://arxiv.org/pdf/2502.04128v2", "arxiv_id": "2502.04128", "doi": "10.48550/arXiv.2502.04128", "citation_count": 100, "influential_citation_count": 16, "has_code": false, "code_url": null, "venue": "arXiv.org", "quality_score": 0.6152}
{"id": "84c84c01d88ad0f4aa20874341fad87ccb2c7714176d36dd3574d67455697ee4", "sources": ["arxiv", "semantic_scholar"], "title": "Demystifying Long Chain-of-Thought Reasoning in LLMs", "abstract": "Scaling inference compute enhances reasoning in large language models (LLMs), with long chains-of-thought (CoTs) enabling strategies like backtracking and error correction. Reinforcement learning (RL) has emerged as a crucial method for developing these capabilities, yet the conditions under which long CoTs emerge remain unclear, and RL training requires careful design choices. In this study, we systematically investigate the mechanics of long CoT reasoning, identifying the key factors that enable models to generate long CoT trajectories. Through extensive supervised fine-tuning (SFT) and RL experiments, we present four main findings: (1) While SFT is not strictly necessary, it simplifies training and improves efficiency; (2) Reasoning capabilities tend to emerge with increased training compute, but their development is not guaranteed, making reward shaping crucial for stabilizing CoT length growth; (3) Scaling verifiable reward signals is critical for RL. We find that leveraging noisy, web-extracted solutions with filtering mechanisms shows strong potential, particularly for out-of-distribution (OOD) tasks such as STEM reasoning; and (4) Core abilities like error correction are inherently present in base models, but incentivizing these skills effectively for complex tasks via RL demands significant compute, and measuring their emergence requires a nuanced approach. These insights provide practical guidance for optimizing training strategies to enhance long CoT reasoning in LLMs. Our code is available at: https://github.com/eddycmu/demystify-long-cot.", "authors": ["Edward Yeo", "Yuxuan Tong", "Morry Niu", "Graham Neubig", "Xiang Yue"], "categories": ["cs.CL", "cs.LG"], "fields_of_study": ["Computer Science"], "published_date": "2025-02-05", "url": "https://arxiv.org/abs/2502.03373", "pdf_url": "https://arxiv.org/pdf/2502.03373v1", "arxiv_id": "2502.03373", "doi": "10.48550/arXiv.2502.03373", "citation_count": 326, "influential_citation_count": 27, "has_code": true, "code_url": "https://github.com/eddycmu/demystify-long-cot", "venue": "arXiv.org", "quality_score": 0.7236}
{"id": "c47491ba7193edb0e9e5c1c51d2a0a2a68baabddcbfa634dc63803b76a5f7ded", "sources": ["arxiv", "semantic_scholar"], "title": "Reasoning-as-Logic-Units: Scaling Test-Time Reasoning in Large Language Models Through Logic Unit Alignment", "abstract": "Chain-of-Thought (CoT) prompting has shown promise in enhancing the reasoning capabilities of large language models (LLMs) by generating natural language (NL) rationales that lead to the final answer. However, it struggles with numerical computation, which has somehow led to the development of program-aided techniques. Despite their potential, a persistent challenge remains: inconsistencies between LLM-reported reasoning steps and the logic in generated programs, which we term ``reasoning hallucinations.\" This stems from the inherent ambiguities of NL and the statistical nature of LLMs, which often lack rigorous logical coherence. To address this challenge, we propose a novel test-time scaling framework, Reasoning-as-Logic-Units (RaLU), which constructs a more reliable reasoning path by aligning logical units between the generated program and their corresponding NL descriptions. By decomposing the initially generated program into discrete units using static analysis, RaLU engages in an iterative dialogue with the LLM to judge, refine, and explain each unit. A rewind-and-correct mechanism ensures alignment between code statements and task requirements in each unit, ultimately forming a cohesive reasoning path under the program's logic, from which the model reaches a final solution. Our experiments demonstrate that RaLU significantly outperforms existing baselines in mathematical reasoning (GSM8K, MATH) and algorithmic reasoning (HumanEval+, MBPP+), underscoring its potential to advance LLM reasoning and programming by offering enhanced accuracy and interpretability.", "authors": ["Cheryl Li", "Tianyuan Xu", "Yiwen Guo"], "categories": ["cs.AI", "cs.LG"], "fields_of_study": ["Computer Science"], "published_date": "2025-02-05", "url": "https://arxiv.org/abs/2502.07803", "pdf_url": "https://arxiv.org/pdf/2502.07803v1", "arxiv_id": "2502.07803", "doi": "10.48550/arXiv.2502.07803", "citation_count": 14, "influential_citation_count": 1, "has_code": false, "code_url": null, "venue": "International Conference on Machine Learning", "quality_score": 0.294}
{"id": "b4d28b78bdda83fe63989cc7e5da8c61d4bf96ec9745bc3ba39db2e057a4c29c", "sources": ["arxiv", "semantic_scholar"], "title": "Lower Bounds for Chain-of-Thought Reasoning in Hard-Attention Transformers", "abstract": "Chain-of-thought reasoning and scratchpads have emerged as critical tools for enhancing the computational capabilities of transformers. While theoretical results show that polynomial-length scratchpads can extend transformers' expressivity from $TC^0$ to $PTIME$, their required length remains poorly understood. Empirical evidence even suggests that transformers need scratchpads even for many problems in $TC^0$, such as Parity or Multiplication, challenging optimistic bounds derived from circuit complexity. In this work, we initiate the study of systematic lower bounds for the number of chain-of-thought steps across different algorithmic problems, in the hard-attention regime. We study a variety of algorithmic problems, and provide bounds that are tight up to logarithmic factors. Overall, these results contribute to emerging understanding of the power and limitations of chain-of-thought reasoning.", "authors": ["Alireza Amiri", "Xinting Huang", "Mark Rofin", "Michael Hahn"], "categories": ["cs.LG", "cs.CC"], "fields_of_study": ["Computer Science"], "published_date": "2025-02-04", "url": "https://arxiv.org/abs/2502.02393", "pdf_url": "https://arxiv.org/pdf/2502.02393v3", "arxiv_id": "2502.02393", "doi": "10.48550/arXiv.2502.02393", "citation_count": 29, "influential_citation_count": 2, "has_code": false, "code_url": null, "venue": "International Conference on Machine Learning", "quality_score": 0.3693}
{"id": "0c141b7cc3cb769cf74819e9091e3e8b65f3af749a2a72f1febf98eb1c21fcc3", "sources": ["arxiv", "semantic_scholar"], "title": "CoAT: Chain-of-Associated-Thoughts Framework for Enhancing Large Language Models Reasoning", "abstract": "Research on LLM technologies is rapidly emerging, with most of them employ a 'fast thinking' approach to inference. Most LLMs generate the final result based solely on a single query and LLM's reasoning capabilities. However, with the advent of OpenAI-o1, 'slow thinking' techniques have garnered increasing attention because its process is closer to the human thought process. Inspired by the human ability to constantly associate and replenish knowledge during thinking, we developed the novel Chain-of-Associated-Thoughts (CoAT) framework, which introduces an innovative synergy between the Monte Carlo Tree Search (MCTS) algorithm and a dynamic mechanism for integrating new key information, termed 'associative memory'. By combining the structured exploration capabilities of MCTS with the adaptive learning capacity of associative memory, CoAT significantly expands the LLM search space, enabling our framework to explore diverse reasoning pathways and dynamically update its knowledge base in real-time. This allows the framework to not only revisit and refine earlier inferences but also adaptively incorporate evolving information, ensuring that the final output is both accurate and comprehensive. We validate CoAT's effectiveness across a variety of generative and reasoning tasks. Quantitative experiments show that CoAT achieves over 10% performance improvement on open-source multi-hop reasoning datasets (HotpotQA, MuSiQue) and more than 15% gain on our proprietary CRB dataset.", "authors": ["Jianfeng Pan", "Senyou Deng", "Shaomang Huang"], "categories": ["cs.CL", "cs.AI"], "fields_of_study": ["Computer Science"], "published_date": "2025-02-04", "url": "https://arxiv.org/abs/2502.02390", "pdf_url": "https://arxiv.org/pdf/2502.02390v3", "arxiv_id": "2502.02390", "doi": "10.48550/arXiv.2502.02390", "citation_count": 15, "influential_citation_count": 2, "has_code": true, "code_url": null, "venue": "Conference on Empirical Methods in Natural Language Processing", "quality_score": 0.301}
{"id": "32d6218156e9ea1e80bdacd8f67a9d7df10134b81dc95cc9378b281b1bcaa749", "sources": ["arxiv", "semantic_scholar"], "title": "Satori: Reinforcement Learning with Chain-of-Action-Thought Enhances LLM Reasoning via Autoregressive Search", "abstract": "Large language models (LLMs) have demonstrated remarkable reasoning capabilities across diverse domains. Recent studies have shown that increasing test-time computation enhances LLMs' reasoning capabilities. This typically involves extensive sampling at inference time guided by an external LLM verifier, resulting in a two-player system. Despite external guidance, the effectiveness of this system demonstrates the potential of a single LLM to tackle complex tasks. Thus, we pose a new research problem: Can we internalize the searching capabilities to fundamentally enhance the reasoning abilities of a single LLM? This work explores an orthogonal direction focusing on post-training LLMs for autoregressive searching (i.e., an extended reasoning process with self-reflection and self-exploration of new strategies). To achieve this, we propose the Chain-of-Action-Thought (COAT) reasoning and a two-stage training paradigm: 1) a small-scale format tuning stage to internalize the COAT reasoning format and 2) a large-scale self-improvement stage leveraging reinforcement learning. Our approach results in Satori, a 7B LLM trained on open-source models and data. Extensive empirical evaluations demonstrate that Satori achieves state-of-the-art performance on mathematical reasoning benchmarks while exhibits strong generalization to out-of-domain tasks. Code, data, and models are fully open-sourced.", "authors": ["Maohao Shen", "Guangtao Zeng", "Zhenting Qi", "Zhang-Wei Hong", "Zhenfang Chen", "Wei Lu", "Gregory Wornell", "Subhro Das", "David Cox", "Chuang Gan"], "categories": ["cs.CL", "cs.AI"], "fields_of_study": ["Computer Science"], "published_date": "2025-02-04", "url": "https://arxiv.org/abs/2502.02508", "pdf_url": "https://arxiv.org/pdf/2502.02508v3", "arxiv_id": "2502.02508", "doi": "10.48550/arXiv.2502.02508", "citation_count": 54, "influential_citation_count": 5, "has_code": true, "code_url": null, "venue": "International Conference on Machine Learning", "quality_score": 0.4351}
{"id": "d350689088ada05df5fc33fadc65854226bd74879381546c40c1fa854359d2e2", "sources": ["arxiv", "semantic_scholar"], "title": "Latent Thought Models with Variational Bayes Inference-Time Computation", "abstract": "We propose a novel class of language models, Latent Thought Models (LTMs), which incorporate explicit latent thought vectors that follow an explicit prior model in latent space. These latent thought vectors guide the autoregressive generation of ground tokens through a Transformer decoder. Training employs a dual-rate optimization process within the classical variational Bayes framework: fast learning of local variational parameters for the posterior distribution of latent vectors (inference-time computation), and slow learning of global decoder parameters. Empirical studies reveal that LTMs possess additional scaling dimensions beyond traditional Large Language Models (LLMs), such as the number of iterations in inference-time computation and number of latent thought vectors. Higher sample efficiency can be achieved by increasing training compute per token, with further gains possible by trading model size for more inference steps. Designed based on these scaling properties, LTMs demonstrate superior sample and parameter efficiency compared to autoregressive models and discrete diffusion models. They significantly outperform these counterparts in validation perplexity and zero-shot language modeling tasks. Additionally, LTMs exhibit emergent few-shot in-context reasoning capabilities that scale with model size, and achieve competitive performance in conditional and unconditional text generation.", "authors": ["Deqian Kong", "Minglu Zhao", "Dehong Xu", "Bo Pang", "Shu Wang", "Edouardo Honig", "Zhangzhang Si", "Chuan Li", "Jianwen Xie", "Sirui Xie", "Ying Nian Wu"], "categories": ["cs.CL", "cs.LG", "stat.ML"], "fields_of_study": ["Computer Science", "Mathematics"], "published_date": "2025-02-03", "url": "https://arxiv.org/abs/2502.01567", "pdf_url": "https://arxiv.org/pdf/2502.01567v2", "arxiv_id": "2502.01567", "doi": null, "citation_count": 18, "influential_citation_count": 1, "has_code": false, "code_url": null, "venue": "International Conference on Machine Learning", "quality_score": 0.3197}
{"id": "8b80a5a4bd79b40ec31b37888692278b9727075db48da3fc3dd89f5fa0ff6117", "sources": ["arxiv", "semantic_scholar"], "title": "Sample, Scrutinize and Scale: Effective Inference-Time Search by Scaling Verification", "abstract": "Sampling-based search, a simple paradigm for utilizing test-time compute, involves generating multiple candidate responses and selecting the best one -- typically by having models self-verify each response for correctness. In this paper, we study the scaling trends governing sampling-based search. Among our findings is that simply scaling up a minimalist implementation of sampling-based search, using only random sampling and direct self-verification, provides a practical inference method that, for example, elevates the reasoning capabilities of Gemini v1.5 Pro above that of o1-Preview on popular benchmarks. We partially attribute the scalability of sampling-based search to a phenomenon of implicit scaling, where sampling a larger pool of responses in turn improves self-verification accuracy. We further identify two useful principles for improving self-verification capabilities with test-time compute: (1) comparing across responses provides helpful signals about the locations of errors and hallucinations, and (2) different model output styles are useful for different contexts -- chains of thought are useful for reasoning but harder to verify. We also find that, though accurate verification can be elicited, frontier models demonstrate remarkably weak out-of-box verification capabilities and introduce a benchmark to measure progress on these deficiencies.", "authors": ["Eric Zhao", "Pranjal Awasthi", "Sreenivas Gollapudi"], "categories": ["cs.LG", "cs.AI"], "fields_of_study": ["Computer Science"], "published_date": "2025-02-03", "url": "https://arxiv.org/abs/2502.01839", "pdf_url": "https://arxiv.org/pdf/2502.01839v2", "arxiv_id": "2502.01839", "doi": "10.48550/arXiv.2502.01839", "citation_count": 42, "influential_citation_count": 5, "has_code": false, "code_url": null, "venue": "International Conference on Machine Learning", "quality_score": 0.4084}
{"id": "c7daadecfa1cde175fa1c8b85ea0dcbaea43f38a7d5ef14e392f81938cb38367", "sources": ["arxiv", "semantic_scholar"], "title": "Metastable Dynamics of Chain-of-Thought Reasoning: Provable Benefits of Search, RL and Distillation", "abstract": "A key paradigm to improve the reasoning capabilities of large language models (LLMs) is to allocate more inference-time compute to search against a verifier or reward model. This process can then be utilized to refine the pretrained model or distill its reasoning patterns into more efficient models. In this paper, we study inference-time compute by viewing chain-of-thought (CoT) generation as a metastable Markov process: easy reasoning steps (e.g., algebraic manipulations) form densely connected clusters, while hard reasoning steps (e.g., applying a relevant theorem) create sparse, low-probability edges between clusters, leading to phase transitions at longer timescales. Under this framework, we prove that implementing a search protocol that rewards sparse edges improves CoT by decreasing the expected number of steps to reach different clusters. In contrast, we establish a limit on reasoning capability when the model is restricted to local information of the pretrained graph. We also show that the information gained by search can be utilized to obtain a better reasoning model: (1) the pretrained model can be directly finetuned to favor sparse edges via policy gradient methods, and moreover (2) a compressed metastable representation of the reasoning dynamics can be distilled into a smaller, more efficient model.", "authors": ["Juno Kim", "Denny Wu", "Jason Lee", "Taiji Suzuki"], "categories": ["cs.AI", "cs.LG", "stat.ML"], "fields_of_study": ["Computer Science", "Mathematics"], "published_date": "2025-02-02", "url": "https://arxiv.org/abs/2502.01694", "pdf_url": "https://arxiv.org/pdf/2502.01694v2", "arxiv_id": "2502.01694", "doi": "10.48550/arXiv.2502.01694", "citation_count": 21, "influential_citation_count": 3, "has_code": false, "code_url": null, "venue": "International Conference on Machine Learning", "quality_score": 0.3356}
{"id": "b6049c69d3e65e13aa4776287a52fc2533bd4ebc1c0341e33b98de1414d754b9", "sources": ["arxiv", "semantic_scholar"], "title": "CDW-CoT: Clustered Distance-Weighted Chain-of-Thoughts Reasoning", "abstract": "Large Language Models (LLMs) have recently achieved impressive results in complex reasoning tasks through Chain of Thought (CoT) prompting. However, most existing CoT methods rely on using the same prompts, whether manually designed or automatically generated, to handle the entire dataset. This one-size-fits-all approach may fail to meet the specific needs arising from the diversities within a single dataset. To solve this problem, we propose the Clustered Distance-Weighted Chain of Thought (CDW-CoT) method, which dynamically constructs prompts tailored to the characteristics of each data instance by integrating clustering and prompt optimization techniques. Our method employs clustering algorithms to categorize the dataset into distinct groups, from which a candidate pool of prompts is selected to reflect the inherent diversity within the dataset. For each cluster, CDW-CoT trains the optimal prompt probability distribution tailored to their specific characteristics. Finally, it dynamically constructs a unique prompt probability distribution for each test instance, based on its proximity to cluster centers, from which prompts are selected for reasoning. CDW-CoT consistently outperforms traditional CoT methods across six datasets, including commonsense, symbolic, and mathematical reasoning tasks. Specifically, when compared to manual CoT, CDW-CoT achieves an average accuracy improvement of 25.34% on LLaMA2 (13B) and 15.72% on LLaMA3 (8B).", "authors": ["Yuanheng Fang", "Guoqing Chao", "Wenqiang Lei", "Shaobo Li", "Dianhui Chu"], "categories": ["cs.LG"], "fields_of_study": ["Computer Science"], "published_date": "2025-01-21", "url": "https://arxiv.org/abs/2501.12226", "pdf_url": "https://arxiv.org/pdf/2501.12226v1", "arxiv_id": "2501.12226", "doi": "10.48550/arXiv.2501.12226", "citation_count": 5, "influential_citation_count": 1, "has_code": false, "code_url": null, "venue": "AAAI Conference on Artificial Intelligence", "quality_score": 0.1945}
{"id": "4627370986942d6ebf4e99b8d8f6894f6c9e74a25c7cfbd685f2a7d830b40c3e", "sources": ["arxiv", "semantic_scholar"], "title": "T1: Advancing Language Model Reasoning through Reinforcement Learning and Inference Scaling", "abstract": "Large language models (LLMs) have demonstrated remarkable capabilities in complex reasoning tasks. However, existing approaches mainly rely on imitation learning and struggle to achieve effective test-time scaling. While reinforcement learning (RL) holds promise for enabling self-exploration, recent attempts yield modest improvements in complex reasoning. In this paper, we present T1 to scale RL by encouraging exploration and understand inference scaling. We first initialize the LLM using synthesized chain-of-thought data that integrates trial-and-error and self-verification. To scale RL training, we promote increased sampling diversity through oversampling. We demonstrate that T1 with open LLMs as its base exhibits inference scaling behavior and achieves superior performance on challenging math reasoning benchmarks. More importantly, we present a simple strategy to examine inference scaling, where increased inference budgets directly lead to T1's better performance without any additional verification.", "authors": ["Zhenyu Hou", "Xin Lv", "Rui Lu", "Jiajie Zhang", "Yujiang Li", "Zijun Yao", "Juanzi Li", "Jie Tang", "Yuxiao Dong"], "categories": ["cs.LG", "cs.CL"], "fields_of_study": ["Computer Science"], "published_date": "2025-01-20", "url": "https://arxiv.org/abs/2501.11651", "pdf_url": "https://arxiv.org/pdf/2501.11651v2", "arxiv_id": "2501.11651", "doi": "10.48550/arXiv.2501.11651", "citation_count": 80, "influential_citation_count": 6, "has_code": false, "code_url": null, "venue": "International Conference on Machine Learning", "quality_score": 0.4771}
{"id": "0785aa0184749c039dcc8475824d0b5889901581cc3360d962d92a5d80437efd", "sources": ["arxiv", "semantic_scholar"], "title": "RedStar: Does Scaling Long-CoT Data Unlock Better Slow-Reasoning Systems?", "abstract": "Can scaling transform reasoning? In this work, we explore the untapped potential of scaling Long Chain-of-Thought (Long-CoT) data to 1000k samples, pioneering the development of a slow-thinking model, RedStar. Through extensive experiments with various LLMs and different sizes, we uncover the ingredients for specialization and scale for Long-CoT training. Surprisingly, even smaller models show significant performance gains with limited data, revealing the sample efficiency of Long-CoT and the critical role of sample difficulty in the learning process. Our findings demonstrate that Long-CoT reasoning can be effectively triggered with just a few thousand examples, while larger models achieve unparalleled improvements. We also introduce reinforcement learning (RL)-scale training as a promising direction for advancing slow-thinking systems. RedStar shines across domains: on the MATH-Hard benchmark, RedStar-code-math boosts performance from 66.2\\% to 81.6\\%, and on the USA Math Olympiad (AIME), it solves 46.7\\% of problems using only 21k mixed-code-math datasets. In multimodal tasks like GeoQA and MathVista-GEO, RedStar-Geo achieves competitive results with minimal Long-CoT data, outperforming other slow-thinking systems like QvQ-Preview. Compared to QwQ, RedStar strikes the perfect balance between reasoning and generalizability. Our work highlights that, with careful tuning, scaling Long-CoT can unlock extraordinary reasoning capabilities-even with limited dataset and set a new standard for slow-thinking models across diverse challenges. Our data and models are released at https://huggingface.co/RedStar-Reasoning.", "authors": ["Haotian Xu", "Xing Wu", "Weinong Wang", "Zhongzhi Li", "Da Zheng", "Boyuan Chen", "Yi Hu", "Shijia Kang", "Jiaming Ji", "Yingying Zhang", "Zhijiang Guo", "Yaodong Yang", "Muhan Zhang", "Debing Zhang"], "categories": ["cs.LG", "cs.AI", "cs.CL"], "fields_of_study": ["Computer Science"], "published_date": "2025-01-20", "url": "https://arxiv.org/abs/2501.11284", "pdf_url": "https://arxiv.org/pdf/2501.11284v1", "arxiv_id": "2501.11284", "doi": "10.48550/arXiv.2501.11284", "citation_count": 45, "influential_citation_count": 5, "has_code": false, "code_url": null, "venue": "arXiv.org", "quality_score": 0.4157}
{"id": "e3ae0f18425e12b434c74f19cc7be12f29fc23f30a5ce16137492c36954b4096", "sources": ["arxiv", "semantic_scholar"], "title": "Enhancing Generalization in Chain of Thought Reasoning for Smaller Models", "abstract": "Chain-of-Thought (CoT) reasoning in smaller language models is a challenging natural language process problem yet highly desirable in many real-life applications. Existing CoT knowledge distillation methods often suffer from overly conservative memorization in smaller LLMs, leading to low generalization confidence. As fully preserving the CoT ability of teacher model is impossible, we hypothesize that adversarial CoT fine-tuning is crucial for developing smaller LLM with robust CoT generalization. To this end, we propose \\textit{PRompt-Assisted Domain-Adversarial fine-tuning} (PRADA), a principled fine-tuning framework that integrates diverse CoT domains. Specifically, PRADA pioneers two CoT improvements in smaller LLM: (1) Recovering the domain-invariant feature insight which typically lost during distillation with domain adversarial fine-tuning; (2) Enhancing the domain adaptability of CoT prompt engineering by employing domain-adversarial approaches. We theoretically demonstrate the effectiveness of our approach and empirically show that it significantly outperforms the state of the arts in a wide range of tasks. Moreover, our empirical findings reveal that the smaller LLM, when leveraging PRADA, aligns closely with domain knowledge, thereby improving the explainability of our approach.", "authors": ["Maxwell J. Yin", "Dingyi Jiang", "Yongbing Chen", "Boyu Wang", "Charles Ling"], "categories": ["cs.LG", "cs.AI", "cs.CL"], "fields_of_study": ["Computer Science"], "published_date": "2025-01-16", "url": "https://arxiv.org/abs/2501.09804", "pdf_url": "https://arxiv.org/pdf/2501.09804v1", "arxiv_id": "2501.09804", "doi": "10.48550/arXiv.2501.09804", "citation_count": 7, "influential_citation_count": 1, "has_code": false, "code_url": null, "venue": "arXiv.org", "quality_score": 0.2258}
{"id": "81d86028f5bc366310934990f54f388678b4c355d0f8d809728336f0308a0ccb", "sources": ["arxiv", "semantic_scholar"], "title": "Audio-CoT: Exploring Chain-of-Thought Reasoning in Large Audio Language Model", "abstract": "Large Audio-Language Models (LALMs) have demonstrated remarkable performance in tasks involving audio perception and understanding, such as speech recognition and audio captioning. However, their reasoning capabilities - critical for solving complex real-world problems - remain underexplored. In this work, we conduct the first exploration into integrating Chain-of-Thought (CoT) reasoning into LALMs to enhance their reasoning ability across auditory modalities. We evaluate representative CoT methods, analyzing their performance in both information extraction and reasoning tasks across sound, music, and speech domains. Our findings reveal that CoT methods significantly improve performance on easy and medium tasks but encounter challenges with hard tasks, where reasoning chains can confuse the model rather than improve accuracy. Additionally, we identify a positive correlation between reasoning path length and accuracy, demonstrating the potential of scaling inference for advanced instruction-following and reasoning. This study not only highlights the promise of CoT in enhancing LALM reasoning capabilities but also identifies key limitations and provides actionable directions for future research.", "authors": ["Ziyang Ma", "Zhuo Chen", "Yuping Wang", "Eng Siong Chng", "Xie Chen"], "categories": ["cs.SD", "cs.CL", "cs.MM", "eess.AS"], "fields_of_study": ["Computer Science", "Engineering"], "published_date": "2025-01-13", "url": "https://arxiv.org/abs/2501.07246", "pdf_url": "https://arxiv.org/pdf/2501.07246v1", "arxiv_id": "2501.07246", "doi": "10.1109/ASRU65441.2025.11434628", "citation_count": 68, "influential_citation_count": 3, "has_code": false, "code_url": null, "venue": "Automatic Speech Recognition & Understanding", "quality_score": 0.4597}
{"id": "2dc6c460bedbad332ae0643b986a295cb7061036e0b948ee91f03560809aa38c", "sources": ["arxiv", "semantic_scholar"], "title": "O1 Replication Journey -- Part 3: Inference-time Scaling for Medical Reasoning", "abstract": "Building upon our previous investigations of O1 replication (Part 1: Journey Learning [Qin et al., 2024] and Part 2: Distillation [Huang et al., 2024]), this work explores the potential of inference-time scaling in large language models (LLMs) for medical reasoning tasks, ranging from diagnostic decision-making to treatment planning. Through extensive experiments on medical benchmarks of varying complexity (MedQA, Medbullets, and JAMA Clinical Challenges), our investigation reveals several key insights: (1) Increasing inference time does lead to improved performance. With a modest training set of 500 samples, our model yields substantial performance improvements of 6%-11%. (2) Task complexity directly correlates with the required length of reasoning chains, confirming the necessity of extended thought processes for challenging problems. (3) The differential diagnoses generated by our model adhere to the principles of the hypothetico-deductive method, producing a list of potential conditions that may explain a patient's symptoms and systematically narrowing these possibilities by evaluating the evidence. These findings demonstrate the promising synergy between inference-time scaling and journey learning in advancing LLMs' real-world clinical reasoning capabilities.", "authors": ["Zhongzhen Huang", "Gui Geng", "Shengyi Hua", "Zhen Huang", "Haoyang Zou", "Shaoting Zhang", "Pengfei Liu", "Xiaofan Zhang"], "categories": ["cs.CL"], "fields_of_study": ["Computer Science"], "published_date": "2025-01-11", "url": "https://arxiv.org/abs/2501.06458", "pdf_url": "https://arxiv.org/pdf/2501.06458v1", "arxiv_id": "2501.06458", "doi": "10.48550/arXiv.2501.06458", "citation_count": 34, "influential_citation_count": 0, "has_code": false, "code_url": null, "venue": "arXiv.org", "quality_score": 0.386}
{"id": "12ef41c79c0b6ca1875cca7b9a2d0cdb8d06761e0aca3bfb294409506f69a296", "sources": ["arxiv", "semantic_scholar"], "title": "Towards System 2 Reasoning in LLMs: Learning How to Think With Meta Chain-of-Thought", "abstract": "We propose a novel framework, Meta Chain-of-Thought (Meta-CoT), which extends traditional Chain-of-Thought (CoT) by explicitly modeling the underlying reasoning required to arrive at a particular CoT. We present empirical evidence from state-of-the-art models exhibiting behaviors consistent with in-context search, and explore methods for producing Meta-CoT via process supervision, synthetic data generation, and search algorithms. Finally, we outline a concrete pipeline for training a model to produce Meta-CoTs, incorporating instruction tuning with linearized search traces and reinforcement learning post-training. Finally, we discuss open research questions, including scaling laws, verifier roles, and the potential for discovering novel reasoning algorithms. This work provides a theoretical and practical roadmap to enable Meta-CoT in LLMs, paving the way for more powerful and human-like reasoning in artificial intelligence.", "authors": ["Violet Xiang", "Charlie Snell", "Kanishk Gandhi", "Alon Albalak", "Anikait Singh", "Chase Blagden", "Duy Phung", "Rafael Rafailov", "Nathan Lile", "Dakota Mahan", "Louis Castricato", "Jan-Philipp Franken", "Nick Haber", "Chelsea Finn"], "categories": ["cs.AI", "cs.CL"], "fields_of_study": ["Computer Science"], "published_date": "2025-01-08", "url": "https://arxiv.org/abs/2501.04682", "pdf_url": "https://arxiv.org/pdf/2501.04682v1", "arxiv_id": "2501.04682", "doi": "10.48550/arXiv.2501.04682", "citation_count": 103, "influential_citation_count": 7, "has_code": false, "code_url": null, "venue": "arXiv.org", "quality_score": 0.5043}
{"id": "0a729e4d19fe75f6fcfeda4f1473f614d28d7de9e06d89419ae1294360303ca5", "sources": ["arxiv", "semantic_scholar"], "title": "A Survey of Test-Time Compute: From Intuitive Inference to Deliberate Reasoning", "abstract": "The remarkable performance of the o1 model in complex reasoning demonstrates that test-time compute scaling can further unlock the model's potential, enabling powerful System-2 thinking. However, there is still a lack of comprehensive surveys for test-time compute scaling. We trace the concept of test-time compute back to System-1 models. In System-1 models, test-time compute addresses distribution shifts and improves robustness and generalization through parameter updating, input modification, representation editing, and output calibration. In System-2 models, it enhances the model's reasoning ability to solve complex problems through repeated sampling, self-correction, and tree search. We organize this survey according to the trend of System-1 to System-2 thinking, highlighting the key role of test-time compute in the transition from System-1 models to weak System-2 models, and then to strong System-2 models. We also point out advanced topics and future directions.", "authors": ["Yixin Ji", "Juntao Li", "Yang Xiang", "Hai Ye", "Kaixin Wu", "Kai Yao", "Jia Xu", "Linjian Mo", "Min Zhang"], "categories": ["cs.AI", "cs.CL", "cs.LG"], "fields_of_study": ["Computer Science"], "published_date": "2025-01-05", "url": "https://arxiv.org/abs/2501.02497", "pdf_url": "https://arxiv.org/pdf/2501.02497v3", "arxiv_id": "2501.02497", "doi": "10.1162/coli.a.629", "citation_count": 19, "influential_citation_count": 1, "has_code": false, "code_url": null, "venue": "Computational Linguistics", "quality_score": 0.3253}
{"id": "df5ae64c921923d68b988bff1cdc545e9ce9349dfbac2a2d5ad466a6080b7084", "sources": ["arxiv", "semantic_scholar"], "title": "Table as Thought: Exploring Structured Thoughts in LLM Reasoning", "abstract": "Large language models' reasoning abilities benefit from methods that organize their thought processes, such as chain-of-thought prompting, which employs a sequential structure to guide the reasoning process step-by-step. However, existing approaches focus primarily on organizing the sequence of thoughts, leaving structure in individual thought steps underexplored. To address this gap, we propose Table as Thought, a framework inspired by cognitive neuroscience theories on human thought. Table as Thought organizes reasoning within a tabular schema, where rows represent sequential thought steps and columns capture critical constraints and contextual information to enhance reasoning. The reasoning process iteratively populates the table until self-verification ensures completeness and correctness. Our experiments show that Table as Thought excels in planning tasks and demonstrates a strong potential for enhancing LLM performance in mathematical reasoning compared to unstructured thought baselines. This work provides a novel exploration of refining thought representation within LLMs, paving the way for advancements in reasoning and AI cognition.", "authors": ["Zhenjie Sun", "Naihao Deng", "Haofei Yu", "Jiaxuan You"], "categories": ["cs.AI", "cs.CL"], "fields_of_study": ["Computer Science"], "published_date": "2025-01-04", "url": "https://arxiv.org/abs/2501.02152", "pdf_url": "https://arxiv.org/pdf/2501.02152v1", "arxiv_id": "2501.02152", "doi": "10.48550/arXiv.2501.02152", "citation_count": 6, "influential_citation_count": 0, "has_code": false, "code_url": null, "venue": "arXiv.org", "quality_score": 0.2113}
{"id": "c134184379c4cf099453ac57ca51c42cb9819963b05df81d8ace24caeb87ce75", "sources": ["arxiv", "semantic_scholar"], "title": "Toward Adaptive Reasoning in Large Language Models with Thought Rollback", "abstract": "Large language models (LLMs) have been routinely used to solve various tasks using step-by-step reasoning. However, the structure of intermediate reasoning steps, or thoughts, is rigid and unidirectional, such as chains, trees, or acyclic-directed graphs. Consequently, the resulting inflexible and forward-only reasoning may not address challenging tasks and fail when the LLM frequently gives false responses, i.e., ``hallucinations''. This paper proposes a new reasoning framework, called Thought Rollback (TR), allowing LLMs to adaptively build thought structure while maintaining effective reasoning toward problem-solving under ``hallucinations''. The core mechanism of TR is rolling back thoughts, which allows LLMs to perform error analysis on thoughts, and thus roll back to any previously mistaken thought for revision. Subsequently, by including such trial-and-error in the prompt to guide the LLM, each rollback leads to one more reliable reasoning path. Therefore, starting with a simple prompt without human annotations, LLM with TR adaptively and gradually explores thoughts for a correct solution. Comprehensive experiments on mathematical problems and multi-task reasoning demonstrate the state-of-the-art performance of TR in terms of problem-solving rate and interaction cost. For instance, the solving rate of GPT-4 with TR outperforms the current best by $9\\%$ on the MATH dataset.", "authors": ["Sijia Chen", "Baochun Li"], "categories": ["cs.AI", "cs.CL", "cs.LG"], "fields_of_study": ["Computer Science"], "published_date": "2024-12-27", "url": "https://arxiv.org/abs/2412.19707", "pdf_url": "https://arxiv.org/pdf/2412.19707v1", "arxiv_id": "2412.19707", "doi": "10.48550/arXiv.2412.19707", "citation_count": 19, "influential_citation_count": 0, "has_code": true, "code_url": "https://github.com/iQua/llmpebase/tree/main/examples/ThoughtRollback", "venue": "International Conference on Machine Learning", "quality_score": 0.3253}
{"id": "4e648ccf95311b11e5b2eb466c4867736cca0c4d3edc32814347dc5115506dde", "sources": ["arxiv", "semantic_scholar"], "title": "DRT: Deep Reasoning Translation via Long Chain-of-Thought", "abstract": "Recently, O1-like models have emerged as representative examples, illustrating the effectiveness of long chain-of-thought (CoT) in reasoning tasks such as math and coding tasks. In this paper, we introduce DRT, an attempt to bring the success of long CoT to neural machine translation (MT). Specifically, in view of the literature books that might involve similes and metaphors, translating these texts to a target language is very difficult in practice due to cultural differences. In such cases, literal translation often fails to convey the intended meaning effectively. Even for professional human translators, considerable thought must be given to preserving semantics throughout the translation process. To simulate LLMs' long thought ability in MT, we first mine sentences containing similes or metaphors from existing literature books, and then develop a multi-agent framework to translate these sentences via long thought. In the multi-agent framework, a translator is used to iteratively translate the source sentence under the suggestions provided by an advisor. To ensure the effectiveness of the long thoughts, an evaluator is also employed to quantify the translation quality in each round. In this way, we collect tens of thousands of long-thought MT data, which is used to train our DRT. Using Qwen2.5 and LLama-3.1 as the backbones, DRT models can learn the thought process during machine translation, and outperform vanilla LLMs as well as LLMs which are simply fine-tuning on the paired sentences without long thought, showing its effectiveness. The synthesized data and model checkpoints are released at https://github.com/krystalan/DRT.", "authors": ["Jiaan Wang", "Fandong Meng", "Yunlong Liang", "Jie Zhou"], "categories": ["cs.CL", "cs.AI"], "fields_of_study": ["Computer Science"], "published_date": "2024-12-23", "url": "https://arxiv.org/abs/2412.17498", "pdf_url": "https://arxiv.org/pdf/2412.17498v4", "arxiv_id": "2412.17498", "doi": "10.18653/v1/2025.findings-acl.351", "citation_count": 28, "influential_citation_count": 2, "has_code": true, "code_url": "https://github.com/krystalan/DRT", "venue": "Annual Meeting of the Association for Computational Linguistics", "quality_score": 0.3656}
{"id": "d49d8bb9a84265bc3cebc1c7280b06efeaa91b95dd31d2fe8e8758db4e03e1f9", "sources": ["arxiv", "semantic_scholar"], "title": "Compressed Chain of Thought: Efficient Reasoning Through Dense Representations", "abstract": "Chain-of-thought (CoT) decoding enables language models to improve reasoning performance at the cost of high generation latency in decoding. Recent proposals have explored variants of contemplation tokens, a term we introduce that refers to special tokens used during inference to allow for extra computation. Prior work has considered fixed-length sequences drawn from a discrete set of embeddings as contemplation tokens. Here we propose Compressed Chain-of-Thought (CCoT), a framework to generate contentful and continuous contemplation tokens of variable sequence length. The generated contemplation tokens are compressed representations of explicit reasoning chains, and our method can be applied to off-the-shelf decoder language models. Through experiments, we illustrate how CCoT enables additional reasoning over dense contentful representations to achieve corresponding improvements in accuracy. Moreover, the reasoning improvements can be adaptively modified on demand by controlling the number of contemplation tokens generated.", "authors": ["Jeffrey Cheng", "Benjamin Van Durme"], "categories": ["cs.CL"], "fields_of_study": ["Computer Science"], "published_date": "2024-12-17", "url": "https://arxiv.org/abs/2412.13171", "pdf_url": "https://arxiv.org/pdf/2412.13171v1", "arxiv_id": "2412.13171", "doi": "10.48550/arXiv.2412.13171", "citation_count": 171, "influential_citation_count": 10, "has_code": false, "code_url": null, "venue": "arXiv.org", "quality_score": 0.5589}
{"id": "1395dfc3b4508e29bac1320c932051075a8aced2dc70e0f94f29afbabda16a08", "sources": ["arxiv", "semantic_scholar"], "title": "Emma-X: An Embodied Multimodal Action Model with Grounded Chain of Thought and Look-ahead Spatial Reasoning", "abstract": "Traditional reinforcement learning-based robotic control methods are often task-specific and fail to generalize across diverse environments or unseen objects and instructions. Visual Language Models (VLMs) demonstrate strong scene understanding and planning capabilities but lack the ability to generate actionable policies tailored to specific robotic embodiments. To address this, Visual-Language-Action (VLA) models have emerged, yet they face challenges in long-horizon spatial reasoning and grounded task planning. In this work, we propose the Embodied Multimodal Action Model with Grounded Chain of Thought and Look-ahead Spatial Reasoning, Emma-X. Emma-X leverages our constructed hierarchical embodiment dataset based on BridgeV2, containing 60,000 robot manipulation trajectories auto-annotated with grounded task reasoning and spatial guidance. Additionally, we introduce a trajectory segmentation strategy based on gripper states and motion trajectories, which can help mitigate hallucination in grounding subtask reasoning generation. Experimental results demonstrate that Emma-X achieves superior performance over competitive baselines, particularly in real-world robotic tasks requiring spatial reasoning.", "authors": ["Qi Sun", "Pengfei Hong", "Tej Deep Pala", "Vernon Toh", "U-Xuan Tan", "Deepanway Ghosal", "Soujanya Poria"], "categories": ["cs.RO", "cs.AI", "cs.CL", "cs.CV"], "fields_of_study": ["Computer Science"], "published_date": "2024-12-16", "url": "https://arxiv.org/abs/2412.11974", "pdf_url": "https://arxiv.org/pdf/2412.11974v2", "arxiv_id": "2412.11974", "doi": "10.48550/arXiv.2412.11974", "citation_count": 38, "influential_citation_count": 5, "has_code": true, "code_url": "https://github.com/declare-lab/Emma-X", "venue": "arXiv.org", "quality_score": 0.3978}
{"id": "2901e91cbb3a3edcf8e42dd4e217388b222c168e4aa3ba88ab5c4205ceac1021", "sources": ["arxiv", "semantic_scholar"], "title": "Forest-of-Thought: Scaling Test-Time Compute for Enhancing LLM Reasoning", "abstract": "Large Language Models (LLMs) have demonstrated remarkable abilities across various language tasks, but solving complex reasoning problems remains a significant challenge. While existing methods, such as Chain-of-Thought (CoT) and Tree-of-Thought (ToT), enhance reasoning by decomposing problems or structuring prompts, they typically perform a single pass of reasoning and may fail to revisit flawed paths, compromising accuracy. To address this limitation, we propose a novel reasoning framework called Forest-of-Thought (FoT), which integrates multiple reasoning trees to leverage collective decision-making for solving complex logical problems. FoT employs sparse activation strategies to select the most relevant reasoning paths, improving both efficiency and accuracy. Additionally, we introduce a dynamic self-correction strategy that enables real-time error correction, along with consensus-guided decision-making strategies to optimize both correctness and computational resources. Experimental results demonstrate that the FoT framework, combined with these strategies, significantly enhances the reasoning capabilities of LLMs, enabling them to solve complex tasks with greater precision and efficiency. Code will be available at https://github.com/iamhankai/Forest-of-Thought.", "authors": ["Zhenni Bi", "Kai Han", "Chuanjian Liu", "Yehui Tang", "Yunhe Wang"], "categories": ["cs.CL", "cs.AI"], "fields_of_study": ["Computer Science"], "published_date": "2024-12-12", "url": "https://arxiv.org/abs/2412.09078", "pdf_url": "https://arxiv.org/pdf/2412.09078v5", "arxiv_id": "2412.09078", "doi": "10.48550/arXiv.2412.09078", "citation_count": 118, "influential_citation_count": 10, "has_code": true, "code_url": "https://github.com/iamhankai/Forest-of-Thought", "venue": "International Conference on Machine Learning", "quality_score": 0.5207}
{"id": "08a704aa0236cb3954cf3f844c718a03a10ab28a54e508e0505f1586d329fd99", "sources": ["arxiv", "semantic_scholar"], "title": "Understanding Hidden Computations in Chain-of-Thought Reasoning", "abstract": "Chain-of-Thought (CoT) prompting has significantly enhanced the reasoning abilities of large language models. However, recent studies have shown that models can still perform complex reasoning tasks even when the CoT is replaced with filler(hidden) characters (e.g., \"...\"), leaving open questions about how models internally process and represent reasoning steps. In this paper, we investigate methods to decode these hidden characters in transformer models trained with filler CoT sequences. By analyzing layer-wise representations using the logit lens method and examining token rankings, we demonstrate that the hidden characters can be recovered without loss of performance. Our findings provide insights into the internal mechanisms of transformer models and open avenues for improving interpretability and transparency in language model reasoning.", "authors": ["Aryasomayajula Ram Bharadwaj"], "categories": ["cs.CL", "cs.LG"], "fields_of_study": ["Computer Science"], "published_date": "2024-12-05", "url": "https://arxiv.org/abs/2412.04537", "pdf_url": "https://arxiv.org/pdf/2412.04537v1", "arxiv_id": "2412.04537", "doi": "10.48550/arXiv.2412.04537", "citation_count": 4, "influential_citation_count": 0, "has_code": false, "code_url": null, "venue": "arXiv.org", "quality_score": 0.1747}
{"id": "b9d95409517675bdf8c7c6a2fd494d17418e8640d8682dc8cc6d200646efbbef", "sources": ["arxiv", "semantic_scholar"], "title": "On the Impact of Fine-Tuning on Chain-of-Thought Reasoning", "abstract": "Large language models have emerged as powerful tools for general intelligence, showcasing advanced natural language processing capabilities that find applications across diverse domains. Despite their impressive performance, recent studies have highlighted the potential for significant enhancements in LLMs' task-specific performance through fine-tuning strategies like Reinforcement Learning with Human Feedback (RLHF), supervised fine-tuning (SFT), and Quantized Low-Rank Adapters (Q-LoRA) method. However, previous works have shown that while fine-tuning offers significant performance gains, it also leads to challenges such as catastrophic forgetting and privacy and safety risks. To this end, there has been little to no work in \\textit{understanding the impact of fine-tuning on the reasoning capabilities of LLMs}. Our research investigates the effect of fine-tuning on the reasoning abilities of LLMs, addressing critical questions regarding the impact of task-specific fine-tuning on overall reasoning capabilities, the influence of fine-tuning on Chain-of-Thought (CoT) reasoning performance, and the implications for the faithfulness of CoT reasonings. By exploring these dimensions, our study shows the impact of fine-tuning on LLM reasoning capabilities, where the faithfulness of CoT reasoning, on average across four datasets, decreases, highlighting potential shifts in internal mechanisms of the LLMs resulting from fine-tuning processes.", "authors": ["Elita Lobo", "Chirag Agarwal", "Himabindu Lakkaraju"], "categories": ["cs.CL"], "fields_of_study": ["Computer Science"], "published_date": "2024-11-22", "url": "https://arxiv.org/abs/2411.15382", "pdf_url": "https://arxiv.org/pdf/2411.15382v2", "arxiv_id": "2411.15382", "doi": "10.48550/arXiv.2411.15382", "citation_count": 40, "influential_citation_count": 0, "has_code": false, "code_url": null, "venue": "North American Chapter of the Association for Computational Linguistics", "quality_score": 0.4032}
{"id": "e1ec9653228f1229bdea0e03b0670d808a750e784ae2026d8f9427efa250a6cf", "sources": ["arxiv", "semantic_scholar"], "title": "Reducing Reasoning Costs: The Path of Optimization for Chain of Thought via Sparse Attention Mechanism", "abstract": "In order to address the chain of thought in the large language model inference cost surge, this research proposes to use a sparse attention mechanism that only focuses on a few relevant tokens. The researcher constructed a new attention mechanism and used GiantRabbit trained with custom GPTs as an experimental tool. The experiment tested and compared the reasoning time, correctness score and chain of thought length of this model and o1 Preview in solving the linear algebra test questions of MIT OpenCourseWare. The results show that GiantRabbit's reasoning time and chain of thought length are significantly lower than o1 Preview. It verifies the feasibility of sparse attention mechanism for optimizing chain of thought reasoning. Detailed architectural details and experimental process have been uploaded to Github, the link is:https://github.com/brucewang123456789/GeniusTrail.git.", "authors": ["Libo Wang"], "categories": ["cs.LG"], "fields_of_study": ["Computer Science"], "published_date": "2024-11-14", "url": "https://arxiv.org/abs/2411.09111", "pdf_url": "https://arxiv.org/pdf/2411.09111v8", "arxiv_id": "2411.09111", "doi": "10.48550/arXiv.2411.09111", "citation_count": 4, "influential_citation_count": 1, "has_code": true, "code_url": "https://github.com/brucewang123456789/GeniusTrail/tree/main/Sparse%20Attention%20Mechanism", "venue": "arXiv.org", "quality_score": 0.1747}
{"id": "54d49d7c65329194facda6356395b3fc4744c37a712dba2c4f5667fe71bb2894", "sources": ["arxiv", "semantic_scholar"], "title": "Blind Spot Navigation in Large Language Model Reasoning with Thought Space Explorer", "abstract": "Large language models have shown strong reasoning capabilities through chain-structured methods such as Chain-of-Thought. Recent studies optimize thought structures by generating parallel or tree-like structures, switching between long and short reasoning modes, or aligning reasoning steps with task performance. However, these approaches mainly rely on previously generated logical directions of the chains, which ignore the unexplored regions of the solution space. Such a phenomenon is defined as blind spots, which limit the diversity and effectiveness of the reasoning process. To this end, we propose the ``Thought Space Explorer'' (TSE), a framework for navigating and expanding thought structures to overcome blind spots in LLM reasoning. Our TSE first identifies key nodes with high impact, then generates new nodes by integrating information from multiple chains. Finally, it extends new branches through connection strategies. We conduct a series of experiments on math and QA benchmarks. Compared with existing baseline methods, TSE improves the accuracy of both the final answer and intermediate reasoning steps, while maintaining a better effectiveness-efficiency trade-off for practical deployment.", "authors": ["Jinghan Zhang", "Fengran Mo", "Tharindu Cyril Weerasooriya", "Xinyue Ye", "Dongjie Wang", "Yanjie Fu", "Kunpeng Liu"], "categories": ["cs.CL"], "fields_of_study": ["Computer Science"], "published_date": "2024-10-31", "url": "https://arxiv.org/abs/2410.24155", "pdf_url": "https://arxiv.org/pdf/2410.24155v4", "arxiv_id": "2410.24155", "doi": "10.18653/v1/2026.findings-eacl.191", "citation_count": 1, "influential_citation_count": 0, "has_code": false, "code_url": null, "venue": "Conference of the European Chapter of the Association for Computational Linguistics", "quality_score": 0.0753}
{"id": "6df8199fe0bfdb2e02decb837ac9f1a8d91787b17d4bc4f68398a9a700247d8d", "sources": ["arxiv", "semantic_scholar"], "title": "Can Language Models Perform Robust Reasoning in Chain-of-thought Prompting with Noisy Rationales?", "abstract": "This paper investigates an under-explored challenge in large language models (LLMs): chain-of-thought prompting with noisy rationales, which include irrelevant or inaccurate reasoning thoughts within examples used for in-context learning. We construct NoRa dataset that is tailored to evaluate the robustness of reasoning in the presence of noisy rationales. Our findings on NoRa dataset reveal a prevalent vulnerability to such noise among current LLMs, with existing robust methods like self-correction and self-consistency showing limited efficacy. Notably, compared to prompting with clean rationales, base LLM drops by 1.4%-19.8% in accuracy with irrelevant thoughts and more drastically by 2.2%-40.4% with inaccurate thoughts. Addressing this challenge necessitates external supervision that should be accessible in practice. Here, we propose the method of contrastive denoising with noisy chain-of-thought (CD-CoT). It enhances LLMs' denoising-reasoning capabilities by contrasting noisy rationales with only one clean rationale, which can be the minimal requirement for denoising-purpose prompting. This method follows a principle of exploration and exploitation: (1) rephrasing and selecting rationales in the input space to achieve explicit denoising and (2) exploring diverse reasoning paths and voting on answers in the output space. Empirically, CD-CoT demonstrates an average improvement of 17.8% in accuracy over the base model and shows significantly stronger denoising capabilities than baseline methods. The source code is publicly available at: https://github.com/tmlr-group/NoisyRationales.", "authors": ["Zhanke Zhou", "Rong Tao", "Jianing Zhu", "Yiwen Luo", "Zengmao Wang", "Bo Han"], "categories": ["cs.CL", "cs.LG"], "fields_of_study": ["Computer Science"], "published_date": "2024-10-31", "url": "https://arxiv.org/abs/2410.23856", "pdf_url": "https://arxiv.org/pdf/2410.23856v1", "arxiv_id": "2410.23856", "doi": "10.48550/arXiv.2410.23856", "citation_count": 67, "influential_citation_count": 2, "has_code": true, "code_url": "https://github.com/tmlr-group/NoisyRationales", "venue": "Neural Information Processing Systems", "quality_score": 0.4581}
{"id": "74bdbc8c77eaa476f9345f1862bb0a53814658be2d8fd33e6a518282b0f51865", "sources": ["arxiv", "semantic_scholar"], "title": "Markov Chain of Thought for Efficient Mathematical Reasoning", "abstract": "Chain of Thought (CoT) of multi-step benefits from the logical structure of the reasoning steps and task-specific actions, significantly enhancing the mathematical reasoning capabilities of large language models. As the prevalence of long CoT, the number of reasoning steps exceeds manageable token limits and leads to higher computational demands. Inspired by the fundamental logic of human cognition, \"derive, then reduce\", we conceptualize the standard multi-step CoT as a novel Markov Chain of Thought (MCoT). In this study, we consider the mathematical reasoning task, defining each reasoning step as text accompanied by a Python code snippet. To facilitate a longer reasoning path, self-correction is enabled through interactions with the code interpreter. Our MCoT aims to compress previous reasoning steps into a simplified question, enabling efficient next-step inference without relying on a lengthy KV cache. In our experiments, we curate the $\\texttt{MCoTInstruct}$ dataset, and the empirical results indicate that MCoT not only significantly enhances efficiency but also maintains comparable accuracy. While much remains to be explored, this work paves the way for exploring the long CoT reasoning abilities of LLMs. The code is available at https://github.com/james-yw/Markov-Chain-of-Thought", "authors": ["Wen Yang", "Minpeng Liao", "Kai Fan"], "categories": ["cs.AI", "cs.CL"], "fields_of_study": ["Computer Science"], "published_date": "2024-10-23", "url": "https://arxiv.org/abs/2410.17635", "pdf_url": "https://arxiv.org/pdf/2410.17635v2", "arxiv_id": "2410.17635", "doi": "10.48550/arXiv.2410.17635", "citation_count": 21, "influential_citation_count": 0, "has_code": true, "code_url": "https://github.com/james-yw/Markov-Chain-of-Thought", "venue": "North American Chapter of the Association for Computational Linguistics", "quality_score": 0.3356}
{"id": "799965f153f22ab96db828770059e52ce2cf3f8b1dbbead9859fd38dfbcb3fa2", "sources": ["arxiv", "semantic_scholar"], "title": "Theorem-Validated Reverse Chain-of-Thought Problem Generation for Geometric Reasoning", "abstract": "Large Multimodal Models (LMMs) face limitations in geometric reasoning due to insufficient Chain of Thought (CoT) image-text training data. While existing approaches leverage template-based or LLM-assisted methods for geometric CoT data creation, they often face challenges in achieving both diversity and precision. To bridge this gap, we introduce a two-stage Theorem-Validated Reverse Chain-of-Thought Reasoning Synthesis (TR-CoT) framework. The first stage, TR-Engine, synthesizes theorem-grounded geometric diagrams with structured descriptions and properties. The second stage, TR-Reasoner, employs reverse reasoning to iteratively refine question-answer pairs by cross-validating geometric properties and description fragments. Our approach expands theorem-type coverage, corrects long-standing misunderstandings, and enhances geometric reasoning. Fine-grained CoT improves theorem understanding and increases logical consistency by 24.5%. Our best models surpass the baselines in MathVista and GeoQA by 10.1% and 4.7%, outperforming advanced closed-source models like GPT-4o.", "authors": ["Linger Deng", "Linghao Zhu", "Yuliang Liu", "Yu Wang", "Qunyi Xie", "Jingjing Wu", "Gang Zhang", "Yingying Zhu", "Xiang Bai"], "categories": ["cs.AI", "cs.CV"], "fields_of_study": ["Computer Science"], "published_date": "2024-10-23", "url": "https://arxiv.org/abs/2410.17885", "pdf_url": "https://arxiv.org/pdf/2410.17885v4", "arxiv_id": "2410.17885", "doi": "10.18653/v1/2025.emnlp-main.38", "citation_count": 15, "influential_citation_count": 2, "has_code": false, "code_url": null, "venue": "Conference on Empirical Methods in Natural Language Processing", "quality_score": 0.301}
{"id": "5751b85d10f299cac0c5139527a359e040352e2d920af8348f2659e108dcef7e", "sources": ["arxiv", "semantic_scholar"], "title": "Improve Vision Language Model Chain-of-thought Reasoning", "abstract": "Chain-of-thought (CoT) reasoning in vision language models (VLMs) is crucial for improving interpretability and trustworthiness. However, current training recipes lack robust CoT reasoning data, relying on datasets dominated by short annotations with minimal rationales. In this work, we show that training VLM on short answers does not generalize well to reasoning tasks that require more detailed responses. To address this, we propose a two-fold approach. First, we distill rationales from GPT-4o model to enrich the training data and fine-tune VLMs, boosting their CoT performance. Second, we apply reinforcement learning to further calibrate reasoning quality. Specifically, we construct positive (correct) and negative (incorrect) pairs of model-generated reasoning chains, by comparing their predictions with annotated short answers. Using this pairwise data, we apply the Direct Preference Optimization algorithm to refine the model's reasoning abilities. Our experiments demonstrate significant improvements in CoT reasoning on benchmark datasets and better generalization to direct answer prediction as well. This work emphasizes the importance of incorporating detailed rationales in training and leveraging reinforcement learning to strengthen the reasoning capabilities of VLMs.", "authors": ["Ruohong Zhang", "Bowen Zhang", "Yanghao Li", "Haotian Zhang", "Zhiqing Sun", "Zhe Gan", "Yinfei Yang", "Ruoming Pang", "Yiming Yang"], "categories": ["cs.AI", "cs.CV"], "fields_of_study": ["Computer Science"], "published_date": "2024-10-21", "url": "https://arxiv.org/abs/2410.16198", "pdf_url": "https://arxiv.org/pdf/2410.16198v1", "arxiv_id": "2410.16198", "doi": "10.48550/arXiv.2410.16198", "citation_count": 144, "influential_citation_count": 3, "has_code": false, "code_url": null, "venue": "Annual Meeting of the Association for Computational Linguistics", "quality_score": 0.5403}
{"id": "df4d39e3e658a55d774668d46d9db85bd812f0c4485f81b243b723d5f43875e6", "sources": ["arxiv", "semantic_scholar"], "title": "A Theoretical Understanding of Chain-of-Thought: Coherent Reasoning and Error-Aware Demonstration", "abstract": "Few-shot Chain-of-Thought (CoT) prompting has demonstrated strong performance in improving the reasoning capabilities of large language models (LLMs). While theoretical investigations have been conducted to understand CoT, the underlying transformer used in these studies isolates the CoT reasoning process into separated in-context learning steps (Stepwise ICL). In this work, we theoretically show that, compared to Stepwise ICL, the transformer gains better error correction ability and more accurate predictions if the reasoning from earlier steps (Coherent CoT) is integrated. Given that this coherent reasoning changes the behavior of the transformer, we further investigate the sensitivity of the transformer with Coherent CoT when the demonstration examples are corrupted at the inference stage. Our theoretical results indicate that the transformer is more sensitive to errors in intermediate reasoning steps than the final outcome. Building upon this observation, we propose an improvement on CoT by incorporating both correct and incorrect reasoning paths in the demonstration. Our experiments validate the effectiveness of the proposed approach.", "authors": ["Yingqian Cui", "Pengfei He", "Xianfeng Tang", "Qi He", "Chen Luo", "Jiliang Tang", "Yue Xing"], "categories": ["cs.CL", "cs.AI", "cs.LG", "stat.ML"], "fields_of_study": ["Computer Science", "Mathematics"], "published_date": "2024-10-21", "url": "https://arxiv.org/abs/2410.16540", "pdf_url": "https://arxiv.org/pdf/2410.16540v1", "arxiv_id": "2410.16540", "doi": "10.48550/arXiv.2410.16540", "citation_count": 16, "influential_citation_count": 1, "has_code": false, "code_url": null, "venue": "International Conference on Artificial Intelligence and Statistics", "quality_score": 0.3076}
{"id": "14659b194e3c956ac5cc841bd201f25ffe6c0c43850f96723df598a56adfb531", "sources": ["arxiv", "semantic_scholar"], "title": "MiCEval: Unveiling Multimodal Chain of Thought's Quality via Image Description and Reasoning Steps", "abstract": "Multimodal Chain of Thought (MCoT) is a popular prompting strategy for improving the performance of multimodal large language models (MLLMs) across a range of complex reasoning tasks. Despite its popularity, there is a notable absence of automated methods for evaluating the quality of reasoning steps in MCoT. To address this gap, we propose Multimodal Chain-of-Thought Evaluation (MiCEval), a framework designed to assess the correctness of reasoning chains by evaluating the quality of both the description and each reasoning step. The evaluation of the description component focuses on the accuracy of the image descriptions, while the reasoning step evaluates the quality of each step as it is conditionally generated based on the preceding steps. MiCEval is built upon a fine-grained dataset with annotations that rate each step according to correctness, relevance, and informativeness. Extensive experiments on four state-of-the-art MLLMs show that step-wise evaluations using MiCEval align more closely with human judgments compared to existing methods based on cosine similarity or fine-tuning approaches. MiCEval datasets and code can be found in https://github.com/alenai97/MiCEval.", "authors": ["Xiongtao Zhou", "Jie He", "Lanyu Chen", "Jingyu Li", "Haojing Chen", "Víctor Gutiérrez-Basulto", "Jeff Z. Pan", "Hanjie Chen"], "categories": ["cs.CL"], "fields_of_study": ["Computer Science"], "published_date": "2024-10-18", "url": "https://arxiv.org/abs/2410.14668", "pdf_url": "https://arxiv.org/pdf/2410.14668v4", "arxiv_id": "2410.14668", "doi": "10.48550/arXiv.2410.14668", "citation_count": 10, "influential_citation_count": 0, "has_code": true, "code_url": "https://github.com/alenai97/MiCEval", "venue": "North American Chapter of the Association for Computational Linguistics", "quality_score": 0.2603}
{"id": "e542e537bcbe96a8071cf61687143bca14fd6ef980558d3990ddc298a82e9364", "sources": ["arxiv", "semantic_scholar"], "title": "Supervised Chain of Thought", "abstract": "Large Language Models (LLMs) have revolutionized natural language processing and hold immense potential for advancing Artificial Intelligence. However, the core architecture of most mainstream LLMs -- the Transformer -- has inherent limitations in computational depth, rendering them theoretically incapable of solving many reasoning tasks that demand increasingly deep computations. Chain of Thought (CoT) prompting has emerged as a technique to address these architectural limitations, as evidenced by several theoretical studies. It offers a promising approach to solving complex reasoning tasks that were previously beyond the capabilities of these models. Despite its successes, CoT and its variants (such as Tree of Thought, Graph of Thought, etc.) rely on a \"one-prompt-for-all\" approach, using a single prompt structure (e.g., \"think step by step\") for a wide range of tasks -- from counting and sorting to solving mathematical and algorithmic problems. This approach poses significant challenges for models to generate the correct reasoning steps, as the model must navigate through a vast prompt template space to find the appropriate template for each task. In this work, we build upon previous theoretical analyses of CoT to demonstrate how the one-prompt-for-all approach can negatively affect the computability of LLMs. We partition the solution search space into two: the prompt space and the answer space. Our findings show that task-specific supervision is essential for navigating the prompt space accurately and achieving optimal performance. Through experiments with state-of-the-art LLMs, we reveal a gap in reasoning performance when supervision is applied versus when it is not.", "authors": ["Xiang Zhang", "Dujian Ding"], "categories": ["cs.CL", "cs.AI"], "fields_of_study": ["Computer Science"], "published_date": "2024-10-18", "url": "https://arxiv.org/abs/2410.14198", "pdf_url": "https://arxiv.org/pdf/2410.14198v1", "arxiv_id": "2410.14198", "doi": "10.48550/arXiv.2410.14198", "citation_count": 4, "influential_citation_count": 0, "has_code": false, "code_url": null, "venue": "arXiv.org", "quality_score": 0.1747}
{"id": "43db16537b372e094ad180356e9eeb889e135cc92d76355a87a6c0b1510bd66f", "sources": ["arxiv", "semantic_scholar"], "title": "CoMAT: Chain of Mathematically Annotated Thought Improves Mathematical Reasoning", "abstract": "Mathematical reasoning remains a significant challenge for large language models (LLMs), despite progress in prompting techniques such as Chain-of-Thought (CoT). We present **Chain of Mathematically Annotated Thought (CoMAT)**, which enhances reasoning through two stages: *Symbolic Conversion* (converting natural language queries into symbolic form) and *Reasoning Execution* (deriving answers from symbolic representations). CoMAT operates entirely with a single LLM and without external solvers. Across four LLMs, CoMAT outperforms traditional CoT on six out of seven benchmarks, achieving gains of 4.48% on MMLU-Redux (MATH) and 4.58% on GaoKao MCQ. In addition to improved performance, CoMAT ensures faithfulness and verifiability, offering a transparent reasoning process for complex mathematical tasks", "authors": ["Joshua Ong Jun Leang", "Aryo Pradipta Gema", "Shay B. Cohen"], "categories": ["cs.AI", "cs.CL", "cs.LG", "cs.SC"], "fields_of_study": ["Computer Science"], "published_date": "2024-10-14", "url": "https://arxiv.org/abs/2410.10336", "pdf_url": "https://arxiv.org/pdf/2410.10336v2", "arxiv_id": "2410.10336", "doi": "10.18653/v1/2025.emnlp-main.1024", "citation_count": 16, "influential_citation_count": 1, "has_code": false, "code_url": null, "venue": "Conference on Empirical Methods in Natural Language Processing", "quality_score": 0.3076}
{"id": "c7046bade344fe1f545c31bfa1608be8b4d563a514ccba9b3de06fc785b0bd3b", "sources": ["arxiv", "semantic_scholar"], "title": "Chain-of-Sketch: Enabling Global Visual Reasoning", "abstract": "Modern vision models have achieved remarkable success in benchmarks where local features provide critical information about the target. There is now a growing interest in tackling tasks requiring more global reasoning, where local features do not provide significant information. Minsky and Papert put forward such tasks in 1969 with their connectivity study, exposing the limitations of the perceptron model. In this paper, we introduce an expanded set of global visual datasets involving graphs, strings, mazes, and image grids. We show that large vision models still struggle to learn these tasks efficiently. Similarly, state-of-the-art multi-modal LLMs perform poorly on these datasets. We explain this learning inefficiency by means of the 'globality degree' measure. To mitigate this, we propose a method called chain-of-sketch (CoS). Similar to the chain-of-thought and scratchpad techniques used in language models, CoS breaks the original task into intermediate visual steps to help learn a complex task. In addition, we show that not all CoS strategies perform equally well. Our key insight is to impose a Markovian structure on the CoS frames. This leads to the introduction of 'inductive CoS' which achieves better out-of-distribution generalization and performs well even with smaller models compared to non-inductive variants.", "authors": ["Aryo Lotfi", "Enrico Fini", "Samy Bengio", "Moin Nabi", "Emmanuel Abbe"], "categories": ["cs.LG", "cs.CV"], "fields_of_study": ["Computer Science"], "published_date": "2024-10-10", "url": "https://arxiv.org/abs/2410.08165", "pdf_url": "https://arxiv.org/pdf/2410.08165v2", "arxiv_id": "2410.08165", "doi": null, "citation_count": 2, "influential_citation_count": 0, "has_code": false, "code_url": null, "venue": null, "quality_score": 0.1193}
{"id": "69de24374167ad7debc59975c7b8b523c9d7246f726560c054018e7ca5395b8e", "sources": ["arxiv", "semantic_scholar"], "title": "Improving LLM Reasoning through Scaling Inference Computation with Collaborative Verification", "abstract": "Despite significant advancements in the general capability of large language models (LLMs), they continue to struggle with consistent and accurate reasoning, especially in complex tasks such as mathematical and code reasoning. One key limitation is that LLMs are trained primarily on correct solutions, reducing their ability to detect and learn from errors, which hampers their ability to reliably verify and rank outputs. To address this, we scale up the inference-time computation by generating multiple reasoning paths and employing verifiers to assess and rank the generated outputs by correctness. To facilitate this, we introduce a comprehensive dataset consisting of correct and incorrect solutions for math and code tasks, generated by multiple LLMs. This diverse set of solutions enables verifiers to more effectively distinguish and rank correct answers from erroneous outputs. The training methods for building verifiers were selected based on an extensive comparison of existing approaches. Moreover, to leverage the unique strengths of different reasoning strategies, we propose a novel collaborative method integrating Chain-of-Thought (CoT) and Program-of-Thought (PoT) solutions for verification. CoT provides a clear, step-by-step reasoning process that enhances interpretability, while PoT, being executable, offers a precise and error-sensitive validation mechanism. By taking both of their strengths, our approach significantly improves the accuracy and reliability of reasoning verification. Our verifiers, Math-Rev and Code-Rev, demonstrate substantial performance gains to existing LLMs, achieving state-of-the-art results on benchmarks such as GSM8k and MATH and even outperforming GPT-4o with Qwen-72B-Instruct as the reasoner.", "authors": ["Zhenwen Liang", "Ye Liu", "Tong Niu", "Xiangliang Zhang", "Yingbo Zhou", "Semih Yavuz"], "categories": ["cs.LG", "cs.AI"], "fields_of_study": ["Computer Science"], "published_date": "2024-10-05", "url": "https://arxiv.org/abs/2410.05318", "pdf_url": "https://arxiv.org/pdf/2410.05318v1", "arxiv_id": "2410.05318", "doi": "10.48550/arXiv.2410.05318", "citation_count": 42, "influential_citation_count": 2, "has_code": false, "code_url": null, "venue": "arXiv.org", "quality_score": 0.4084}
{"id": "d7c2de1f432067741589391a1a3df8d09c9b0c1fdf6097148dee231df17295e0", "sources": ["arxiv", "semantic_scholar"], "title": "TS-Reasoner: Domain-Oriented Time Series Inference Agents for Reasoning and Automated Analysis", "abstract": "Time series analysis is crucial in real-world applications, yet traditional methods focus on isolated tasks only, and recent studies on time series reasoning remain limited to either single-step inference or are constrained to natural language answers. In this work, we introduce TS-Reasoner, a domain-specialized agent designed for multi-step time series inference. By integrating large language model (LLM) reasoning with domain-specific computational tools and an error feedback loop, TS-Reasoner enables domain-informed, constraint-aware analytical workflows that combine symbolic reasoning with precise numerical analysis. We assess the system's capabilities along two axes: (1) fundamental time series understanding assessed by TimeSeriesExam and (2) complex, multi-step inference evaluated by a newly proposed dataset designed to test both compositional reasoning and computational precision in time series analysis. Experiments show that our approach outperforms standalone general-purpose LLMs in both basic time series concept understanding as well as the multi-step time series inference task, highlighting the promise of domain-specialized agents for automating real-world time series reasoning and analysis.", "authors": ["Wen Ye", "Wei Yang", "Defu Cao", "Yizhou Zhang", "Lumingyuan Tang", "Jie Cai", "Yan Liu"], "categories": ["cs.LG", "cs.AI"], "fields_of_study": ["Computer Science"], "published_date": "2024-10-05", "url": "https://arxiv.org/abs/2410.04047", "pdf_url": "https://arxiv.org/pdf/2410.04047v6", "arxiv_id": "2410.04047", "doi": null, "citation_count": 17, "influential_citation_count": 0, "has_code": false, "code_url": null, "venue": null, "quality_score": 0.3138}
{"id": "64539d8a807a9cff447ee9cdae611efe0d5e7f82782cd0336bd25b1c3a66474d", "sources": ["arxiv", "semantic_scholar"], "title": "Understanding Reasoning in Chain-of-Thought from the Hopfieldian View", "abstract": "Large Language Models have demonstrated remarkable abilities across various tasks, with Chain-of-Thought (CoT) prompting emerging as a key technique to enhance reasoning capabilities. However, existing research primarily focuses on improving performance, lacking a comprehensive framework to explain and understand the fundamental factors behind CoT's success. To bridge this gap, we introduce a novel perspective grounded in the Hopfieldian view of cognition in cognitive neuroscience. We establish a connection between CoT reasoning and key cognitive elements such as stimuli, actions, neural populations, and representation spaces. From our view, we can understand the reasoning process as the movement between these representation spaces. Building on this insight, we develop a method for localizing reasoning errors in the response of CoTs. Moreover, we propose the Representation-of-Thought (RoT) framework, which leverages the robustness of low-dimensional representation spaces to enhance the robustness of the reasoning process in CoTs. Experimental results demonstrate that RoT improves the robustness and interpretability of CoT reasoning while offering fine-grained control over the reasoning process.", "authors": ["Lijie Hu", "Liang Liu", "Shu Yang", "Xin Chen", "Zhen Tan", "Muhammad Asif Ali", "Mengdi Li", "Di Wang"], "categories": ["cs.AI", "cs.CL", "cs.LG"], "fields_of_study": ["Computer Science"], "published_date": "2024-10-04", "url": "https://arxiv.org/abs/2410.03595", "pdf_url": "https://arxiv.org/pdf/2410.03595v1", "arxiv_id": "2410.03595", "doi": "10.48550/arXiv.2410.03595", "citation_count": 10, "influential_citation_count": 1, "has_code": false, "code_url": null, "venue": "arXiv.org", "quality_score": 0.2603}
{"id": "0a74129afdb790f4a7031b2c7a435f07a1cd171b17c818976493beb49d4afb7a", "sources": ["arxiv", "semantic_scholar"], "title": "Exploring the Limitations of Mamba in COPY and CoT Reasoning", "abstract": "Transformers have become the backbone of modern Large Language Models (LLMs); however, their inference overhead grows linearly with the sequence length, posing challenges for modeling long sequences. In light of this, Mamba has attracted attention for maintaining a constant inference size, with empirical evidence demonstrating that it can match Transformer performance in sequence modeling while significantly reducing computational costs. However, an open question remains: can Mamba always bring savings while achieving performance comparable to Transformers? In this paper, we focus on analyzing the expressive ability of Mamba to perform our defined COPY operation and Chain of Thought (CoT) reasoning. First, inspired by the connection between Mamba and linear attention, we show that constant-sized Mamba may struggle to perform COPY operations while Transformers can handle them more easily. However, when the size of Mamba grows linearly with the input sequence length, it can accurately perform COPY, but in this case, Mamba no longer provides overhead savings. Based on this observation, we further analyze Mamba's ability to tackle CoT tasks, which can be described by the Dynamic Programming (DP) problems. Our findings suggest that to solve arbitrary DP problems, the total cost of Mamba is still comparable to standard Transformers. However, similar to efficient Transformers, when facing DP problems with favorable properties such as locality, Mamba can provide savings in overhead. Our experiments on the copy and CoT tasks further demonstrate Mamba's limitations compared to Transformers in learning these tasks.", "authors": ["Ruifeng Ren", "Zhicong Li", "Yong Liu"], "categories": ["cs.LG", "cs.AI", "cs.CL"], "fields_of_study": ["Computer Science"], "published_date": "2024-10-04", "url": "https://arxiv.org/abs/2410.03810", "pdf_url": "https://arxiv.org/pdf/2410.03810v3", "arxiv_id": "2410.03810", "doi": "10.18653/v1/2025.emnlp-main.634", "citation_count": 7, "influential_citation_count": 0, "has_code": false, "code_url": null, "venue": "Conference on Empirical Methods in Natural Language Processing", "quality_score": 0.2258}
{"id": "bc7b3a86263bce8a5da9dd17b179261f1358657d35a50575b9b6e91ac739a50f", "sources": ["arxiv", "semantic_scholar"], "title": "Proof of Thought : Neurosymbolic Program Synthesis allows Robust and Interpretable Reasoning", "abstract": "Large Language Models (LLMs) have revolutionized natural language processing, yet they struggle with inconsistent reasoning, particularly in novel domains and complex logical sequences. This research introduces Proof of Thought, a framework that enhances the reliability and transparency of LLM outputs. Our approach bridges LLM-generated ideas with formal logic verification, employing a custom interpreter to convert LLM outputs into First Order Logic constructs for theorem prover scrutiny. Central to our method is an intermediary JSON-based Domain-Specific Language, which by design balances precise logical structures with intuitive human concepts. This hybrid representation enables both rigorous validation and accessible human comprehension of LLM reasoning processes. Key contributions include a robust type system with sort management for enhanced logical integrity, explicit representation of rules for clear distinction between factual and inferential knowledge, and a flexible architecture that allows for easy extension to various domain-specific applications. We demonstrate Proof of Thought's effectiveness through benchmarking on StrategyQA and a novel multimodal reasoning task, showing improved performance in open-ended scenarios. By providing verifiable and interpretable results, our technique addresses critical needs for AI system accountability and sets a foundation for human-in-the-loop oversight in high-stakes domains.", "authors": ["Debargha Ganguly", "Srinivasan Iyengar", "Vipin Chaudhary", "Shivkumar Kalyanaraman"], "categories": ["cs.AI", "cs.CL", "cs.LG", "cs.LO", "cs.NE"], "fields_of_study": ["Computer Science"], "published_date": "2024-09-25", "url": "https://arxiv.org/abs/2409.17270", "pdf_url": "https://arxiv.org/pdf/2409.17270v2", "arxiv_id": "2409.17270", "doi": "10.48550/arXiv.2409.17270", "citation_count": 17, "influential_citation_count": 0, "has_code": false, "code_url": null, "venue": "arXiv.org", "quality_score": 0.3138}
{"id": "10c04e8b385d2b09777436234e33717c4bbb532db209758968cfdf45a0d3b82c", "sources": ["arxiv", "semantic_scholar"], "title": "Towards Time Series Reasoning with LLMs", "abstract": "Multi-modal large language models (MLLMs) have enabled numerous advances in understanding and reasoning in domains like vision, but we have not yet seen this broad success for time-series. Although prior works on time-series MLLMs have shown promising performance in time-series forecasting, very few works show how an LLM could be used for time-series reasoning in natural language. We propose a novel multi-modal time-series LLM approach that learns generalizable information across various domains with powerful zero-shot performance. First, we train a lightweight time-series encoder on top of an LLM to directly extract time-series information. Then, we fine-tune our model with chain-of-thought augmented time-series tasks to encourage the model to generate reasoning paths. We show that our model learns a latent representation that reflects specific time-series features (e.g. slope, frequency), as well as outperforming GPT-4o on a set of zero-shot reasoning tasks on a variety of domains.", "authors": ["Winnie Chow", "Lauren Gardiner", "Haraldur T. Hallgrímsson", "Maxwell A. Xu", "Shirley You Ren"], "categories": ["cs.LG"], "fields_of_study": ["Computer Science"], "published_date": "2024-09-17", "url": "https://arxiv.org/abs/2409.11376", "pdf_url": "https://arxiv.org/pdf/2409.11376v2", "arxiv_id": "2409.11376", "doi": "10.48550/arXiv.2409.11376", "citation_count": 43, "influential_citation_count": 5, "has_code": false, "code_url": null, "venue": "arXiv.org", "quality_score": 0.4109}
{"id": "9ede8fbe2fd4fa08844c56f0bb65740fa674b1d51e1dcb13c9012edc499a70cb", "sources": ["arxiv", "semantic_scholar"], "title": "Expediting and Elevating Large Language Model Reasoning via Hidden Chain-of-Thought Decoding", "abstract": "Large language models (LLMs) have demonstrated remarkable capabilities in tasks requiring reasoning and multi-step problem-solving through the use of chain-of-thought (CoT) prompting. However, generating the full CoT process results in significantly longer output sequences, leading to increased computational costs and latency during inference. To address this challenge, we propose a novel approach to compress the CoT process through semantic alignment, enabling more efficient decoding while preserving the benefits of CoT reasoning. Our method introduces an auxiliary CoT model that learns to generate and compress the full thought process into a compact special token representation semantically aligned with the original CoT output. This compressed representation is then integrated into the input of the Hidden Chain-of-Thought (HCoT) model. The training process follows a two-stage procedure: First, the CoT model is optimized to generate the compressed token representations aligned with the ground-truth CoT outputs using a contrastive loss. Subsequently, with the CoT model parameters frozen, the HCoT model is fine-tuned to generate accurate subsequent predictions conditioned on the prefix instruction and the compressed CoT representations from the CoT model. Extensive experiments across three challenging domains - mathematical reasoning, agent invocation, and question answering - demonstrate that our semantic compression approach achieves competitive or improved performance compared to the full CoT baseline, while providing significant speedups of at least 1.5x in decoding time. Moreover, incorporating contrastive learning objectives further enhances the quality of the compressed representations, leading to better CoT prompting and improved task accuracy. Our work paves the way for more efficient exploitation of multi-step reasoning capabilities in LLMs across a wide range of applications.", "authors": ["Tianqiao Liu", "Zui Chen", "Zitao Liu", "Mi Tian", "Weiqi Luo"], "categories": ["cs.CL", "cs.AI"], "fields_of_study": ["Computer Science"], "published_date": "2024-09-13", "url": "https://arxiv.org/abs/2409.08561", "pdf_url": "https://arxiv.org/pdf/2409.08561v1", "arxiv_id": "2409.08561", "doi": "10.48550/arXiv.2409.08561", "citation_count": 14, "influential_citation_count": 2, "has_code": false, "code_url": null, "venue": "arXiv.org", "quality_score": 0.294}
{"id": "fad3ab7bf1a530b94d53f801cc12296004e30c0d544a8733c922fd2c3feb3e6c", "sources": ["arxiv", "semantic_scholar"], "title": "Self-Harmonized Chain of Thought", "abstract": "Chain-of-thought (CoT) prompting has demonstrated the capacity of large language models to perform complex reasoning through intermediate steps. While effective, current CoT methods face challenges: Zero-shot-CoT can lead to reasoning errors, and Few-shot-CoT requires labor-intensive manual demonstrations. Auto-CoT attempts to address these issues by automatically generating diverse demonstrations, but this diversity can lead to inconsistent reasoning patterns. We propose ECHO (Self-Harmonized Chain of Thought), a novel method that unifies diverse solution paths into a consistent and effective reasoning pattern. ECHO employs an iterative process to refine and harmonize automatically generated demonstrations, mitigating the limitations of existing approaches. Our comprehensive experiments across arithmetic, commonsense, and symbolic reasoning tasks demonstrate that ECHO outperforms Auto-CoT by an average of 2.8%. These findings suggest that ECHO represents a significant step towards more robust and generalizable automated reasoning in large language models.", "authors": ["Ziqi Jin", "Wei Lu"], "categories": ["cs.CL"], "fields_of_study": ["Computer Science"], "published_date": "2024-09-06", "url": "https://arxiv.org/abs/2409.04057", "pdf_url": "https://arxiv.org/pdf/2409.04057v2", "arxiv_id": "2409.04057", "doi": "10.48550/arXiv.2409.04057", "citation_count": 7, "influential_citation_count": 0, "has_code": false, "code_url": null, "venue": "North American Chapter of the Association for Computational Linguistics", "quality_score": 0.2258}
{"id": "adca2d18988fbaf03811b5a980b4ac2c720655f31958130c932c4e5794b2a017", "sources": ["arxiv", "semantic_scholar"], "title": "Strategic Chain-of-Thought: Guiding Accurate Reasoning in LLMs through Strategy Elicitation", "abstract": "The Chain-of-Thought (CoT) paradigm has emerged as a critical approach for enhancing the reasoning capabilities of large language models (LLMs). However, despite their widespread adoption and success, CoT methods often exhibit instability due to their inability to consistently ensure the quality of generated reasoning paths, leading to sub-optimal reasoning performance. To address this challenge, we propose the \\textbf{Strategic Chain-of-Thought} (SCoT), a novel methodology designed to refine LLM performance by integrating strategic knowledge prior to generating intermediate reasoning steps. SCoT employs a two-stage approach within a single prompt: first eliciting an effective problem-solving strategy, which is then used to guide the generation of high-quality CoT paths and final answers. Our experiments across eight challenging reasoning datasets demonstrate significant improvements, including a 21.05\\% increase on the GSM8K dataset and 24.13\\% on the Tracking\\_Objects dataset, respectively, using the Llama3-8b model. Additionally, we extend the SCoT framework to develop a few-shot method with automatically matched demonstrations, yielding even stronger results. These findings underscore the efficacy of SCoT, highlighting its potential to substantially enhance LLM performance in complex reasoning tasks.", "authors": ["Yu Wang", "Shiwan Zhao", "Zhihu Wang", "Heyuan Huang", "Ming Fan", "Yubo Zhang", "Zhixing Wang", "Haijun Wang", "Ting Liu"], "categories": ["cs.AI", "cs.CL", "cs.HC"], "fields_of_study": ["Computer Science"], "published_date": "2024-09-05", "url": "https://arxiv.org/abs/2409.03271", "pdf_url": "https://arxiv.org/pdf/2409.03271v1", "arxiv_id": "2409.03271", "doi": "10.48550/arXiv.2409.03271", "citation_count": 25, "influential_citation_count": 2, "has_code": false, "code_url": null, "venue": "arXiv.org", "quality_score": 0.3537}
{"id": "7107ba1905bdcb630c3dcdb38c8205671ca1f3030fd80fd76d2b14b2eb40e428", "sources": ["arxiv", "semantic_scholar"], "title": "Scaling LLM Test-Time Compute Optimally can be More Effective than Scaling Model Parameters", "abstract": "Enabling LLMs to improve their outputs by using more test-time computation is a critical step towards building generally self-improving agents that can operate on open-ended natural language. In this paper, we study the scaling of inference-time computation in LLMs, with a focus on answering the question: if an LLM is allowed to use a fixed but non-trivial amount of inference-time compute, how much can it improve its performance on a challenging prompt? Answering this question has implications not only on the achievable performance of LLMs, but also on the future of LLM pretraining and how one should tradeoff inference-time and pre-training compute. Despite its importance, little research attempted to understand the scaling behaviors of various test-time inference methods. Moreover, current work largely provides negative results for a number of these strategies. In this work, we analyze two primary mechanisms to scale test-time computation: (1) searching against dense, process-based verifier reward models; and (2) updating the model's distribution over a response adaptively, given the prompt at test time. We find that in both cases, the effectiveness of different approaches to scaling test-time compute critically varies depending on the difficulty of the prompt. This observation motivates applying a \"compute-optimal\" scaling strategy, which acts to most effectively allocate test-time compute adaptively per prompt. Using this compute-optimal strategy, we can improve the efficiency of test-time compute scaling by more than 4x compared to a best-of-N baseline. Additionally, in a FLOPs-matched evaluation, we find that on problems where a smaller base model attains somewhat non-trivial success rates, test-time compute can be used to outperform a 14x larger model.", "authors": ["Charlie Snell", "Jaehoon Lee", "Kelvin Xu", "Aviral Kumar"], "categories": ["cs.LG", "cs.CL"], "fields_of_study": ["Computer Science"], "published_date": "2024-08-06", "url": "https://arxiv.org/abs/2408.03314", "pdf_url": "https://arxiv.org/pdf/2408.03314v1", "arxiv_id": "2408.03314", "doi": "10.48550/arXiv.2408.03314", "citation_count": 1866, "influential_citation_count": 134, "has_code": false, "code_url": null, "venue": "arXiv.org", "quality_score": 1.0}
{"id": "f093b44c29cfb018b47fcc5acbe99b9e5467e0a35a083fb7851cb7cfc4f72c98", "sources": ["arxiv", "semantic_scholar"], "title": "Inference Scaling Laws: An Empirical Analysis of Compute-Optimal Inference for Problem-Solving with Language Models", "abstract": "While the scaling laws of large language models (LLMs) training have been extensively studied, optimal inference configurations of LLMs remain underexplored. We study inference scaling laws (aka test-time scaling laws) and compute-optimal inference, focusing on the trade-offs between model sizes and generating additional tokens with different inference strategies. As a first step towards understanding and designing compute-optimal inference methods, we studied cost-performance trade-offs for inference strategies such as greedy search, majority voting, best-of-$n$, weighted voting, and two different tree search algorithms, using different model sizes and compute budgets. Our findings suggest that scaling inference compute with inference strategies can be more computationally efficient than scaling model parameters. Additionally, smaller models combined with advanced inference algorithms offer Pareto-optimal trade-offs in cost and performance. For example, the Llemma-7B model, when paired with our novel tree search algorithm, consistently outperforms the Llemma-34B model across all tested inference strategies on the MATH benchmark. We hope these insights contribute to a deeper understanding of inference scaling laws (test-time scaling laws) for LLMs.", "authors": ["Yangzhen Wu", "Zhiqing Sun", "Shanda Li", "Sean Welleck", "Yiming Yang"], "categories": ["cs.AI"], "fields_of_study": ["Computer Science"], "published_date": "2024-08-01", "url": "https://arxiv.org/abs/2408.00724", "pdf_url": "https://arxiv.org/pdf/2408.00724v3", "arxiv_id": "2408.00724", "doi": null, "citation_count": 195, "influential_citation_count": 14, "has_code": false, "code_url": null, "venue": "International Conference on Learning Representations", "quality_score": 0.588}
{"id": "0dc2194f8fef2dc52eaf4563a81b9da7f6ee79c44a36b07f80dad19773c5ed42", "sources": ["arxiv", "semantic_scholar"], "title": "Robotic Control via Embodied Chain-of-Thought Reasoning", "abstract": "A key limitation of learned robot control policies is their inability to generalize outside their training data. Recent works on vision-language-action models (VLAs) have shown that the use of large, internet pre-trained vision-language models as the backbone of learned robot policies can substantially improve their robustness and generalization ability. Yet, one of the most exciting capabilities of large vision-language models in other domains is their ability to reason iteratively through complex problems. Can that same capability be brought into robotics to allow policies to improve performance by reasoning about a given task before acting? Naive use of \"chain-of-thought\" (CoT) style prompting is significantly less effective with standard VLAs because of the relatively simple training examples that are available to them. Additionally, purely semantic reasoning about sub-tasks, as is common in regular CoT, is insufficient for robot policies that need to ground their reasoning in sensory observations and the robot state. To this end, we introduce Embodied Chain-of-Thought Reasoning (ECoT) for VLAs, in which we train VLAs to perform multiple steps of reasoning about plans, sub-tasks, motions, and visually grounded features like object bounding boxes and end effector positions, before predicting the robot action. We design a scalable pipeline for generating synthetic training data for ECoT on large robot datasets. We demonstrate, that ECoT increases the absolute success rate of OpenVLA, the current strongest open-source VLA policy, by 28% across challenging generalization tasks, without any additional robot training data. Additionally, ECoT makes it easier for humans to interpret a policy's failures and correct its behavior using natural language.", "authors": ["Michał Zawalski", "William Chen", "Karl Pertsch", "Oier Mees", "Chelsea Finn", "Sergey Levine"], "categories": ["cs.RO", "cs.LG"], "fields_of_study": ["Computer Science"], "published_date": "2024-07-11", "url": "https://arxiv.org/abs/2407.08693", "pdf_url": "https://arxiv.org/pdf/2407.08693v3", "arxiv_id": "2407.08693", "doi": "10.48550/arXiv.2407.08693", "citation_count": 325, "influential_citation_count": 31, "has_code": true, "code_url": null, "venue": "Conference on Robot Learning", "quality_score": 0.7526}
{"id": "1ecc851b2692dcb6b68c685008b497b7b231f9e208495f9c78a5fc63083d8bc3", "sources": ["arxiv", "semantic_scholar"], "title": "Chain-of-Thought Augmentation with Logit Contrast for Enhanced Reasoning in Language Models", "abstract": "Rapidly increasing model scales coupled with steering methods such as chain-of-thought prompting have led to drastic improvements in language model reasoning. At the same time, models struggle with compositional generalization and are far from human performance on many reasoning-based benchmarks. Leveraging the success of chain-of-thought prompting, and also taking inspiration from context-aware decoding (CAD), we explore input-based contrasting methods to further encourage the type of reasoning induced by chain-of-thought prompting. While work remains to stabilize these results across datasets and models, the improvements we find warrant further investigation into input-based steering methods for context-aware reasoning.", "authors": ["Jay Shim", "Grant Kruttschnitt", "Alyssa Ma", "Daniel Kim", "Benjamin Chek", "Athul Anand", "Kevin Zhu", "Sean O'Brien"], "categories": ["cs.CL"], "fields_of_study": ["Computer Science"], "published_date": "2024-07-04", "url": "https://arxiv.org/abs/2407.03600", "pdf_url": "https://arxiv.org/pdf/2407.03600v2", "arxiv_id": "2407.03600", "doi": null, "citation_count": 0, "influential_citation_count": 0, "has_code": false, "code_url": null, "venue": null, "quality_score": 0.0}
{"id": "c03b36b5ae932dff042108e09e130d6e93e80156b5ef54dac4dc4f6d8b306552", "sources": ["arxiv", "semantic_scholar"], "title": "Fine-Tuning on Diverse Reasoning Chains Drives Within-Inference CoT Refinement in LLMs", "abstract": "Requiring a large language model (LLM) to generate intermediary reasoning steps, known as Chain of Thought (CoT), has been shown to be an effective way of boosting performance. Previous approaches have focused on generating multiple independent CoTs, combining them through ensembling or other post-hoc strategies to enhance reasoning. In this work, we introduce a novel approach where LLMs are fine-tuned to generate a sequence of Diverse Chains of Thought (DCoT) within a single inference step, which is fundamentally different from prior work that primarily operate on parallel CoT generations. DCoT allows LLMs to gain the ability to perform within-inference refinement of reasoning chains without requiring external feedback. Through a rigorous set of experiments spanning a wide range of tasks that require various reasoning types, we show that fine-tuning on DCoT improves performance over the CoT baseline across model families and scales (1.3B to 70B). These improvements are particularly impactful for tasks with a large result state space, such as those involving numeric answers. Our work is also significant because both quantitative analyses and manual evaluations reveal the observed gains stem from the models' ability to refine an initial reasoning chain by generating a second, improved chain within the same inference step, demonstrating previously elusive self-improvement. Our code and data are publicly available at https://github.com/UKPLab/acl2025-diverse-cot.", "authors": ["Haritz Puerto", "Tilek Chubakov", "Xiaodan Zhu", "Harish Tayyar Madabushi", "Iryna Gurevych"], "categories": ["cs.CL"], "fields_of_study": ["Computer Science"], "published_date": "2024-07-03", "url": "https://arxiv.org/abs/2407.03181", "pdf_url": "https://arxiv.org/pdf/2407.03181v2", "arxiv_id": "2407.03181", "doi": "10.18653/v1/2025.acl-long.191", "citation_count": 14, "influential_citation_count": 0, "has_code": true, "code_url": "https://github.com/UKPLab/acl2025-diverse-cot", "venue": "Annual Meeting of the Association for Computational Linguistics", "quality_score": 0.294}
{"id": "056c07025de4e43407db9647c7109cb0eb52004a911705dfea6561fed7b23c8f", "sources": ["arxiv", "semantic_scholar"], "title": "On the Representational Capacity of Neural Language Models with Chain-of-Thought Reasoning", "abstract": "The performance of modern language models (LMs) has been improved by chain-of-thought (CoT) reasoning, i.e., the process of generating intermediate results that guide the model towards a final answer. A possible explanation for this improvement is that CoT reasoning extends an LM's computational power, as RNNs and transformers with additional scratch space are known to be Turing complete. Comparing LMs to Turing machines, however, introduces a category error - Turing machines decide language membership, whereas LMs define distributions over strings. To bridge this gap, we formalize CoT reasoning in a probabilistic setting. We present several results on the representational capacity of recurrent and transformer LMs with CoT reasoning, showing that they can represent the same family of distributions over strings as probabilistic Turing machines.", "authors": ["Franz Nowak", "Anej Svete", "Alexandra Butoi", "Ryan Cotterell"], "categories": ["cs.CL", "cs.FL"], "fields_of_study": ["Computer Science"], "published_date": "2024-06-20", "url": "https://arxiv.org/abs/2406.14197", "pdf_url": "https://arxiv.org/pdf/2406.14197v2", "arxiv_id": "2406.14197", "doi": "10.48550/arXiv.2406.14197", "citation_count": 29, "influential_citation_count": 3, "has_code": false, "code_url": null, "venue": "Annual Meeting of the Association for Computational Linguistics", "quality_score": 0.3693}
{"id": "3a373ddf4dd3b29564442b508f288a388b4fe735f03d8875f233d746f02ff4bf", "sources": ["arxiv", "semantic_scholar"], "title": "LLMs as Models for Analogical Reasoning", "abstract": "Analogical reasoning -- the capacity to identify and map structural relationships between different domains -- is fundamental to human cognition and learning. Recent studies have shown that large language models (LLMs) can sometimes match humans in analogical reasoning tasks, opening the possibility that analogical reasoning might emerge from domain-general processes. However, it is still debated whether these emergent capacities are largely superficial and limited to simple relations seen during training or whether they encompass the flexible representational and mapping capabilities which are the focus of leading cognitive models of analogy. In this study, we introduce novel analogical reasoning tasks that require participants to map between semantically contentful words and sequences of letters and other abstract characters. This task necessitates the ability to flexibly re-represent rich semantic information -- an ability which is known to be central to human analogy but which is thus far not well captured by existing cognitive theories and models. We assess the performance of both human participants and LLMs on tasks focusing on reasoning from semantic structure and semantic content, introducing variations that test the robustness of their analogical inferences. Advanced LLMs match human performance across several conditions, though humans and LLMs respond differently to certain task variations and semantic distractors. Our results thus provide new evidence that LLMs might offer a how-possibly explanation of human analogical reasoning in contexts that are not yet well modeled by existing theories, but that even today's best models are unlikely to yield how-actually explanations.", "authors": ["Sam Musker", "Alex Duchnowski", "Raphaël Millière", "Ellie Pavlick"], "categories": ["cs.CL"], "fields_of_study": ["Computer Science"], "published_date": "2024-06-19", "url": "https://arxiv.org/abs/2406.13803", "pdf_url": "https://arxiv.org/pdf/2406.13803v3", "arxiv_id": "2406.13803", "doi": "10.1016/j.jml.2025.104676", "citation_count": 21, "influential_citation_count": 0, "has_code": false, "code_url": null, "venue": "Journal of Memory and Language", "quality_score": 0.3356}
{"id": "1b043f1ce1dcab811d3e91d374240a58edd6468b129b93adb7fe0f22e254bd23", "sources": ["arxiv", "semantic_scholar"], "title": "Chain of Preference Optimization: Improving Chain-of-Thought Reasoning in LLMs", "abstract": "The recent development of chain-of-thought (CoT) decoding has enabled large language models (LLMs) to generate explicit logical reasoning paths for complex problem-solving. However, research indicates that these paths are not always deliberate and optimal. The tree-of-thought (ToT) method employs tree-searching to extensively explore the reasoning space and find better reasoning paths that CoT decoding might overlook. This deliberation, however, comes at the cost of significantly increased inference complexity. In this work, we demonstrate that fine-tuning LLMs leveraging the search tree constructed by ToT allows CoT to achieve similar or better performance, thereby avoiding the substantial inference burden. This is achieved through Chain of Preference Optimization (CPO), where LLMs are fine-tuned to align each step of the CoT reasoning paths with those of ToT using the inherent preference information in the tree-search process. Extensive experimental results show that CPO significantly improves LLM performance in solving a variety of complex problems, including question answering, fact verification, and arithmetic reasoning, demonstrating its effectiveness. Our code is available at https://github.com/sail-sg/CPO.", "authors": ["Xuan Zhang", "Chao Du", "Tianyu Pang", "Qian Liu", "Wei Gao", "Min Lin"], "categories": ["cs.CL", "cs.LG"], "fields_of_study": ["Computer Science"], "published_date": "2024-06-13", "url": "https://arxiv.org/abs/2406.09136", "pdf_url": "https://arxiv.org/pdf/2406.09136v2", "arxiv_id": "2406.09136", "doi": "10.48550/arXiv.2406.09136", "citation_count": 155, "influential_citation_count": 10, "has_code": true, "code_url": "https://github.com/sail-sg/CPO", "venue": "Neural Information Processing Systems", "quality_score": 0.5483}
{"id": "040d58e948d13d1c42df2ece358ff0d35724c67f4f523d0e1621599377d898ea", "sources": ["arxiv", "semantic_scholar"], "title": "FairCoT: Enhancing Fairness in Text-to-Image Generation via Chain of Thought Reasoning with Multimodal Large Language Models", "abstract": "In the domain of text-to-image generative models, biases inherent in training datasets often propagate into generated content, posing significant ethical challenges, particularly in socially sensitive contexts. We introduce FairCoT, a novel framework that enhances fairness in text to image models through Chain of Thought (CoT) reasoning within multimodal generative large language models. FairCoT employs iterative CoT refinement to systematically mitigate biases, and dynamically adjusts textual prompts in real time, ensuring diverse and equitable representation in generated images. By integrating iterative reasoning processes, FairCoT addresses the limitations of zero shot CoT in sensitive scenarios, balancing creativity with ethical responsibility. Experimental evaluations across popular text-to-image systems including DALLE and various Stable Diffusion variants, demonstrate that FairCoT significantly enhances fairness and diversity without sacrificing image quality or semantic fidelity. By combining robust reasoning, lightweight deployment, and extensibility to multiple models, FairCoT represents a promising step toward more socially responsible and transparent AI driven content generation.", "authors": ["Zahraa Al Sahili", "Ioannis Patras", "Matthew Purver"], "categories": ["cs.LG", "cs.AI", "cs.CV"], "fields_of_study": ["Computer Science"], "published_date": "2024-06-13", "url": "https://arxiv.org/abs/2406.09070", "pdf_url": "https://arxiv.org/pdf/2406.09070v4", "arxiv_id": "2406.09070", "doi": "10.18653/v1/2025.findings-emnlp.42", "citation_count": 4, "influential_citation_count": 0, "has_code": false, "code_url": null, "venue": "Conference on Empirical Methods in Natural Language Processing", "quality_score": 0.1747}
{"id": "605088a96f8252fd772042c256815397384c8f915eed65ed71bbcafdc6eafca8", "sources": ["arxiv", "semantic_scholar"], "title": "Buffer of Thoughts: Thought-Augmented Reasoning with Large Language Models", "abstract": "We introduce Buffer of Thoughts (BoT), a novel and versatile thought-augmented reasoning approach for enhancing accuracy, efficiency and robustness of large language models (LLMs). Specifically, we propose meta-buffer to store a series of informative high-level thoughts, namely thought-template, distilled from the problem-solving processes across various tasks. Then for each problem, we retrieve a relevant thought-template and adaptively instantiate it with specific reasoning structures to conduct efficient reasoning. To guarantee the scalability and stability, we further propose buffer-manager to dynamically update the meta-buffer, thus enhancing the capacity of meta-buffer as more tasks are solved. We conduct extensive experiments on 10 challenging reasoning-intensive tasks, and achieve significant performance improvements over previous SOTA methods: 11% on Game of 24, 20% on Geometric Shapes and 51% on Checkmate-in-One. Further analysis demonstrate the superior generalization ability and model robustness of our BoT, while requiring only 12% of the cost of multi-query prompting methods (e.g., tree/graph of thoughts) on average. Notably, we find that our Llama3-8B+BoT has the potential to surpass Llama3-70B model. Our project is available at: https://github.com/YangLing0818/buffer-of-thought-llm", "authors": ["Ling Yang", "Zhaochen Yu", "Tianjun Zhang", "Shiyi Cao", "Minkai Xu", "Wentao Zhang", "Joseph E. Gonzalez", "Bin Cui"], "categories": ["cs.CL"], "fields_of_study": ["Computer Science"], "published_date": "2024-06-06", "url": "https://arxiv.org/abs/2406.04271", "pdf_url": "https://arxiv.org/pdf/2406.04271v2", "arxiv_id": "2406.04271", "doi": "10.48550/arXiv.2406.04271", "citation_count": 112, "influential_citation_count": 10, "has_code": true, "code_url": "https://github.com/YangLing0818/buffer-of-thought-llm", "venue": "Neural Information Processing Systems", "quality_score": 0.5207}
{"id": "22415ed66ac4e2da9b87c9a654bbf1ecb53307bff42c2cb9a27613549e7dbf6e", "sources": ["arxiv", "semantic_scholar"], "title": "RATT: A Thought Structure for Coherent and Correct LLM Reasoning", "abstract": "Large Language Models (LLMs) gain substantial reasoning and decision-making capabilities from thought structures. However, existing methods such as Tree of Thought and Retrieval Augmented Thoughts often fall short in complex tasks due to the limitations of insufficient local retrieval of factual knowledge and inadequate global selection of strategies. These limitations make it challenging for these methods to balance factual accuracy and comprehensive logical optimization effectively. To address these limitations, we introduce the Retrieval Augmented Thought Tree (RATT), a novel thought structure that considers both overall logical soundness and factual correctness at each step of the thinking process. Specifically, at every point of a thought branch, RATT performs planning and lookahead to explore and evaluate multiple potential reasoning steps, and integrate the fact-checking ability of Retrieval-Augmented Generation (RAG) with LLM's ability to assess overall strategy. Through this combination of factual knowledge and strategic feasibility, the RATT adjusts and integrates the thought tree structure to search for the most promising branches within the search space. This thought structure significantly enhances the model's coherence in logical inference and efficiency in decision-making, and thus increases the limit of the capacity of LLM to generate reliable inferences and decisions based on thought structures. A broad range of experiments on different types of tasks showcases that the RATT structure significantly outperforms existing methods in factual correctness and logical coherence.", "authors": ["Jinghan Zhang", "Xiting Wang", "Weijieying Ren", "Lu Jiang", "Dongjie Wang", "Kunpeng Liu"], "categories": ["cs.CL"], "fields_of_study": ["Computer Science"], "published_date": "2024-06-04", "url": "https://arxiv.org/abs/2406.02746", "pdf_url": "https://arxiv.org/pdf/2406.02746v5", "arxiv_id": "2406.02746", "doi": "10.48550/arXiv.2406.02746", "citation_count": 55, "influential_citation_count": 2, "has_code": false, "code_url": null, "venue": "AAAI Conference on Artificial Intelligence", "quality_score": 0.437}
{"id": "c8a92b165776445a9efa0ca6de711edd47e0863227ce14f4976f78febd3c7416", "sources": ["arxiv", "semantic_scholar"], "title": "Preemptive Answer \"Attacks\" on Chain-of-Thought Reasoning", "abstract": "Large language models (LLMs) showcase impressive reasoning capabilities when coupled with Chain-of-Thought (CoT) prompting. However, the robustness of this approach warrants further investigation. In this paper, we introduce a novel scenario termed preemptive answers, where the LLM obtains an answer before engaging in reasoning. This situation can arise inadvertently or induced by malicious users by prompt injection attacks. Experiments reveal that preemptive answers significantly impair the model's reasoning capability across various CoT methods and a broad spectrum of datasets. To bolster the robustness of reasoning, we propose two measures aimed at mitigating this issue to some extent.", "authors": ["Rongwu Xu", "Zehan Qi", "Wei Xu"], "categories": ["cs.CL", "cs.AI", "cs.CR"], "fields_of_study": ["Computer Science"], "published_date": "2024-05-31", "url": "https://arxiv.org/abs/2405.20902", "pdf_url": "https://arxiv.org/pdf/2405.20902v1", "arxiv_id": "2405.20902", "doi": "10.48550/arXiv.2405.20902", "citation_count": 25, "influential_citation_count": 4, "has_code": false, "code_url": null, "venue": "Annual Meeting of the Association for Computational Linguistics", "quality_score": 0.3537}
{"id": "d13fdee1fafc6f75a6166c90be33b7668c9aa2a45fba40989e0e723d4e000602", "sources": ["arxiv", "semantic_scholar"], "title": "Faithful Logical Reasoning via Symbolic Chain-of-Thought", "abstract": "While the recent Chain-of-Thought (CoT) technique enhances the reasoning ability of large language models (LLMs) with the theory of mind, it might still struggle in handling logical reasoning that relies much on symbolic expressions and rigid deducing rules. To strengthen the logical reasoning capability of LLMs, we propose a novel Symbolic Chain-of-Thought, namely SymbCoT, a fully LLM-based framework that integrates symbolic expressions and logic rules with CoT prompting. Technically, building upon an LLM, SymbCoT 1) first translates the natural language context into the symbolic format, and then 2) derives a step-by-step plan to solve the problem with symbolic logical rules, 3) followed by a verifier to check the translation and reasoning chain. Via thorough evaluations on 5 standard datasets with both First-Order Logic and Constraint Optimization symbolic expressions, SymbCoT shows striking improvements over the CoT method consistently, meanwhile refreshing the current state-of-the-art performances. We further demonstrate that our system advances in more faithful, flexible, and explainable logical reasoning. To our knowledge, this is the first to combine symbolic expressions and rules into CoT for logical reasoning with LLMs. Code is open at https://github.com/Aiden0526/SymbCoT.", "authors": ["Jundong Xu", "Hao Fei", "Liangming Pan", "Qian Liu", "Mong-Li Lee", "Wynne Hsu"], "categories": ["cs.CL"], "fields_of_study": ["Computer Science"], "published_date": "2024-05-28", "url": "https://arxiv.org/abs/2405.18357", "pdf_url": "https://arxiv.org/pdf/2405.18357v2", "arxiv_id": "2405.18357", "doi": "10.48550/arXiv.2405.18357", "citation_count": 171, "influential_citation_count": 10, "has_code": true, "code_url": "https://github.com/Aiden0526/SymbCoT", "venue": "Annual Meeting of the Association for Computational Linguistics", "quality_score": 0.5589}
{"id": "3ca03dc53abeaba33badc351f7d598fea3b9ca3b0c82d491f326f891cb251c71", "sources": ["arxiv", "semantic_scholar"], "title": "Enhancing Chain of Thought Prompting in Large Language Models via Reasoning Patterns", "abstract": "Chain of Thought (CoT) prompting can encourage language models to engage in multi-step logical reasoning. The quality of the provided demonstrations significantly influences the success of downstream inference tasks. Current unsupervised CoT methods primarily select examples based on the semantics of the questions, which can introduce noise and lack interpretability. In this paper, we propose leveraging reasoning patterns to enhance CoT prompting effectiveness. Reasoning patterns represent the process by which language models arrive at their final results. By utilizing prior knowledge and prompt-based methods from large models, we first construct task-specific pattern sets. We then select diverse demonstrations based on different reasoning patterns. This approach not only mitigates the impact of noise but also provides explicit interpretability to help us understand the mechanisms of CoT. Extensive experiments demonstrate that our method is more robust and consistently leads to improvements across various reasoning tasks.", "authors": ["Yufeng Zhang", "Xuepeng Wang", "Lingxiang Wu", "Jinqiao Wang"], "categories": ["cs.CL"], "fields_of_study": ["Computer Science"], "published_date": "2024-04-23", "url": "https://arxiv.org/abs/2404.14812", "pdf_url": "https://arxiv.org/pdf/2404.14812v2", "arxiv_id": "2404.14812", "doi": "10.1609/aaai.v39i24.34793", "citation_count": 27, "influential_citation_count": 1, "has_code": false, "code_url": null, "venue": "AAAI Conference on Artificial Intelligence", "quality_score": 0.3618}
{"id": "5e577b5a7c82e58e578f6bebcda09f7cef490a3ca5605cd288b10f75251fea8a", "sources": ["arxiv", "semantic_scholar"], "title": "Mitigating Misleading Chain-of-Thought Reasoning with Selective Filtering", "abstract": "Large language models have manifested remarkable capabilities by leveraging chain-of-thought (CoT) reasoning techniques to solve intricate questions through step-by-step reasoning chains. Despite its success, the efficacy of such reasoning is inherently contingent upon the quality of CoT. However, flawless CoT reasoning cannot be guaranteed due to the presence of indecomposable questions and the potential for erroneous reasoning chains, particularly in the case of small-scale language models. To tackle this challenge, we propose a novel approach called the selective filtering reasoner (SelF-Reasoner) that assesses the entailment relationship between the question and the candidate reasoning chain. Then, we proceed with CoT reasoning when the reasoning chain demonstrates confidence; otherwise, we opt to predict the answer directly. SelF-Reasoner improves the fine-tuned T5 baseline consistently over the ScienceQA, ECQA, and LastLetter tasks. Code is available at \\texttt{https://github.com/LibroWu/SelF-Reasoner}.", "authors": ["Yexin Wu", "Zhuosheng Zhang", "Hai Zhao"], "categories": ["cs.CL", "cs.AI"], "fields_of_study": ["Computer Science"], "published_date": "2024-03-28", "url": "https://arxiv.org/abs/2403.19167", "pdf_url": "https://arxiv.org/pdf/2403.19167v1", "arxiv_id": "2403.19167", "doi": "10.48550/arXiv.2403.19167", "citation_count": 11, "influential_citation_count": 0, "has_code": true, "code_url": "https://github.com/LibroWu/SelF-Reasoner}", "venue": "International Conference on Language Resources and Evaluation", "quality_score": 0.2698}
{"id": "ce7e156fec1fe9e3800ece307b9c4404e7cbc2b69cb0d7d78e34f4dc41d8f7f3", "sources": ["arxiv", "semantic_scholar"], "title": "DriveCoT: Integrating Chain-of-Thought Reasoning with End-to-End Driving", "abstract": "End-to-end driving has made significant progress in recent years, demonstrating benefits such as system simplicity and competitive driving performance under both open-loop and closed-loop settings. Nevertheless, the lack of interpretability and controllability in its driving decisions hinders real-world deployment for end-to-end driving systems. In this paper, we collect a comprehensive end-to-end driving dataset named DriveCoT, leveraging the CARLA simulator. It contains sensor data, control decisions, and chain-of-thought labels to indicate the reasoning process. We utilize the challenging driving scenarios from the CARLA leaderboard 2.0, which involve high-speed driving and lane-changing, and propose a rule-based expert policy to control the vehicle and generate ground truth labels for its reasoning process across different driving aspects and the final decisions. This dataset can serve as an open-loop end-to-end driving benchmark, enabling the evaluation of accuracy in various chain-of-thought aspects and the final decision. In addition, we propose a baseline model called DriveCoT-Agent, trained on our dataset, to generate chain-of-thought predictions and final decisions. The trained model exhibits strong performance in both open-loop and closed-loop evaluations, demonstrating the effectiveness of our proposed dataset.", "authors": ["Tianqi Wang", "Enze Xie", "Ruihang Chu", "Zhenguo Li", "Ping Luo"], "categories": ["cs.CV", "cs.RO"], "fields_of_study": ["Computer Science"], "published_date": "2024-03-25", "url": "https://arxiv.org/abs/2403.16996", "pdf_url": "https://arxiv.org/pdf/2403.16996v1", "arxiv_id": "2403.16996", "doi": "10.48550/arXiv.2403.16996", "citation_count": 57, "influential_citation_count": 3, "has_code": false, "code_url": null, "venue": "arXiv.org", "quality_score": 0.4409}
{"id": "b208b636dbaa34ffe8fb38c739988fe772eb42e2c829c8623ad2bafd62616b8d", "sources": ["arxiv", "semantic_scholar"], "title": "Markov chain models for inspecting response dynamics in psychological testing", "abstract": "The importance of considering contextual probabilities in shaping response patterns within psychological testing is underscored, despite the ubiquitous nature of order effects discussed extensively in methodological literature. Drawing from concepts such as path-dependency, first-order autocorrelation, state-dependency, and hysteresis, the present study is an attempt to address how earlier responses serve as an anchor for subsequent answers in tests, surveys, and questionnaires. Introducing the notion of non-commuting observables derived from quantum physics, I highlight their role in characterizing psychological processes and the impact of measurement instruments on participants' responses. We advocate for the utilization of first-order Markov chain modeling to capture and forecast sequential dependencies in survey and test responses. The employment of the first-order Markov chain model lies in individuals' propensity to exhibit partial focus to preceding responses, with recent items most likely exerting a substantial influence on subsequent response selection. This study contributes to advancing our understanding of the dynamics inherent in sequential data within psychological research and provides a methodological framework for conducting longitudinal analyses of response patterns of test and questionnaire.", "authors": ["Andrea Bosco"], "categories": ["stat.ME", "cs.LG", "math.PR"], "fields_of_study": ["Mathematics", "Computer Science"], "published_date": "2024-03-24", "url": "https://arxiv.org/abs/2403.17982", "pdf_url": "https://arxiv.org/pdf/2403.17982v1", "arxiv_id": "2403.17982", "doi": "10.48550/arXiv.2403.17982", "citation_count": 0, "influential_citation_count": 0, "has_code": false, "code_url": null, "venue": "arXiv.org", "quality_score": 0.0}
{"id": "5ebc4f5711bbc99418fb0fce0d3768dab1f95e6969bf4a26c2a903b9e9939578", "sources": ["arxiv", "semantic_scholar"], "title": "Chain-of-Spot: Interactive Reasoning Improves Large Vision-Language Models", "abstract": "In the realm of vision-language understanding, the proficiency of models in interpreting and reasoning over visual content has become a cornerstone for numerous applications. However, it is challenging for the visual encoder in Large Vision-Language Models (LVLMs) to extract useful features tailored to questions that aid the language model's response. Furthermore, a common practice among existing LVLMs is to utilize lower-resolution images, which restricts the ability for visual recognition. Our work introduces the Chain-of-Spot (CoS) method, which we describe as Interactive Reasoning, a novel approach that enhances feature extraction by focusing on key regions of interest (ROI) within the image, corresponding to the posed questions or instructions. This technique allows LVLMs to access more detailed visual information without altering the original image resolution, thereby offering multi-granularity image features. By integrating Chain-of-Spot with instruct-following LLaVA-1.5 models, the process of image reasoning consistently improves performance across a wide range of multimodal datasets and benchmarks without bells and whistles and achieves new state-of-the-art results. Our empirical findings demonstrate a significant improvement in LVLMs' ability to understand and reason about visual content, paving the way for more sophisticated visual instruction-following applications. Code and models are available at https://github.com/dongyh20/Chain-of-Spot", "authors": ["Zuyan Liu", "Yuhao Dong", "Yongming Rao", "Jie Zhou", "Jiwen Lu"], "categories": ["cs.CV"], "fields_of_study": ["Computer Science"], "published_date": "2024-03-19", "url": "https://arxiv.org/abs/2403.12966", "pdf_url": "https://arxiv.org/pdf/2403.12966v2", "arxiv_id": "2403.12966", "doi": "10.48550/arXiv.2403.12966", "citation_count": 58, "influential_citation_count": 1, "has_code": true, "code_url": "https://github.com/dongyh20/Chain-of-Spot", "venue": "arXiv.org", "quality_score": 0.4427}
{"id": "3f93690d0bd8853f429608452dd65b403cabf05c6a8cf367a68b22b6d40cb5d5", "sources": ["arxiv", "semantic_scholar"], "title": "Stop Reasoning! When Multimodal LLM with Chain-of-Thought Reasoning Meets Adversarial Image", "abstract": "Multimodal LLMs (MLLMs) with a great ability of text and image understanding have received great attention. To achieve better reasoning with MLLMs, Chain-of-Thought (CoT) reasoning has been widely explored, which further promotes MLLMs' explainability by giving intermediate reasoning steps. Despite the strong power demonstrated by MLLMs in multimodal reasoning, recent studies show that MLLMs still suffer from adversarial images. This raises the following open questions: Does CoT also enhance the adversarial robustness of MLLMs? What do the intermediate reasoning steps of CoT entail under adversarial attacks? To answer these questions, we first generalize existing attacks to CoT-based inferences by attacking the two main components, i.e., rationale and answer. We find that CoT indeed improves MLLMs' adversarial robustness against the existing attack methods by leveraging the multi-step reasoning process, but not substantially. Based on our findings, we further propose a novel attack method, termed as stop-reasoning attack, that attacks the model while bypassing the CoT reasoning process. Experiments on three MLLMs and two visual reasoning datasets verify the effectiveness of our proposed method. We show that stop-reasoning attack can result in misled predictions and outperform baseline attacks by a significant margin.", "authors": ["Zefeng Wang", "Zhen Han", "Shuo Chen", "Fan Xue", "Zifeng Ding", "Xun Xiao", "Volker Tresp", "Philip Torr", "Jindong Gu"], "categories": ["cs.CV", "cs.AI", "cs.CR", "cs.LG"], "fields_of_study": ["Computer Science"], "published_date": "2024-02-22", "url": "https://arxiv.org/abs/2402.14899", "pdf_url": "https://arxiv.org/pdf/2402.14899v3", "arxiv_id": "2402.14899", "doi": null, "citation_count": 29, "influential_citation_count": 2, "has_code": false, "code_url": null, "venue": null, "quality_score": 0.3693}
{"id": "9bfcfafe00c3455745abaa6eb4af25fa1fb3e84a81006e55326643e59b15d462", "sources": ["arxiv", "semantic_scholar"], "title": "Chain-of-Thought Reasoning Without Prompting", "abstract": "In enhancing the reasoning capabilities of large language models (LLMs), prior research primarily focuses on specific prompting techniques such as few-shot or zero-shot chain-of-thought (CoT) prompting. These methods, while effective, often involve manually intensive prompt engineering. Our study takes a novel approach by asking: Can LLMs reason effectively without prompting? Our findings reveal that, intriguingly, CoT reasoning paths can be elicited from pre-trained LLMs by simply altering the \\textit{decoding} process. Rather than conventional greedy decoding, we investigate the top-$k$ alternative tokens, uncovering that CoT paths are frequently inherent in these sequences. This approach not only bypasses the confounders of prompting but also allows us to assess the LLMs' \\textit{intrinsic} reasoning abilities. Moreover, we observe that the presence of a CoT in the decoding path correlates with a higher confidence in the model's decoded answer. This confidence metric effectively differentiates between CoT and non-CoT paths. Extensive empirical studies on various reasoning benchmarks show that the proposed CoT-decoding effectively elicits reasoning capabilities from language models, which were previously obscured by standard greedy decoding.", "authors": ["Xuezhi Wang", "Denny Zhou"], "categories": ["cs.CL"], "fields_of_study": ["Computer Science"], "published_date": "2024-02-15", "url": "https://arxiv.org/abs/2402.10200", "pdf_url": "https://arxiv.org/pdf/2402.10200v2", "arxiv_id": "2402.10200", "doi": "10.48550/arXiv.2402.10200", "citation_count": 273, "influential_citation_count": 19, "has_code": false, "code_url": null, "venue": "Neural Information Processing Systems", "quality_score": 0.6505}
{"id": "d5bf213b7f83179d2e22affe3dea839ae6b459d26f235746abbc25e2480f44ab", "sources": ["arxiv", "semantic_scholar"], "title": "Medical Image Segmentation with InTEnt: Integrated Entropy Weighting for Single Image Test-Time Adaptation", "abstract": "Test-time adaptation (TTA) refers to adapting a trained model to a new domain during testing. Existing TTA techniques rely on having multiple test images from the same domain, yet this may be impractical in real-world applications such as medical imaging, where data acquisition is expensive and imaging conditions vary frequently. Here, we approach such a task, of adapting a medical image segmentation model with only a single unlabeled test image. Most TTA approaches, which directly minimize the entropy of predictions, fail to improve performance significantly in this setting, in which we also observe the choice of batch normalization (BN) layer statistics to be a highly important yet unstable factor due to only having a single test domain example. To overcome this, we propose to instead integrate over predictions made with various estimates of target domain statistics between the training and test statistics, weighted based on their entropy statistics. Our method, validated on 24 source/target domain splits across 3 medical image datasets surpasses the leading method by 2.9% Dice coefficient on average.", "authors": ["Haoyu Dong", "Nicholas Konz", "Hanxue Gu", "Maciej A. Mazurowski"], "categories": ["cs.CV", "cs.AI"], "fields_of_study": ["Computer Science"], "published_date": "2024-02-14", "url": "https://arxiv.org/abs/2402.09604", "pdf_url": "https://arxiv.org/pdf/2402.09604v2", "arxiv_id": "2402.09604", "doi": "10.1109/CVPRW63382.2024.00511", "citation_count": 11, "influential_citation_count": 2, "has_code": true, "code_url": "https://github.com/mazurowski-lab/single-image-test-time-adaptation", "venue": null, "quality_score": 0.2698}
{"id": "f631cd13a21e25847322bcb84101f5a036049534db7d4603606ac80db9bc4d70", "sources": ["arxiv", "semantic_scholar"], "title": "Diffusion of Thoughts: Chain-of-Thought Reasoning in Diffusion Language Models", "abstract": "Recently, diffusion models have garnered significant interest in the field of text processing due to their many potential advantages compared to conventional autoregressive models. In this work, we propose Diffusion-of-Thought (DoT), a novel approach that integrates diffusion models with Chain-of-Thought, a well-established technique for improving the reasoning ability of autoregressive language models. In contrast to autoregressive language models that make decisions in a left-to-right, token-by-token manner, DoT allows reasoning steps to diffuse over time through a diffusion language model and offers greater flexibility in trading-off computation for reasoning performance. Our experimental results demonstrate the effectiveness of DoT in multi-digit multiplication, boolean logic, and grade school math problems, with a small diffusion model outperforming a much larger autoregressive model in both efficiency and accuracy. In addition to that, DoT showcases promising self-correction abilities and benefits from existing reasoning-enhancing techniques like self-consistency decoding. Our findings contribute to the understanding and development of reasoning with diffusion language models.", "authors": ["Jiacheng Ye", "Shansan Gong", "Liheng Chen", "Lin Zheng", "Jiahui Gao", "Han Shi", "Chuan Wu", "Xin Jiang", "Zhenguo Li", "Wei Bi", "Lingpeng Kong"], "categories": ["cs.CL", "cs.AI", "cs.LG"], "fields_of_study": ["Computer Science"], "published_date": "2024-02-12", "url": "https://arxiv.org/abs/2402.07754", "pdf_url": "https://arxiv.org/pdf/2402.07754v3", "arxiv_id": "2402.07754", "doi": "10.48550/arXiv.2402.07754", "citation_count": 25, "influential_citation_count": 1, "has_code": false, "code_url": null, "venue": "arXiv.org", "quality_score": 0.3537}
{"id": "667dd6cd61897dd2ae17c485fb4d481aaa1cbb5b94b3965d7e149c3f6f98289c", "sources": ["arxiv", "semantic_scholar"], "title": "A Chain-of-Thought Is as Strong as Its Weakest Link: A Benchmark for Verifiers of Reasoning Chains", "abstract": "Prompting language models to provide step-by-step answers (e.g., \"Chain-of-Thought\") is the prominent approach for complex reasoning tasks, where more accurate reasoning chains typically improve downstream task performance. Recent literature discusses automatic methods to verify reasoning to evaluate and improve their correctness. However, no fine-grained step-level datasets are available to enable thorough evaluation of such verification methods, hindering progress in this direction. We introduce REVEAL: Reasoning Verification Evaluation, a dataset to benchmark automatic verifiers of complex Chain-of-Thought reasoning in open-domain question-answering settings. REVEAL includes comprehensive labels for the relevance, attribution to evidence passages, and logical correctness of each reasoning step in a language model's answer, across a variety of datasets and state-of-the-art language models. Evaluation on REVEAL shows that verifiers struggle at verifying reasoning chains - in particular, verifying logical correctness and detecting contradictions. Available at https://reveal-dataset.github.io/ .", "authors": ["Alon Jacovi", "Yonatan Bitton", "Bernd Bohnet", "Jonathan Herzig", "Or Honovich", "Michael Tseng", "Michael Collins", "Roee Aharoni", "Mor Geva"], "categories": ["cs.CL"], "fields_of_study": ["Computer Science"], "published_date": "2024-02-01", "url": "https://arxiv.org/abs/2402.00559", "pdf_url": "https://arxiv.org/pdf/2402.00559v4", "arxiv_id": "2402.00559", "doi": "10.48550/arXiv.2402.00559", "citation_count": 73, "influential_citation_count": 8, "has_code": false, "code_url": null, "venue": "Annual Meeting of the Association for Computational Linguistics", "quality_score": 0.4771}
{"id": "b53d160f39039a181c1b6765f5bc4c12fc14812f5855ae575bad3e2997f4a200", "sources": ["arxiv", "semantic_scholar"], "title": "Deductive Beam Search: Decoding Deducible Rationale for Chain-of-Thought Reasoning", "abstract": "Recent advancements have significantly augmented the reasoning capabilities of Large Language Models (LLMs) through various methodologies, especially chain-of-thought (CoT) reasoning. However, previous methods fail to address reasoning errors in intermediate steps, leading to accumulative errors. In this paper, we propose Deductive Beam Search (DBS), which seamlessly integrates CoT and deductive reasoning with step-wise beam search for LLMs. Our approach deploys a verifier, verifying the deducibility of a reasoning step and its premises, thus alleviating the error accumulation. Furthermore, we introduce a scalable and labor-free data construction method to amplify our model's verification capabilities. Extensive experiments demonstrate that our approach significantly enhances the base performance of LLMs of various scales (7B, 13B, 70B, and ChatGPT) across 8 reasoning datasets from 3 diverse reasoning genres, including arithmetic, commonsense, and symbolic. Moreover, our analysis proves DBS's capability of detecting diverse and subtle reasoning errors and robustness on different model scales.", "authors": ["Tinghui Zhu", "Kai Zhang", "Jian Xie", "Yu Su"], "categories": ["cs.CL"], "fields_of_study": ["Computer Science"], "published_date": "2024-01-31", "url": "https://arxiv.org/abs/2401.17686", "pdf_url": "https://arxiv.org/pdf/2401.17686v3", "arxiv_id": "2401.17686", "doi": "10.48550/arXiv.2401.17686", "citation_count": 27, "influential_citation_count": 0, "has_code": false, "code_url": null, "venue": "arXiv.org", "quality_score": 0.3618}
{"id": "65bdd70f480adb6f1b22b88275e2bbbfb12c2d385204965dee5401b819b72188", "sources": ["arxiv", "semantic_scholar"], "title": "Efficient Tool Use with Chain-of-Abstraction Reasoning", "abstract": "To achieve faithful reasoning that aligns with human expectations, large language models (LLMs) need to ground their reasoning to real-world knowledge (e.g., web facts, math and physical rules). Tools help LLMs access this external knowledge, but there remains challenges for fine-tuning LLM agents (e.g., Toolformer) to invoke tools in multi-step reasoning problems, where inter-connected tool calls require holistic and efficient tool usage planning. In this work, we propose a new method for LLMs to better leverage tools in multi-step reasoning. Our method, Chain-of-Abstraction (CoA), trains LLMs to first decode reasoning chains with abstract placeholders, and then call domain tools to reify each reasoning chain by filling in specific knowledge. This planning with abstract chains enables LLMs to learn more general reasoning strategies, which are robust to shifts of domain knowledge (e.g., math results) relevant to different reasoning questions. It also allows LLMs to perform decoding and calling of external tools in parallel, which avoids the inference delay caused by waiting for tool responses. In mathematical reasoning and Wiki QA domains, we show that our method consistently outperforms previous chain-of-thought and tool-augmented baselines on both in-distribution and out-of-distribution test sets, with an average ~6% absolute QA accuracy improvement. LLM agents trained with our method also show more efficient tool use, with inference speed being on average ~1.4x faster than baseline tool-augmented LLMs.", "authors": ["Silin Gao", "Jane Dwivedi-Yu", "Ping Yu", "Xiaoqing Ellen Tan", "Ramakanth Pasunuru", "Olga Golovneva", "Koustuv Sinha", "Asli Celikyilmaz", "Antoine Bosselut", "Tianlu Wang"], "categories": ["cs.CL"], "fields_of_study": ["Computer Science"], "published_date": "2024-01-30", "url": "https://arxiv.org/abs/2401.17464", "pdf_url": "https://arxiv.org/pdf/2401.17464v3", "arxiv_id": "2401.17464", "doi": "10.48550/arXiv.2401.17464", "citation_count": 46, "influential_citation_count": 5, "has_code": false, "code_url": null, "venue": "International Conference on Computational Linguistics", "quality_score": 0.418}
{"id": "c199d4a74b0c615ec638d193fe0ea16105eb1330158d4bcfa0c696b54b9b77ac", "sources": ["arxiv", "semantic_scholar"], "title": "KAM-CoT: Knowledge Augmented Multimodal Chain-of-Thoughts Reasoning", "abstract": "Large Language Models (LLMs) have demonstrated impressive performance in natural language processing tasks by leveraging chain of thought (CoT) that enables step-by-step thinking. Extending LLMs with multimodal capabilities is the recent interest, but incurs computational cost and requires substantial hardware resources. To address these challenges, we propose KAM-CoT a framework that integrates CoT reasoning, Knowledge Graphs (KGs), and multiple modalities for a comprehensive understanding of multimodal tasks. KAM-CoT adopts a two-stage training process with KG grounding to generate effective rationales and answers. By incorporating external knowledge from KGs during reasoning, the model gains a deeper contextual understanding reducing hallucinations and enhancing the quality of answers. This knowledge-augmented CoT reasoning empowers the model to handle questions requiring external context, providing more informed answers. Experimental findings show KAM-CoT outperforms the state-of-the-art methods. On the ScienceQA dataset, we achieve an average accuracy of 93.87%, surpassing GPT-3.5 (75.17%) by 18% and GPT-4 (83.99%) by 10%. Remarkably, KAM-CoT achieves these results with only 280M trainable parameters at a time, demonstrating its cost-efficiency and effectiveness.", "authors": ["Debjyoti Mondal", "Suraj Modi", "Subhadarshi Panda", "Rituraj Singh", "Godawari Sudhakar Rao"], "categories": ["cs.CL", "cs.AI"], "fields_of_study": ["Computer Science"], "published_date": "2024-01-23", "url": "https://arxiv.org/abs/2401.12863", "pdf_url": "https://arxiv.org/pdf/2401.12863v1", "arxiv_id": "2401.12863", "doi": "10.48550/arXiv.2401.12863", "citation_count": 102, "influential_citation_count": 3, "has_code": false, "code_url": null, "venue": "AAAI Conference on Artificial Intelligence", "quality_score": 0.5032}
{"id": "fb0344e575db907ae9440fa1bfdca2f65ac309379b1318ef6bc7ddf3d6f0631f", "sources": ["arxiv", "semantic_scholar"], "title": "Code Prompting Elicits Conditional Reasoning Abilities in Text+Code LLMs", "abstract": "Reasoning is a fundamental component of language understanding. Recent prompting techniques, such as chain of thought, have consistently improved LLMs' performance on various reasoning tasks. Nevertheless, there is still little understanding of what triggers reasoning abilities in LLMs in the inference stage. In this paper, we introduce code prompting, a chain of prompts that transforms a natural language problem into code and directly prompts the LLM using the generated code without resorting to external code execution. We hypothesize that code prompts can elicit certain reasoning capabilities of LLMs trained on text and code and utilize the proposed method to improve conditional reasoning, the ability to infer different conclusions depending on the fulfillment of certain conditions. We find that code prompting exhibits a high-performance boost for multiple LLMs (up to 22.52 percentage points on GPT 3.5, 7.75 on Mixtral, and 16.78 on Mistral) across multiple conditional reasoning datasets. We then conduct comprehensive experiments to understand how code prompts trigger reasoning abilities and which capabilities are elicited in the underlying models. Our analysis of GPT 3.5 reveals that the code formatting of the input problem is essential for performance improvement. Furthermore, code prompts improve sample efficiency of in-context learning and facilitate state tracking of variables or entities.", "authors": ["Haritz Puerto", "Martin Tutek", "Somak Aditya", "Xiaodan Zhu", "Iryna Gurevych"], "categories": ["cs.CL"], "fields_of_study": ["Computer Science"], "published_date": "2024-01-18", "url": "https://arxiv.org/abs/2401.10065", "pdf_url": "https://arxiv.org/pdf/2401.10065v3", "arxiv_id": "2401.10065", "doi": "10.48550/arXiv.2401.10065", "citation_count": 23, "influential_citation_count": 4, "has_code": true, "code_url": "https://github.com/UKPLab/arxiv2024-conditional-reasoning-llms", "venue": "Conference on Empirical Methods in Natural Language Processing", "quality_score": 0.3495}
{"id": "45000629058572ddeb9f37be47ff74048a19883404b19d5e5682cb9b3cb3b134", "sources": ["arxiv", "semantic_scholar"], "title": "Chain-of-Table: Evolving Tables in the Reasoning Chain for Table Understanding", "abstract": "Table-based reasoning with large language models (LLMs) is a promising direction to tackle many table understanding tasks, such as table-based question answering and fact verification. Compared with generic reasoning, table-based reasoning requires the extraction of underlying semantics from both free-form questions and semi-structured tabular data. Chain-of-Thought and its similar approaches incorporate the reasoning chain in the form of textual context, but it is still an open question how to effectively leverage tabular data in the reasoning chain. We propose the Chain-of-Table framework, where tabular data is explicitly used in the reasoning chain as a proxy for intermediate thoughts. Specifically, we guide LLMs using in-context learning to iteratively generate operations and update the table to represent a tabular reasoning chain. LLMs can therefore dynamically plan the next operation based on the results of the previous ones. This continuous evolution of the table forms a chain, showing the reasoning process for a given tabular problem. The chain carries structured information of the intermediate results, enabling more accurate and reliable predictions. Chain-of-Table achieves new state-of-the-art performance on WikiTQ, FeTaQA, and TabFact benchmarks across multiple LLM choices.", "authors": ["Zilong Wang", "Hao Zhang", "Chun-Liang Li", "Julian Martin Eisenschlos", "Vincent Perot", "Zifeng Wang", "Lesly Miculicich", "Yasuhisa Fujii", "Jingbo Shang", "Chen-Yu Lee", "Tomas Pfister"], "categories": ["cs.CL"], "fields_of_study": ["Computer Science"], "published_date": "2024-01-09", "url": "https://arxiv.org/abs/2401.04398", "pdf_url": "https://arxiv.org/pdf/2401.04398v2", "arxiv_id": "2401.04398", "doi": "10.48550/arXiv.2401.04398", "citation_count": 260, "influential_citation_count": 43, "has_code": false, "code_url": null, "venue": "International Conference on Learning Representations", "quality_score": 0.8217}
{"id": "6c3ad83b9b1a8a3042385bf33952111a160c19d1a2a3ee2a898bc590f971ebed", "sources": ["arxiv", "semantic_scholar"], "title": "A Survey of Reasoning with Foundation Models", "abstract": "Reasoning, a crucial ability for complex problem-solving, plays a pivotal role in various real-world settings such as negotiation, medical diagnosis, and criminal investigation. It serves as a fundamental methodology in the field of Artificial General Intelligence (AGI). With the ongoing development of foundation models, e.g., Large Language Models (LLMs), there is a growing interest in exploring their abilities in reasoning tasks. In this paper, we introduce seminal foundation models proposed or adaptable for reasoning, highlighting the latest advancements in various reasoning tasks, methods, and benchmarks. We then delve into the potential future directions behind the emergence of reasoning abilities within foundation models. We also discuss the relevance of multimodal learning, autonomous agents, and super alignment in the context of reasoning. By discussing these future research directions, we hope to inspire researchers in their exploration of this field, stimulate further advancements in reasoning with foundation models, and contribute to the development of AGI.", "authors": ["Jiankai Sun", "Chuanyang Zheng", "Enze Xie", "Zhengying Liu", "Ruihang Chu", "Jianing Qiu", "Jiaqi Xu", "Mingyu Ding", "Hongyang Li", "Mengzhe Geng", "Yue Wu", "Wenhai Wang", "Junsong Chen", "Zhangyue Yin", "Xiaozhe Ren", "Jie Fu", "Junxian He", "Wu Yuan", "Qi Liu", "Xihui Liu", "Yu Li", "Hao Dong", "Yu Cheng", "Ming Zhang", "Pheng Ann Heng", "Jifeng Dai", "Ping Luo", "Jingdong Wang", "Ji-Rong Wen", "Xipeng Qiu", "Yike Guo", "Hui Xiong", "Qun Liu", "Zhenguo Li"], "categories": ["cs.AI", "cs.CL", "cs.CV", "cs.LG"], "fields_of_study": ["Computer Science"], "published_date": "2023-12-17", "url": "https://arxiv.org/abs/2312.11562", "pdf_url": "https://arxiv.org/pdf/2312.11562v5", "arxiv_id": "2312.11562", "doi": "10.48550/arXiv.2312.11562", "citation_count": 68, "influential_citation_count": 5, "has_code": true, "code_url": "https://github.com/reasoning-survey/Awesome-Reasoning-Foundation-Models", "venue": "arXiv.org", "quality_score": 0.4597}
{"id": "99e5782c52c350796d46589da5581e2d25d274d51300bcb355d144e7e57e2d48", "sources": ["arxiv", "semantic_scholar"], "title": "ThinkBot: Embodied Instruction Following with Thought Chain Reasoning", "abstract": "Embodied Instruction Following (EIF) requires agents to complete human instruction by interacting objects in complicated surrounding environments. Conventional methods directly consider the sparse human instruction to generate action plans for agents, which usually fail to achieve human goals because of the instruction incoherence in action descriptions. On the contrary, we propose ThinkBot that reasons the thought chain in human instruction to recover the missing action descriptions, so that the agent can successfully complete human goals by following the coherent instruction. Specifically, we first design an instruction completer based on large language models to recover the missing actions with interacted objects between consecutive human instruction, where the perceived surrounding environments and the completed sub-goals are considered for instruction completion. Based on the partially observed scene semantic maps, we present an object localizer to infer the position of interacted objects for agents to achieve complex human goals. Extensive experiments in the simulated environment show that our ThinkBot outperforms the state-of-the-art EIF methods by a sizable margin in both success rate and execution efficiency.", "authors": ["Guanxing Lu", "Ziwei Wang", "Changliu Liu", "Jiwen Lu", "Yansong Tang"], "categories": ["cs.CV"], "fields_of_study": ["Computer Science"], "published_date": "2023-12-12", "url": "https://arxiv.org/abs/2312.07062", "pdf_url": "https://arxiv.org/pdf/2312.07062v2", "arxiv_id": "2312.07062", "doi": "10.48550/arXiv.2312.07062", "citation_count": 24, "influential_citation_count": 2, "has_code": false, "code_url": null, "venue": "arXiv.org", "quality_score": 0.3495}
{"id": "f88c0fb3a5cd3c5fff16e16a9b789ed776ed15941bcb542df3ea4e2dc315eb63", "sources": ["arxiv", "semantic_scholar"], "title": "Chain of Code: Reasoning with a Language Model-Augmented Code Emulator", "abstract": "Code provides a general syntactic structure to build complex programs and perform precise computations when paired with a code interpreter - we hypothesize that language models (LMs) can leverage code-writing to improve Chain of Thought reasoning not only for logic and arithmetic tasks, but also for semantic ones (and in particular, those that are a mix of both). For example, consider prompting an LM to write code that counts the number of times it detects sarcasm in an essay: the LM may struggle to write an implementation for \"detect_sarcasm(string)\" that can be executed by the interpreter (handling the edge cases would be insurmountable). However, LMs may still produce a valid solution if they not only write code, but also selectively \"emulate\" the interpreter by generating the expected output of \"detect_sarcasm(string)\". In this work, we propose Chain of Code (CoC), a simple yet surprisingly effective extension that improves LM code-driven reasoning. The key idea is to encourage LMs to format semantic sub-tasks in a program as flexible pseudocode that the interpreter can explicitly catch undefined behaviors and hand off to simulate with an LM (as an \"LMulator\"). Experiments demonstrate that Chain of Code outperforms Chain of Thought and other baselines across a variety of benchmarks; on BIG-Bench Hard, Chain of Code achieves 84%, a gain of 12% over Chain of Thought. In a nutshell, CoC broadens the scope of reasoning questions that LMs can answer by \"thinking in code\".", "authors": ["Chengshu Li", "Jacky Liang", "Andy Zeng", "Xinyun Chen", "Karol Hausman", "Dorsa Sadigh", "Sergey Levine", "Li Fei-Fei", "Fei Xia", "Brian Ichter"], "categories": ["cs.CL", "cs.AI", "cs.LG", "cs.RO"], "fields_of_study": ["Computer Science"], "published_date": "2023-12-07", "url": "https://arxiv.org/abs/2312.04474", "pdf_url": "https://arxiv.org/pdf/2312.04474v4", "arxiv_id": "2312.04474", "doi": "10.48550/arXiv.2312.04474", "citation_count": 165, "influential_citation_count": 15, "has_code": true, "code_url": null, "venue": "International Conference on Machine Learning", "quality_score": 0.6021}
{"id": "33aa8593b817dfea4edc4056a927d1797d178a0c13092b1315892a038d751c8c", "sources": ["arxiv", "semantic_scholar"], "title": "LaRS: Latent Reasoning Skills for Chain-of-Thought Reasoning", "abstract": "Chain-of-thought (CoT) prompting is a popular in-context learning (ICL) approach for large language models (LLMs), especially when tackling complex reasoning tasks. Traditional ICL approaches construct prompts using examples that contain questions similar to the input question. However, CoT prompting, which includes crucial intermediate reasoning steps (rationales) within its examples, necessitates selecting examples based on these rationales rather than the questions themselves. Existing methods require human experts or pre-trained LLMs to describe the skill, a high-level abstraction of rationales, to guide the selection. These methods, however, are often costly and difficult to scale. Instead, this paper introduces a new approach named Latent Reasoning Skills (LaRS) that employs unsupervised learning to create a latent space representation of rationales, with a latent variable called a reasoning skill. Concurrently, LaRS learns a reasoning policy to determine the required reasoning skill for a given question. Then the ICL examples are selected by aligning the reasoning skills between past examples and the question. This approach is theoretically grounded and compute-efficient, eliminating the need for auxiliary LLM inference or manual prompt design. Empirical results demonstrate that LaRS consistently outperforms SOTA skill-based selection methods, processing example banks four times faster, reducing LLM inferences during the selection stage by half, and showing greater robustness to sub-optimal example banks.", "authors": ["Zifan Xu", "Haozhu Wang", "Dmitriy Bespalov", "Xian Wu", "Peter Stone", "Yanjun Qi"], "categories": ["cs.CL", "cs.AI"], "fields_of_study": ["Computer Science"], "published_date": "2023-12-07", "url": "https://arxiv.org/abs/2312.04684", "pdf_url": "https://arxiv.org/pdf/2312.04684v4", "arxiv_id": "2312.04684", "doi": "10.18653/v1/2024.findings-emnlp.206", "citation_count": 12, "influential_citation_count": 0, "has_code": false, "code_url": null, "venue": "Conference on Empirical Methods in Natural Language Processing", "quality_score": 0.2785}
{"id": "7ecb606e1ac7beefc09a345d82b167414b5cf78308b6ec6cf689aaab404abdd5", "sources": ["arxiv", "semantic_scholar"], "title": "Retrieval-augmented Multi-modal Chain-of-Thoughts Reasoning for Large Language Models", "abstract": "The advancement of Large Language Models (LLMs) has brought substantial attention to the Chain of Thought (CoT) approach, primarily due to its ability to enhance the capability of LLMs on complex reasoning tasks. Moreover, the significance of CoT approaches extends to the application of LLMs for multi-modal tasks. However, the selection of optimal CoT demonstration examples in multi-modal reasoning remains less explored for LLMs due to the inherent complexity of multi-modal examples. In this paper, we introduce a novel approach that addresses this challenge by using retrieval mechanisms to dynamically and automatically select demonstration examples based on cross-modal and intra-modal similarities. Furthermore, we employ a Stratified Sampling method of categorising demonstration examples into groups based on their types and then retrieving examples from different groups respectively to promote the diversity of demonstration examples. Through a series of experiments on two popular benchmark datasets: ScienceQA and MathVista, we demonstrate that our approach significantly improves the performance of GPT-4 by 6% on ScienceQA and 12.9% on MathVista, and enhances the performance of GPT-4V on two datasets by 2.7%, substantially improving the performance of the most advanced LLMs and LMMs for complex multi-modal reasoning tasks.", "authors": ["Bingshuai Liu", "Chenyang Lyu", "Zijun Min", "Zhanyu Wang", "Jinsong Su", "Longyue Wang"], "categories": ["cs.CL"], "fields_of_study": ["Computer Science"], "published_date": "2023-12-04", "url": "https://arxiv.org/abs/2312.01714", "pdf_url": "https://arxiv.org/pdf/2312.01714v2", "arxiv_id": "2312.01714", "doi": "10.1109/IJCNN64981.2025.11228391", "citation_count": 15, "influential_citation_count": 1, "has_code": false, "code_url": null, "venue": "IEEE International Joint Conference on Neural Network", "quality_score": 0.301}
{"id": "f0ca9344eb2e7bfe9ef7e7f4d3f9bb02f2afb1f7ef1929ce562fc86763343310", "sources": ["arxiv", "semantic_scholar"], "title": "Contrastive Chain-of-Thought Prompting", "abstract": "Despite the success of chain of thought in enhancing language model reasoning, the underlying process remains less well understood. Although logically sound reasoning appears inherently crucial for chain of thought, prior studies surprisingly reveal minimal impact when using invalid demonstrations instead. Furthermore, the conventional chain of thought does not inform language models on what mistakes to avoid, which potentially leads to more errors. Hence, inspired by how humans can learn from both positive and negative examples, we propose contrastive chain of thought to enhance language model reasoning. Compared to the conventional chain of thought, our approach provides both valid and invalid reasoning demonstrations, to guide the model to reason step-by-step while reducing reasoning mistakes. To improve generalization, we introduce an automatic method to construct contrastive demonstrations. Our experiments on reasoning benchmarks demonstrate that contrastive chain of thought can serve as a general enhancement of chain-of-thought prompting.", "authors": ["Yew Ken Chia", "Guizhen Chen", "Luu Anh Tuan", "Soujanya Poria", "Lidong Bing"], "categories": ["cs.CL"], "fields_of_study": ["Computer Science"], "published_date": "2023-11-15", "url": "https://arxiv.org/abs/2311.09277", "pdf_url": "https://arxiv.org/pdf/2311.09277v1", "arxiv_id": "2311.09277", "doi": "10.48550/arXiv.2311.09277", "citation_count": 56, "influential_citation_count": 3, "has_code": false, "code_url": null, "venue": "arXiv.org", "quality_score": 0.439}
{"id": "d749c9fbcdf1c27c4b72d16dba937e4fb5226552066039f4240bb55ed13a87bd", "sources": ["arxiv", "semantic_scholar"], "title": "The Role of Chain-of-Thought in Complex Vision-Language Reasoning Task", "abstract": "The study explores the effectiveness of the Chain-of-Thought approach, known for its proficiency in language tasks by breaking them down into sub-tasks and intermediate steps, in improving vision-language tasks that demand sophisticated perception and reasoning. We present the \"Description then Decision\" strategy, which is inspired by how humans process signals. This strategy significantly improves probing task performance by 50%, establishing the groundwork for future research on reasoning paradigms in complex vision-language tasks.", "authors": ["Yifan Wu", "Pengchuan Zhang", "Wenhan Xiong", "Barlas Oguz", "James C. Gee", "Yixin Nie"], "categories": ["cs.CL", "cs.AI", "cs.CV"], "fields_of_study": ["Computer Science"], "published_date": "2023-11-15", "url": "https://arxiv.org/abs/2311.09193", "pdf_url": "https://arxiv.org/pdf/2311.09193v1", "arxiv_id": "2311.09193", "doi": "10.48550/arXiv.2311.09193", "citation_count": 31, "influential_citation_count": 3, "has_code": false, "code_url": null, "venue": "arXiv.org", "quality_score": 0.3763}
{"id": "443fafae473a7429ea6d9c06c7764e1e46506fdf9da106798d6ace44d02e8770", "sources": ["arxiv", "semantic_scholar"], "title": "Chain of Thought with Explicit Evidence Reasoning for Few-shot Relation Extraction", "abstract": "Few-shot relation extraction involves identifying the type of relationship between two specific entities within a text, using a limited number of annotated samples. A variety of solutions to this problem have emerged by applying meta-learning and neural graph techniques which typically necessitate a training process for adaptation. Recently, the strategy of in-context learning has been demonstrating notable results without the need of training. Few studies have already utilized in-context learning for zero-shot information extraction. Unfortunately, the evidence for inference is either not considered or implicitly modeled during the construction of chain-of-thought prompts. In this paper, we propose a novel approach for few-shot relation extraction using large language models, named CoT-ER, chain-of-thought with explicit evidence reasoning. In particular, CoT-ER first induces large language models to generate evidences using task-specific and concept-level knowledge. Then these evidences are explicitly incorporated into chain-of-thought prompting for relation extraction. Experimental results demonstrate that our CoT-ER approach (with 0% training data) achieves competitive performance compared to the fully-supervised (with 100% training data) state-of-the-art approach on the FewRel1.0 and FewRel2.0 datasets.", "authors": ["Xilai Ma", "Jing Li", "Min Zhang"], "categories": ["cs.CL"], "fields_of_study": ["Computer Science"], "published_date": "2023-11-10", "url": "https://arxiv.org/abs/2311.05922", "pdf_url": "https://arxiv.org/pdf/2311.05922v3", "arxiv_id": "2311.05922", "doi": "10.48550/arXiv.2311.05922", "citation_count": 45, "influential_citation_count": 4, "has_code": false, "code_url": null, "venue": "Conference on Empirical Methods in Natural Language Processing", "quality_score": 0.4157}
{"id": "20428839697f9fbdc9b2c66e171e530b6990ef7e04b82954f4a831eaebe68613", "sources": ["arxiv", "semantic_scholar"], "title": "Implicit Chain of Thought Reasoning via Knowledge Distillation", "abstract": "To augment language models with the ability to reason, researchers usually prompt or finetune them to produce chain of thought reasoning steps before producing the final answer. However, although people use natural language to reason effectively, it may be that LMs could reason more effectively with some intermediate computation that is not in natural language. In this work, we explore an alternative reasoning approach: instead of explicitly producing the chain of thought reasoning steps, we use the language model's internal hidden states to perform implicit reasoning. The implicit reasoning steps are distilled from a teacher model trained on explicit chain-of-thought reasoning, and instead of doing reasoning \"horizontally\" by producing intermediate words one-by-one, we distill it such that the reasoning happens \"vertically\" among the hidden states in different layers. We conduct experiments on a multi-digit multiplication task and a grade school math problem dataset and find that this approach enables solving tasks previously not solvable without explicit chain-of-thought, at a speed comparable to no chain-of-thought.", "authors": ["Yuntian Deng", "Kiran Prasad", "Roland Fernandez", "Paul Smolensky", "Vishrav Chaudhary", "Stuart Shieber"], "categories": ["cs.CL", "cs.AI", "cs.LG"], "fields_of_study": ["Computer Science"], "published_date": "2023-11-02", "url": "https://arxiv.org/abs/2311.01460", "pdf_url": "https://arxiv.org/pdf/2311.01460v1", "arxiv_id": "2311.01460", "doi": "10.48550/arXiv.2311.01460", "citation_count": 122, "influential_citation_count": 15, "has_code": false, "code_url": null, "venue": "arXiv.org", "quality_score": 0.6021}
{"id": "d950c907e9393f981c2ac8dd4409829b9eb84c2be951f340953aade656b49256", "sources": ["arxiv", "semantic_scholar"], "title": "DDCoT: Duty-Distinct Chain-of-Thought Prompting for Multimodal Reasoning in Language Models", "abstract": "A long-standing goal of AI systems is to perform complex multimodal reasoning like humans. Recently, large language models (LLMs) have made remarkable strides in such multi-step reasoning on the language modality solely by leveraging the chain of thought (CoT) to mimic human thinking. However, the transfer of these advancements to multimodal contexts introduces heightened challenges, including but not limited to the impractical need for labor-intensive annotation and the limitations in terms of flexibility, generalizability, and explainability. To evoke CoT reasoning in multimodality, this work first conducts an in-depth analysis of these challenges posed by multimodality and presents two key insights: \"keeping critical thinking\" and \"letting everyone do their jobs\" in multimodal CoT reasoning. Furthermore, this study proposes a novel DDCoT prompting that maintains a critical attitude through negative-space prompting and incorporates multimodality into reasoning by first dividing the reasoning responsibility of LLMs into reasoning and recognition and then integrating the visual recognition capability of visual models into the joint reasoning process. The rationales generated by DDCoT not only improve the reasoning abilities of both large and small language models in zero-shot prompting and fine-tuning learning, significantly outperforming state-of-the-art methods but also exhibit impressive generalizability and explainability.", "authors": ["Ge Zheng", "Bin Yang", "Jiajin Tang", "Hong-Yu Zhou", "Sibei Yang"], "categories": ["cs.CV", "cs.CL"], "fields_of_study": ["Computer Science"], "published_date": "2023-10-25", "url": "https://arxiv.org/abs/2310.16436", "pdf_url": "https://arxiv.org/pdf/2310.16436v2", "arxiv_id": "2310.16436", "doi": "10.48550/arXiv.2310.16436", "citation_count": 233, "influential_citation_count": 11, "has_code": false, "code_url": null, "venue": "Neural Information Processing Systems", "quality_score": 0.5923}
{"id": "9746aeee3a4f12e3469028ea2596e5ccd2ada4b233adc3e8dfc8654d27908939", "sources": ["arxiv", "semantic_scholar"], "title": "MuSR: Testing the Limits of Chain-of-thought with Multistep Soft Reasoning", "abstract": "While large language models (LLMs) equipped with techniques like chain-of-thought prompting have demonstrated impressive capabilities, they still fall short in their ability to reason robustly in complex settings. However, evaluating LLM reasoning is challenging because system capabilities continue to grow while benchmark datasets for tasks like logical deduction have remained static. We introduce MuSR, a dataset for evaluating language models on multistep soft reasoning tasks specified in a natural language narrative. This dataset has two crucial features. First, it is created through a novel neurosymbolic synthetic-to-natural generation algorithm, enabling the construction of complex reasoning instances that challenge GPT-4 (e.g., murder mysteries roughly 1000 words in length) and which can be scaled further as more capable LLMs are released. Second, our dataset instances are free text narratives corresponding to real-world domains of reasoning; this makes it simultaneously much more challenging than other synthetically-crafted benchmarks while remaining realistic and tractable for human annotators to solve with high accuracy. We evaluate a range of LLMs and prompting techniques on this dataset and characterize the gaps that remain for techniques like chain-of-thought to perform robust reasoning.", "authors": ["Zayne Sprague", "Xi Ye", "Kaj Bostrom", "Swarat Chaudhuri", "Greg Durrett"], "categories": ["cs.CL"], "fields_of_study": ["Computer Science"], "published_date": "2023-10-24", "url": "https://arxiv.org/abs/2310.16049", "pdf_url": "https://arxiv.org/pdf/2310.16049v2", "arxiv_id": "2310.16049", "doi": "10.48550/arXiv.2310.16049", "citation_count": 193, "influential_citation_count": 24, "has_code": false, "code_url": null, "venue": "arXiv.org", "quality_score": 0.699}
{"id": "97158e7738a0bfad7304bbfb1705e5b87f7af77459614327fef44bf401209a46", "sources": ["arxiv", "semantic_scholar"], "title": "Cross-lingual Prompting: Improving Zero-shot Chain-of-Thought Reasoning across Languages", "abstract": "Chain-of-thought (CoT) is capable of eliciting models to explicitly generate reasoning paths, thus promoting reasoning accuracy and attracting increasing attention. Specifically, zero-shot CoT achieves remarkable improvements in a wide range of reasoning tasks by simply instructing the LLM with the prompt \"Let's think step by step!\". Despite the success of zero-shot CoT, the existing zero-shot prompting techniques remain limited to a single language, making it challenging to generalize to other languages and hindering global development. In this work, we introduce cross-lingual prompting (CLP), aiming to improve zero-shot CoT reasoning across languages. Specifically, CLP consists of two main components: (1) cross-lingual alignment prompting and (2) task-specific solver prompting. The cross-lingual alignment prompting is responsible for aligning representations across different languages, whereas the task-specific solver prompting is used to generate the final chain of thoughts and results for the reasoning task. In addition, we further introduce cross-lingual self-consistent prompting (CLSP) to ensemble different reasoning paths across languages. Our experimental evaluations on several benchmarks demonstrate that CLP and CLSP significantly outperform the existing prompting methods and achieve state-of-the-art performance. We hope this work will inspire further breakthroughs in cross-lingual CoT.", "authors": ["Libo Qin", "Qiguang Chen", "Fuxuan Wei", "Shijue Huang", "Wanxiang Che"], "categories": ["cs.CL", "cs.AI"], "fields_of_study": ["Computer Science"], "published_date": "2023-10-23", "url": "https://arxiv.org/abs/2310.14799", "pdf_url": "https://arxiv.org/pdf/2310.14799v1", "arxiv_id": "2310.14799", "doi": "10.48550/arXiv.2310.14799", "citation_count": 163, "influential_citation_count": 16, "has_code": false, "code_url": null, "venue": "Conference on Empirical Methods in Natural Language Processing", "quality_score": 0.6152}
{"id": "5d12df8bc11e29252b9b6a33583f403c54800b768fcd0e0704bb5b6d174c1481", "sources": ["arxiv", "semantic_scholar"], "title": "Plan, Verify and Switch: Integrated Reasoning with Diverse X-of-Thoughts", "abstract": "As large language models (LLMs) have shown effectiveness with different prompting methods, such as Chain of Thought, Program of Thought, we find that these methods have formed a great complementarity to each other on math reasoning tasks. In this work, we propose XoT, an integrated problem solving framework by prompting LLMs with diverse reasoning thoughts. For each question, XoT always begins with selecting the most suitable method then executes each method iteratively. Within each iteration, XoT actively checks the validity of the generated answer and incorporates the feedback from external executors, allowing it to dynamically switch among different prompting methods. Through extensive experiments on 10 popular math reasoning datasets, we demonstrate the effectiveness of our proposed approach and thoroughly analyze the strengths of each module. Moreover, empirical results suggest that our framework is orthogonal to recent work that makes improvements on single reasoning methods and can further generalise to logical reasoning domain. By allowing method switching, XoT provides a fresh perspective on the collaborative integration of diverse reasoning thoughts in a unified framework. The code is available at https://github.com/tengxiaoliu/XoT.", "authors": ["Tengxiao Liu", "Qipeng Guo", "Yuqing Yang", "Xiangkun Hu", "Yue Zhang", "Xipeng Qiu", "Zheng Zhang"], "categories": ["cs.CL"], "fields_of_study": ["Computer Science"], "published_date": "2023-10-23", "url": "https://arxiv.org/abs/2310.14628", "pdf_url": "https://arxiv.org/pdf/2310.14628v2", "arxiv_id": "2310.14628", "doi": "10.48550/arXiv.2310.14628", "citation_count": 47, "influential_citation_count": 4, "has_code": true, "code_url": "https://github.com/tengxiaoliu/XoT", "venue": "Conference on Empirical Methods in Natural Language Processing", "quality_score": 0.4203}
{"id": "e8e16f328353cf2ad62072ec9dfee2cb0161878ab07ac682255d6adb7051dc82", "sources": ["arxiv", "semantic_scholar"], "title": "Self-prompted Chain-of-Thought on Large Language Models for Open-domain Multi-hop Reasoning", "abstract": "In open-domain question-answering (ODQA), most existing questions require single-hop reasoning on commonsense. To further extend this task, we officially introduce open-domain multi-hop reasoning (ODMR) by answering multi-hop questions with explicit reasoning steps in open-domain setting. Recently, large language models (LLMs) have found significant utility in facilitating ODQA without external corpus. Furthermore, chain-of-thought (CoT) prompting boosts the reasoning capability of LLMs to a greater extent with manual or automated paradigms. However, existing automated methods lack of quality assurance, while manual approaches suffer from limited scalability and poor diversity, hindering the capabilities of LLMs. In this paper, we propose Self-prompted Chain-of-Thought (SP-CoT), an automated framework to mass-produce high quality CoTs of LLMs, by LLMs and for LLMs. SP-CoT introduces an automated generation pipeline of high quality ODMR datasets, an adaptive sampler for in-context CoT selection and self-prompted inference via in-context learning. Extensive experiments on four multi-hop question-answering benchmarks show that our proposed SP-CoT not only significantly surpasses the previous SOTA methods on large-scale (175B) LLMs, but also nearly doubles the zero-shot performance of small-scale (13B) LLMs. Further analysis reveals the remarkable capability of SP-CoT to elicit direct and concise intermediate reasoning steps by recalling $\\sim$50\\% of intermediate answers on MuSiQue-Ans dataset.", "authors": ["Jinyuan Wang", "Junlong Li", "Hai Zhao"], "categories": ["cs.CL", "cs.AI"], "fields_of_study": ["Computer Science"], "published_date": "2023-10-20", "url": "https://arxiv.org/abs/2310.13552", "pdf_url": "https://arxiv.org/pdf/2310.13552v2", "arxiv_id": "2310.13552", "doi": "10.48550/arXiv.2310.13552", "citation_count": 35, "influential_citation_count": 3, "has_code": false, "code_url": null, "venue": "Conference on Empirical Methods in Natural Language Processing", "quality_score": 0.3891}
{"id": "3288baa7068c7931afefb6c1f95ccf24fa5968d0805bc8a92a5465e9d32c4968", "sources": ["arxiv", "semantic_scholar"], "title": "The Expressive Power of Transformers with Chain of Thought", "abstract": "Recent theoretical work has identified surprisingly simple reasoning problems, such as checking if two nodes in a graph are connected or simulating finite-state machines, that are provably unsolvable by standard transformers that answer immediately after reading their input. However, in practice, transformers' reasoning can be improved by allowing them to use a \"chain of thought\" or \"scratchpad\", i.e., generate and condition on a sequence of intermediate tokens before answering. Motivated by this, we ask: Does such intermediate generation fundamentally extend the computational power of a decoder-only transformer? We show that the answer is yes, but the amount of increase depends crucially on the amount of intermediate generation. For instance, we find that transformer decoders with a logarithmic number of decoding steps (w.r.t. the input length) push the limits of standard transformers only slightly, while a linear number of decoding steps, assuming projected pre-norm (a slight generalization of standard pre-norm), adds a clear new ability (under standard complexity conjectures): recognizing all regular languages. Our results also imply that linear steps keep transformer decoders within context-sensitive languages, and polynomial steps with generalized pre-norm make them recognize exactly the class of polynomial-time solvable problems -- the first exact characterization of a type of transformers in terms of standard complexity classes. Together, this provides a nuanced framework for understanding how the length of a transformer's chain of thought or scratchpad impacts its reasoning power.", "authors": ["William Merrill", "Ashish Sabharwal"], "categories": ["cs.LG", "cs.CC", "cs.CL", "cs.LO"], "fields_of_study": ["Computer Science"], "published_date": "2023-10-11", "url": "https://arxiv.org/abs/2310.07923", "pdf_url": "https://arxiv.org/pdf/2310.07923v5", "arxiv_id": "2310.07923", "doi": null, "citation_count": 0, "influential_citation_count": 0, "has_code": false, "code_url": null, "venue": null, "quality_score": 0.0}
{"id": "bc969ed60278c96727fddd3ddc2cf926dfe7e460a718a62ec9fe063d6a748635", "sources": ["arxiv", "semantic_scholar"], "title": "Navigate through Enigmatic Labyrinth A Survey of Chain of Thought Reasoning: Advances, Frontiers and Future", "abstract": "Reasoning, a fundamental cognitive process integral to human intelligence, has garnered substantial interest within artificial intelligence. Notably, recent studies have revealed that chain-of-thought prompting significantly enhances LLM's reasoning capabilities, which attracts widespread attention from both academics and industry. In this paper, we systematically investigate relevant research, summarizing advanced methods through a meticulous taxonomy that offers novel perspectives. Moreover, we delve into the current frontiers and delineate the challenges and future directions, thereby shedding light on future research. Furthermore, we engage in a discussion about open questions. We hope this paper serves as an introduction for beginners and fosters future research. Resources have been made publicly available at https://github.com/zchuz/CoT-Reasoning-Survey", "authors": ["Zheng Chu", "Jingchang Chen", "Qianglong Chen", "Weijiang Yu", "Tao He", "Haotian Wang", "Weihua Peng", "Ming Liu", "Bing Qin", "Ting Liu"], "categories": ["cs.CL", "cs.AI"], "fields_of_study": ["Computer Science"], "published_date": "2023-09-27", "url": "https://arxiv.org/abs/2309.15402", "pdf_url": "https://arxiv.org/pdf/2309.15402v3", "arxiv_id": "2309.15402", "doi": "10.18653/v1/2024.acl-long.65", "citation_count": 257, "influential_citation_count": 12, "has_code": true, "code_url": "https://github.com/zchuz/CoT-Reasoning-Survey", "venue": "Annual Meeting of the Association for Computational Linguistics", "quality_score": 0.6029}
{"id": "e0aae59a52781a01aae82d383fe0242c71030f3587e4458bd83729ede129f3d8", "sources": ["arxiv", "semantic_scholar"], "title": "Enhancing Zero-Shot Chain-of-Thought Reasoning in Large Language Models through Logic", "abstract": "Recent advancements in large language models have showcased their remarkable generalizability across various domains. However, their reasoning abilities still have significant room for improvement, especially when confronted with scenarios requiring multi-step reasoning. Although large language models possess extensive knowledge, their reasoning often fails to effectively utilize this knowledge to establish a coherent thinking paradigm. These models sometimes show hallucinations as their reasoning procedures are unconstrained by logical principles. Aiming at improving the zero-shot chain-of-thought reasoning ability of large language models, we propose LoT (Logical Thoughts), a self-improvement prompting framework that leverages principles rooted in symbolic logic, particularly Reductio ad Absurdum, to systematically verify and rectify the reasoning processes step by step. Experimental evaluations conducted on language tasks in diverse domains, including arithmetic, commonsense, symbolic, causal inference, and social problems, demonstrate the efficacy of enhanced reasoning by logic. The implementation code for LoT can be accessed at: https://github.com/xf-zhao/LoT.", "authors": ["Xufeng Zhao", "Mengdi Li", "Wenhao Lu", "Cornelius Weber", "Jae Hee Lee", "Kun Chu", "Stefan Wermter"], "categories": ["cs.CL", "cs.AI", "cs.LG", "cs.SC"], "fields_of_study": ["Computer Science"], "published_date": "2023-09-23", "url": "https://arxiv.org/abs/2309.13339", "pdf_url": "https://arxiv.org/pdf/2309.13339v4", "arxiv_id": "2309.13339", "doi": "10.48550/arXiv.2309.13339", "citation_count": 77, "influential_citation_count": 2, "has_code": true, "code_url": "https://github.com/xf-zhao/LoT", "venue": "International Conference on Language Resources and Evaluation", "quality_score": 0.473}
{"id": "9d13cdec6f5cf9717c1823369a87bfe360e618c73be2d2d4d43ae59c967fe964", "sources": ["arxiv", "semantic_scholar"], "title": "Chain-of-Thought Reasoning is a Policy Improvement Operator", "abstract": "Large language models have astounded the world with fascinating new capabilities. However, they currently lack the ability to teach themselves new skills, relying instead on large amounts of human-generated training data. We introduce SECToR (Self-Education via Chain-of-Thought Reasoning), a proof-of-concept demonstration that language models can teach themselves new skills using chain-of-thought reasoning. During the self-learning loop, SECToR asks models to solve addition problems using chain-of-thought reasoning before training the next version of the model to solve those same problems directly without using such reasoning. This process often results in an improved model which can, when again augmented with chain-of-thought reasoning, solve even harder problems than the original model, allowing the self-learning loop to continue. Language models trained via SECToR autonomously learn to add up to the longest-length-digit numbers without access to any ground truth examples beyond an initial supervised fine-tuning phase consisting only of numbers with 6 or fewer digits. Our central hypothesis is that chain-of-thought reasoning can act as a policy improvement operator, similarly to how Monte-Carlo Tree Search is used in AlphaZero (Silver et al., 2017). We hope that this research can lead to new directions in which language models can learn to teach themselves without the need for human demonstrations.", "authors": ["Hugh Zhang", "David C. Parkes"], "categories": ["cs.LG", "cs.AI", "cs.CL"], "fields_of_study": ["Computer Science"], "published_date": "2023-09-15", "url": "https://arxiv.org/abs/2309.08589", "pdf_url": "https://arxiv.org/pdf/2309.08589v2", "arxiv_id": "2309.08589", "doi": "10.48550/arXiv.2309.08589", "citation_count": 18, "influential_citation_count": 2, "has_code": false, "code_url": null, "venue": "arXiv.org", "quality_score": 0.3197}
{"id": "7e5ea6cea0adb1b333b6de7af5b718d7c60e1417dfab5795ea336236e7fcd07f", "sources": ["arxiv", "semantic_scholar"], "title": "Multi-Point Detection of the Powerful Gamma Ray Burst GRB221009A Propagation through the Heliosphere on October 9, 2022", "abstract": "We present the results of processing the effects of the powerful Gamma Ray Burst GRB221009A captured by the charged particle detectors (electrostatic analyzers and solid-state detectors) onboard spacecraft at different points in the heliosphere on October 9, 2022. To follow the GRB221009A propagation through the heliosphere we used the electron and proton flux measurements from solar missions Solar Orbiter and STEREO-A; Earth magnetosphere and the solar wind missions THEMIS and Wind; meteorological satellites POES15, POES19, MetOp3; and MAVEN - a NASA mission orbiting Mars. GRB221009A had a structure of four bursts: less intense Pulse 1 - the triggering impulse - was detected by gamma-ray observatories at 131659 UT (near the Earth); the most intense Pulses 2 and 3 were detected on board all the spacecraft from the list, and Pulse 4 detected in more than 500 s after Pulse 1. Due to their different scientific objectives, the spacecraft, which data was used in this study, were separated by more than 1 AU (Solar Orbiter and MAVEN). This enabled tracking GRB221009A as it was propagating across the heliosphere. STEREO-A was the first to register Pulse 2 and 3 of the GRB, almost 100 seconds before their detection by spacecraft in the vicinity of Earth. MAVEN detected GRB221009A Pulses 2, 3, and 4 at the orbit of Mars about 237 seconds after their detection near Earth. By processing the time delays observed we show that the source location of the GRB221009A was at RA 288.5 degrees, Dec 18.5 degrees (J2000) with an error cone of 2 degrees", "authors": ["Andrii Voshchepynets", "Oleksiy Agapitov", "Lynn Wilson", "Vassilis Angelopoulos", "Samer T. Alnussirat", "Michael Balikhin", "Myroslava Hlebena", "Ihor Korol", "Davin Larson", "David Mitchell", "Christopher Owen", "Ali Rahmati", "Department of System Analysis", "Optimization Theory", "Uzhhorod National University", " Uzhhorod", " Ukraine", "Space Sciences Laboratory", " :", "University of California Berkeley Berkeley", "CA 94720", " Astronomy", "Space Physics Department", "National Taras Shevchenko University of Kyiv", " Kyiv", " Ukraine", "Goddard Space Flight Center", "National Aeronautics", "Space Administration", " Greenbelt", " MD", "Department of Earth", " Planetary", "Space Sciences", "University of California Los Angeles", "Los Angeles", " CA}", "University of Sheffield", " Sheffield", " UK", "Department of Algebra", "Differential Equation", "Uzhhorod National University", " Uzhhorod", " Ukraine", "Department of Mathematical Analysis", "The John Paul II Catholic University of Lublin", " Lublin", " Poland}"], "categories": ["astro-ph.HE", "astro-ph.IM", "astro-ph.SR"], "fields_of_study": ["Physics"], "published_date": "2023-09-11", "url": "https://arxiv.org/abs/2309.05856", "pdf_url": "https://arxiv.org/pdf/2309.05856v1", "arxiv_id": "2309.05856", "doi": null, "citation_count": 0, "influential_citation_count": 0, "has_code": false, "code_url": null, "venue": null, "quality_score": 0.0}
{"id": "b13149c47279cb44640f36f162be2c298816e2463bb01df6a9cee302cd5385ff", "sources": ["arxiv", "semantic_scholar"], "title": "Measuring and Improving Chain-of-Thought Reasoning in Vision-Language Models", "abstract": "Vision-language models (VLMs) have recently demonstrated strong efficacy as visual assistants that can parse natural queries about the visual content and generate human-like outputs. In this work, we explore the ability of these models to demonstrate human-like reasoning based on the perceived information. To address a crucial concern regarding the extent to which their reasoning capabilities are fully consistent and grounded, we also measure the reasoning consistency of these models. We achieve this by proposing a chain-of-thought (CoT) based consistency measure. However, such an evaluation requires a benchmark that encompasses both high-level inference and detailed reasoning chains, which is costly. We tackle this challenge by proposing a LLM-Human-in-the-Loop pipeline, which notably reduces cost while simultaneously ensuring the generation of a high-quality dataset. Based on this pipeline and the existing coarse-grained annotated dataset, we build the CURE benchmark to measure both the zero-shot reasoning performance and consistency of VLMs. We evaluate existing state-of-the-art VLMs, and find that even the best-performing model is unable to demonstrate strong visual reasoning capabilities and consistency, indicating that substantial efforts are required to enable VLMs to perform visual reasoning as systematically and consistently as humans. As an early step, we propose a two-stage training framework aimed at improving both the reasoning performance and consistency of VLMs. The first stage involves employing supervised fine-tuning of VLMs using step-by-step reasoning samples automatically generated by LLMs. In the second stage, we further augment the training process by incorporating feedback provided by LLMs to produce reasoning chains that are highly consistent and grounded. We empirically highlight the effectiveness of our framework in both reasoning performance and consistency.", "authors": ["Yangyi Chen", "Karan Sikka", "Michael Cogswell", "Heng Ji", "Ajay Divakaran"], "categories": ["cs.CL", "cs.CV", "cs.LG"], "fields_of_study": ["Computer Science"], "published_date": "2023-09-08", "url": "https://arxiv.org/abs/2309.04461", "pdf_url": "https://arxiv.org/pdf/2309.04461v2", "arxiv_id": "2309.04461", "doi": "10.48550/arXiv.2309.04461", "citation_count": 57, "influential_citation_count": 2, "has_code": true, "code_url": "https://github.com/Yangyi-Chen/CoTConsistency", "venue": "North American Chapter of the Association for Computational Linguistics", "quality_score": 0.4409}
{"id": "332bc1275e0dd761d0dbfe4c2b28ed33a0a01fd0cc36e54b1c87c873f6771d45", "sources": ["arxiv", "semantic_scholar"], "title": "Making Large Language Models Better Reasoners with Alignment", "abstract": "Reasoning is a cognitive process of using evidence to reach a sound conclusion. The reasoning capability is essential for large language models (LLMs) to serve as the brain of the artificial general intelligence agent. Recent studies reveal that fine-tuning LLMs on data with the chain of thought (COT) reasoning process can significantly enhance their reasoning capabilities. However, we find that the fine-tuned LLMs suffer from an \\textit{Assessment Misalignment} problem, i.e., they frequently assign higher scores to subpar COTs, leading to potential limitations in their reasoning abilities. To address this problem, we introduce an \\textit{Alignment Fine-Tuning (AFT)} paradigm, which involves three steps: 1) fine-tuning LLMs with COT training data; 2) generating multiple COT responses for each question, and categorizing them into positive and negative ones based on whether they achieve the correct answer; 3) calibrating the scores of positive and negative responses given by LLMs with a novel constraint alignment loss. Specifically, the constraint alignment loss has two objectives: a) Alignment, which guarantees that positive scores surpass negative scores to encourage answers with high-quality COTs; b) Constraint, which keeps the negative scores confined to a reasonable range to prevent the model degradation. Beyond just the binary positive and negative feedback, the constraint alignment loss can be seamlessly adapted to the ranking situations when ranking feedback is accessible. Furthermore, we also delve deeply into recent ranking-based alignment methods, such as DPO, RRHF, and PRO, and discover that the constraint, which has been overlooked by these approaches, is also crucial for their performance. Extensive experiments on four reasoning benchmarks with both binary and ranking feedback demonstrate the effectiveness of AFT.", "authors": ["Peiyi Wang", "Lei Li", "Liang Chen", "Feifan Song", "Binghuai Lin", "Yunbo Cao", "Tianyu Liu", "Zhifang Sui"], "categories": ["cs.CL", "cs.AI", "cs.LG"], "fields_of_study": ["Computer Science"], "published_date": "2023-09-05", "url": "https://arxiv.org/abs/2309.02144", "pdf_url": "https://arxiv.org/pdf/2309.02144v1", "arxiv_id": "2309.02144", "doi": "10.48550/arXiv.2309.02144", "citation_count": 90, "influential_citation_count": 4, "has_code": false, "code_url": null, "venue": "arXiv.org", "quality_score": 0.4898}
{"id": "16887b45dc5e7ac17245718bc935d75080a1e01496cc301623bd7dbeea5c58a9", "sources": ["arxiv", "semantic_scholar"], "title": "Cumulative Reasoning with Large Language Models", "abstract": "Recent advancements in large language models (LLMs) have shown remarkable progress, yet their ability to solve complex problems remains limited. In this work, we introduce Cumulative Reasoning (CR), a structured framework that enhances LLM problem-solving by emulating human-like iterative and cumulative thought processes. CR orchestrates LLMs in three distinct roles: Proposer, Verifier(s), and Reporter, to systematically decompose tasks, generate and validate intermediate reasoning steps, and compose them into a solution by building a dynamic Directed Acyclic Graph (DAG) of verified propositions. This approach substantially enhances problem-solving capabilities. We demonstrate CR's advantage through several complex reasoning tasks: it outperforms existing methods in logical inference tasks with up to a 9.3% improvement, achieving 98.04% accuracy on the curated FOLIO wiki dataset. In the Game of 24, it achieves 98% accuracy, marking a 24% improvement over previous methods. In solving MATH problems, CR achieves a 4.2% increase from previous methods and a 43% relative improvement in the most challenging level 5 problems. When incorporating a code environment with CR, we further harness LLMs' reasoning capabilities and outperform the Program of Thought (PoT) method by 38.8%.", "authors": ["Yifan Zhang", "Jingqin Yang", "Yang Yuan", "Andrew Chi-Chih Yao"], "categories": ["cs.AI"], "fields_of_study": ["Computer Science"], "published_date": "2023-08-08", "url": "https://arxiv.org/abs/2308.04371", "pdf_url": "https://arxiv.org/pdf/2308.04371v11", "arxiv_id": "2308.04371", "doi": "10.48550/arXiv.2308.04371", "citation_count": 98, "influential_citation_count": 7, "has_code": true, "code_url": "https://github.com/iiis-ai/cumulative-reasoning", "venue": null, "quality_score": 0.4989}
{"id": "dbf35b5b3a6dccf5e7047298f6fab08b25616efc1c840b20f9f3350a191b55eb", "sources": ["arxiv", "semantic_scholar"], "title": "Measuring Faithfulness in Chain-of-Thought Reasoning", "abstract": "Large language models (LLMs) perform better when they produce step-by-step, \"Chain-of-Thought\" (CoT) reasoning before answering a question, but it is unclear if the stated reasoning is a faithful explanation of the model's actual reasoning (i.e., its process for answering the question). We investigate hypotheses for how CoT reasoning may be unfaithful, by examining how the model predictions change when we intervene on the CoT (e.g., by adding mistakes or paraphrasing it). Models show large variation across tasks in how strongly they condition on the CoT when predicting their answer, sometimes relying heavily on the CoT and other times primarily ignoring it. CoT's performance boost does not seem to come from CoT's added test-time compute alone or from information encoded via the particular phrasing of the CoT. As models become larger and more capable, they produce less faithful reasoning on most tasks we study. Overall, our results suggest that CoT can be faithful if the circumstances such as the model size and task are carefully chosen.", "authors": ["Tamera Lanham", "Anna Chen", "Ansh Radhakrishnan", "Benoit Steiner", "Carson Denison", "Danny Hernandez", "Dustin Li", "Esin Durmus", "Evan Hubinger", "Jackson Kernion", "Kamilė Lukošiūtė", "Karina Nguyen", "Newton Cheng", "Nicholas Joseph", "Nicholas Schiefer", "Oliver Rausch", "Robin Larson", "Sam McCandlish", "Sandipan Kundu", "Saurav Kadavath", "Shannon Yang", "Thomas Henighan", "Timothy Maxwell", "Timothy Telleen-Lawton", "Tristan Hume", "Zac Hatfield-Dodds", "Jared Kaplan", "Jan Brauner", "Samuel R. Bowman", "Ethan Perez"], "categories": ["cs.AI", "cs.CL", "cs.LG"], "fields_of_study": ["Computer Science"], "published_date": "2023-07-17", "url": "https://arxiv.org/abs/2307.13702", "pdf_url": "https://arxiv.org/pdf/2307.13702v1", "arxiv_id": "2307.13702", "doi": "10.48550/arXiv.2307.13702", "citation_count": 454, "influential_citation_count": 43, "has_code": false, "code_url": null, "venue": "arXiv.org", "quality_score": 0.8217}
{"id": "ba3ebfe4ab7ed0649b87af019b5ec78d58a9ac8646633ac14ea7dfe59b50bfa9", "sources": ["arxiv", "semantic_scholar"], "title": "Test-Time Training on Video Streams", "abstract": "Prior work has established Test-Time Training (TTT) as a general framework to further improve a trained model at test time. Before making a prediction on each test instance, the model is first trained on the same instance using a self-supervised task such as reconstruction. We extend TTT to the streaming setting, where multiple test instances - video frames in our case - arrive in temporal order. Our extension is online TTT: The current model is initialized from the previous model, then trained on the current frame and a small window of frames immediately before. Online TTT significantly outperforms the fixed-model baseline for four tasks, on three real-world datasets. The improvements are more than 2.2x and 1.5x for instance and panoptic segmentation. Surprisingly, online TTT also outperforms its offline variant that accesses strictly more information, training on all frames from the entire test video regardless of temporal order. This finding challenges those in prior work using synthetic videos. We formalize a notion of locality as the advantage of online over offline TTT, and analyze its role with ablations and a theory based on bias-variance trade-off.", "authors": ["Renhao Wang", "Yu Sun", "Arnuv Tandon", "Yossi Gandelsman", "Xinlei Chen", "Alexei A. Efros", "Xiaolong Wang"], "categories": ["cs.CV", "cs.LG"], "fields_of_study": ["Computer Science"], "published_date": "2023-07-11", "url": "https://arxiv.org/abs/2307.05014", "pdf_url": "https://arxiv.org/pdf/2307.05014v3", "arxiv_id": "2307.05014", "doi": "10.48550/arXiv.2307.05014", "citation_count": 39, "influential_citation_count": 1, "has_code": false, "code_url": null, "venue": "Journal of machine learning research", "quality_score": 0.4005}
{"id": "52ab17d3e82aae42dbac08b9c3b29cbdc50614178ac0e009c51f616b512fe39c", "sources": ["arxiv", "semantic_scholar"], "title": "Let's Do a Thought Experiment: Using Counterfactuals to Improve Moral Reasoning", "abstract": "Language models still struggle on moral reasoning, despite their impressive performance in many other tasks. In particular, the Moral Scenarios task in MMLU (Multi-task Language Understanding) is among the worst performing tasks for many language models, including GPT-3. In this work, we propose a new prompting framework, Thought Experiments, to teach language models to do better moral reasoning using counterfactuals. Experiment results show that our framework elicits counterfactual questions and answers from the model, which in turn helps improve the accuracy on Moral Scenarios task by 9-16% compared to other zero-shot baselines. Interestingly, unlike math reasoning tasks, zero-shot Chain-of-Thought (CoT) reasoning doesn't work out of the box, and even reduces accuracy by around 4% compared to direct zero-shot. We further observed that with minimal human supervision in the form of 5 few-shot examples, the accuracy of the task can be improved to as much as 80%.", "authors": ["Xiao Ma", "Swaroop Mishra", "Ahmad Beirami", "Alex Beutel", "Jilin Chen"], "categories": ["cs.CL", "cs.AI"], "fields_of_study": ["Computer Science"], "published_date": "2023-06-25", "url": "https://arxiv.org/abs/2306.14308", "pdf_url": "https://arxiv.org/pdf/2306.14308v1", "arxiv_id": "2306.14308", "doi": "10.48550/arXiv.2306.14308", "citation_count": 20, "influential_citation_count": 1, "has_code": false, "code_url": null, "venue": "arXiv.org", "quality_score": 0.3306}
{"id": "97a0603e8752c90099c61cb56317e0d4cca35a75c77ebb4ddcfca60e1c10d90c", "sources": ["arxiv", "semantic_scholar"], "title": "Deductive Verification of Chain-of-Thought Reasoning", "abstract": "Large Language Models (LLMs) significantly benefit from Chain-of-Thought (CoT) prompting in performing various reasoning tasks. While CoT allows models to produce more comprehensive reasoning processes, its emphasis on intermediate reasoning steps can inadvertently introduce hallucinations and accumulated errors, thereby limiting models' ability to solve complex reasoning tasks. Inspired by how humans engage in careful and meticulous deductive logical reasoning processes to solve tasks, we seek to enable language models to perform explicit and rigorous deductive reasoning, and also ensure the trustworthiness of their reasoning process through self-verification. However, directly verifying the validity of an entire deductive reasoning process is challenging, even with advanced models like ChatGPT. In light of this, we propose to decompose a reasoning verification process into a series of step-by-step subprocesses, each only receiving their necessary context and premises. To facilitate this procedure, we propose Natural Program, a natural language-based deductive reasoning format. Our approach enables models to generate precise reasoning steps where subsequent steps are more rigorously grounded on prior steps. It also empowers language models to carry out reasoning self-verification in a step-by-step manner. By integrating this verification process into each deductive reasoning stage, we significantly enhance the rigor and trustfulness of generated reasoning steps. Along this process, we also improve the answer correctness on complex reasoning tasks. Code will be released at https://github.com/lz1oceani/verify_cot.", "authors": ["Zhan Ling", "Yunhao Fang", "Xuanlin Li", "Zhiao Huang", "Mingu Lee", "Roland Memisevic", "Hao Su"], "categories": ["cs.CL", "cs.AI", "cs.LG"], "fields_of_study": ["Computer Science"], "published_date": "2023-06-06", "url": "https://arxiv.org/abs/2306.03872", "pdf_url": "https://arxiv.org/pdf/2306.03872v3", "arxiv_id": "2306.03872", "doi": "10.52202/075280-1580", "citation_count": 214, "influential_citation_count": 13, "has_code": true, "code_url": "https://github.com/lz1oceani/verify_cot", "venue": "Neural Information Processing Systems", "quality_score": 0.5831}
{"id": "0bbcbe2ddddac08166f03a8bad76ad285e8b6941cbd722cf4f1f1e1581f4e189", "sources": ["arxiv", "semantic_scholar"], "title": "Chain-of-Thought Hub: A Continuous Effort to Measure Large Language Models' Reasoning Performance", "abstract": "As large language models (LLMs) are continuously being developed, their evaluation becomes increasingly important yet challenging. This work proposes Chain-of-Thought Hub, an open-source evaluation suite on the multi-step reasoning capabilities of large language models. We are interested in this setting for two reasons: (1) from the behavior of GPT and PaLM model family, we observe that complex reasoning is likely to be a key differentiator between weaker and stronger LLMs; (2) we envisage large language models to become the next-generation computational platform and foster an ecosystem of LLM-based new applications, this naturally requires the foundation models to perform complex tasks that often involve the composition of linguistic and logical operations. Our approach is to compile a suite of challenging reasoning benchmarks to track the progress of LLMs. Our current results show that: (1) model scale clearly correlates with reasoning capabilities; (2) As of May 2023, Claude-v1.3 and PaLM-2 are the only two models that are comparable with GPT-4, while open-sourced models still lag behind; (3) LLaMA-65B performs closely to code-davinci-002, indicating that with successful further development such as reinforcement learning from human feedback (RLHF), it has great potential to be close to GPT-3.5-Turbo. Our results also suggest that for the open-source efforts to catch up, the community may focus more on building better base models and exploring RLHF.", "authors": ["Yao Fu", "Litu Ou", "Mingyu Chen", "Yuhao Wan", "Hao Peng", "Tushar Khot"], "categories": ["cs.CL", "cs.AI", "cs.LG"], "fields_of_study": ["Computer Science"], "published_date": "2023-05-26", "url": "https://arxiv.org/abs/2305.17306", "pdf_url": "https://arxiv.org/pdf/2305.17306v1", "arxiv_id": "2305.17306", "doi": "10.48550/arXiv.2305.17306", "citation_count": 131, "influential_citation_count": 8, "has_code": true, "code_url": "https://github.com/FranxYao/chain-of-thought-hub", "venue": "arXiv.org", "quality_score": 0.5301}
{"id": "c5efac5a2d8ca7d852b2588926c68393308a4df6099c3b37eceed6a4dec10e10", "sources": ["arxiv", "semantic_scholar"], "title": "Beyond Chain-of-Thought, Effective Graph-of-Thought Reasoning in Language Models", "abstract": "With the widespread use of language models (LMs) in NLP tasks, researchers have discovered the potential of Chain-of-thought (CoT) to assist LMs in accomplishing complex reasoning tasks by generating intermediate steps. However, human thought processes are often non-linear, rather than simply sequential chains of thoughts. Therefore, we propose Graph-of-Thought (GoT) reasoning, which models human thought processes not only as a chain but also as a graph. By representing thought units as nodes and connections between them as edges, our approach captures the non-sequential nature of human thinking and allows for a more realistic modeling of thought processes. GoT adopts a two-stage framework with an additional GoT encoder for thought graph representation and fuses the graph representation with the original input representation through a gated fusion mechanism. We evaluate GoT's performance on a text-only reasoning task (AQUA-RAT) and a multimodal reasoning task (ScienceQA). Our model achieves significant improvement over the strong CoT baseline on the AQUA-RAT test set and boosts accuracy from 85.19% to 87.59% using the T5-base model over the state-of-the-art Multimodal-CoT on the ScienceQA test set.", "authors": ["Yao Yao", "Zuchao Li", "Hai Zhao"], "categories": ["cs.CL"], "fields_of_study": ["Computer Science"], "published_date": "2023-05-26", "url": "https://arxiv.org/abs/2305.16582", "pdf_url": "https://arxiv.org/pdf/2305.16582v2", "arxiv_id": "2305.16582", "doi": null, "citation_count": 52, "influential_citation_count": 2, "has_code": false, "code_url": null, "venue": null, "quality_score": 0.4311}
{"id": "61fa6ca7e66d9792f4becd20df4376069bcb092d0bf75c7adc252830a881e383", "sources": ["arxiv", "semantic_scholar"], "title": "GRACE: Discriminator-Guided Chain-of-Thought Reasoning", "abstract": "In the context of multi-step reasoning, e.g., with chain-of-thought, language models (LMs) can easily assign a high likelihood to incorrect steps. As a result, decoding strategies that optimize for solution likelihood often yield incorrect solutions. To address this issue, we propose Guiding chain-of-thought ReAsoning with a CorrectnEss Discriminator (GRACE), a stepwise decoding approach that steers the decoding process towards producing correct reasoning steps. GRACE employs a step-level verifier or discriminator trained with a contrastive loss over correct and incorrect steps, which is used during decoding to score next-step candidates based on their correctness. Importantly, GRACE only requires sampling from the LM, without the need for LM training or fine-tuning. Using models from FLAN-T5 and LLaMA families, we evaluate GRACE over four math and two symbolic reasoning tasks, where it exhibits substantial performance gains compared to greedy decoding, verifiers, and self-consistency in most settings. When further combined with self-consistency, GRACE outperforms all the baselines by sizeable margins. Human and LLM evaluations over GSM8K show that GRACE not only improves the final answer accuracy but also the correctness of the intermediate reasoning. Our implementation can be accessed at https://github.com/mukhal/grace.", "authors": ["Muhammad Khalifa", "Lajanugen Logeswaran", "Moontae Lee", "Honglak Lee", "Lu Wang"], "categories": ["cs.CL", "cs.AI"], "fields_of_study": ["Computer Science"], "published_date": "2023-05-24", "url": "https://arxiv.org/abs/2305.14934", "pdf_url": "https://arxiv.org/pdf/2305.14934v3", "arxiv_id": "2305.14934", "doi": "10.18653/v1/2023.findings-emnlp.1022", "citation_count": 56, "influential_citation_count": 3, "has_code": true, "code_url": "https://github.com/mukhal/grace", "venue": "Conference on Empirical Methods in Natural Language Processing", "quality_score": 0.439}
{"id": "34eab023da90ee9df2b0382c3bc3cadd28b0bf7842967a43dcb0f792a0475355", "sources": ["arxiv", "semantic_scholar"], "title": "ChatCoT: Tool-Augmented Chain-of-Thought Reasoning on Chat-based Large Language Models", "abstract": "Although large language models (LLMs) have achieved excellent performance in a variety of evaluation benchmarks, they still struggle in complex reasoning tasks which require specific knowledge and multi-hop reasoning. To improve the reasoning abilities, we propose ChatCoT, a tool-augmented chain-of-thought reasoning framework for chat-based LLMs (e.g., ChatGPT). In ChatCoT, we model the chain-of-thought (CoT) reasoning as multi-turn conversations, to utilize tools in a more natural way through chatting. At each turn, LLMs can either interact with tools or perform the reasoning. Our approach can effectively leverage the multi-turn conversation ability of chat-based LLMs, and integrate the thought chain following and tools manipulation in a unified way. Specially, we initialize the early turns of the conversation by the knowledge about tools, tasks, and reasoning format, and propose an iterative tool-augmented reasoning step to perform step-by-step tool-augmented reasoning. The experiment results on two complex reasoning datasets (MATH and HotpotQA) have shown the effectiveness of ChatCoT on complex reasoning tasks, achieving a 7.9% relative improvement over the state-of-the-art baseline. Our code and data are available at: \\url{https://github.com/RUCAIBOX/ChatCoT}.", "authors": ["Zhipeng Chen", "Kun Zhou", "Beichen Zhang", "Zheng Gong", "Wayne Xin Zhao", "Ji-Rong Wen"], "categories": ["cs.CL"], "fields_of_study": ["Computer Science"], "published_date": "2023-05-23", "url": "https://arxiv.org/abs/2305.14323", "pdf_url": "https://arxiv.org/pdf/2305.14323v3", "arxiv_id": "2305.14323", "doi": "10.48550/arXiv.2305.14323", "citation_count": 48, "influential_citation_count": 2, "has_code": true, "code_url": "https://github.com/RUCAIBOX/ChatCoT}", "venue": "Conference on Empirical Methods in Natural Language Processing", "quality_score": 0.4225}
{"id": "01de2cc74a6e1b7d7dfed8ac6b044e30a91bfe66bc97bc4691bde3b7e8c5f618", "sources": ["arxiv", "semantic_scholar"], "title": "\"Is the Pope Catholic?\" Applying Chain-of-Thought Reasoning to Understanding Conversational Implicatures", "abstract": "Conversational implicatures are pragmatic inferences that require listeners to deduce the intended meaning conveyed by a speaker from their explicit utterances. Although such inferential reasoning is fundamental to human communication, recent research indicates that large language models struggle to comprehend these implicatures as effectively as the average human. This paper demonstrates that by incorporating Grice's Four Maxims into the model through chain-of-thought prompting, we can significantly enhance its performance, surpassing even the average human performance on this task.", "authors": ["Zae Myung Kim", "David E. Taylor", "Dongyeop Kang"], "categories": ["cs.CL"], "fields_of_study": ["Computer Science"], "published_date": "2023-05-23", "url": "https://arxiv.org/abs/2305.13826", "pdf_url": "https://arxiv.org/pdf/2305.13826v1", "arxiv_id": "2305.13826", "doi": "10.48550/arXiv.2305.13826", "citation_count": 7, "influential_citation_count": 1, "has_code": false, "code_url": null, "venue": "arXiv.org", "quality_score": 0.2258}
{"id": "688217e60a68a24afcc2eafe4e94677e308afe25a7961765ba082d83020e2f1c", "sources": ["arxiv", "semantic_scholar"], "title": "Reasoning Implicit Sentiment with Chain-of-Thought Prompting", "abstract": "While sentiment analysis systems try to determine the sentiment polarities of given targets based on the key opinion expressions in input texts, in implicit sentiment analysis (ISA) the opinion cues come in an implicit and obscure manner. Thus detecting implicit sentiment requires the common-sense and multi-hop reasoning ability to infer the latent intent of opinion. Inspired by the recent chain-of-thought (CoT) idea, in this work we introduce a Three-hop Reasoning (THOR) CoT framework to mimic the human-like reasoning process for ISA. We design a three-step prompting principle for THOR to step-by-step induce the implicit aspect, opinion, and finally the sentiment polarity. Our THOR+Flan-T5 (11B) pushes the state-of-the-art (SoTA) by over 6% F1 on supervised setup. More strikingly, THOR+GPT3 (175B) boosts the SoTA by over 50% F1 on zero-shot setting. Our code is open at https://github.com/scofield7419/THOR-ISA.", "authors": ["Hao Fei", "Bobo Li", "Qian Liu", "Lidong Bing", "Fei Li", "Tat-Seng Chua"], "categories": ["cs.CL"], "fields_of_study": ["Computer Science"], "published_date": "2023-05-18", "url": "https://arxiv.org/abs/2305.11255", "pdf_url": "https://arxiv.org/pdf/2305.11255v4", "arxiv_id": "2305.11255", "doi": "10.48550/arXiv.2305.11255", "citation_count": 170, "influential_citation_count": 19, "has_code": true, "code_url": "https://github.com/scofield7419/THOR-ISA", "venue": "Annual Meeting of the Association for Computational Linguistics", "quality_score": 0.6505}
{"id": "4224462bfa53a453fe75e3923662faa7f30b148421bc3d186be87ae81f02138f", "sources": ["arxiv", "semantic_scholar"], "title": "Tree of Thoughts: Deliberate Problem Solving with Large Language Models", "abstract": "Language models are increasingly being deployed for general problem solving across a wide range of tasks, but are still confined to token-level, left-to-right decision-making processes during inference. This means they can fall short in tasks that require exploration, strategic lookahead, or where initial decisions play a pivotal role. To surmount these challenges, we introduce a new framework for language model inference, Tree of Thoughts (ToT), which generalizes over the popular Chain of Thought approach to prompting language models, and enables exploration over coherent units of text (thoughts) that serve as intermediate steps toward problem solving. ToT allows LMs to perform deliberate decision making by considering multiple different reasoning paths and self-evaluating choices to decide the next course of action, as well as looking ahead or backtracking when necessary to make global choices. Our experiments show that ToT significantly enhances language models' problem-solving abilities on three novel tasks requiring non-trivial planning or search: Game of 24, Creative Writing, and Mini Crosswords. For instance, in Game of 24, while GPT-4 with chain-of-thought prompting only solved 4% of tasks, our method achieved a success rate of 74%. Code repo with all prompts: https://github.com/princeton-nlp/tree-of-thought-llm.", "authors": ["Shunyu Yao", "Dian Yu", "Jeffrey Zhao", "Izhak Shafran", "Thomas L. Griffiths", "Yuan Cao", "Karthik Narasimhan"], "categories": ["cs.CL", "cs.AI", "cs.LG"], "fields_of_study": ["Computer Science"], "published_date": "2023-05-17", "url": "https://arxiv.org/abs/2305.10601", "pdf_url": "https://arxiv.org/pdf/2305.10601v2", "arxiv_id": "2305.10601", "doi": "10.48550/arXiv.2305.10601", "citation_count": 4251, "influential_citation_count": 286, "has_code": true, "code_url": "https://github.com/princeton-nlp/tree-of-thought-llm", "venue": "Neural Information Processing Systems", "quality_score": 1.0}
{"id": "6c32da792777b7df0fe00edefb99e26264d767bdcde0423fc123544cf892fa18", "sources": ["arxiv", "semantic_scholar"], "title": "Answering Questions by Meta-Reasoning over Multiple Chains of Thought", "abstract": "Modern systems for multi-hop question answering (QA) typically break questions into a sequence of reasoning steps, termed chain-of-thought (CoT), before arriving at a final answer. Often, multiple chains are sampled and aggregated through a voting mechanism over the final answers, but the intermediate steps themselves are discarded. While such approaches improve performance, they do not consider the relations between intermediate steps across chains and do not provide a unified explanation for the predicted answer. We introduce Multi-Chain Reasoning (MCR), an approach which prompts large language models to meta-reason over multiple chains of thought, rather than aggregating their answers. MCR examines different reasoning chains, mixes information between them and selects the most relevant facts in generating an explanation and predicting the answer. MCR outperforms strong baselines on 7 multi-hop QA datasets. Moreover, our analysis reveals that MCR explanations exhibit high quality, enabling humans to verify its answers.", "authors": ["Ori Yoran", "Tomer Wolfson", "Ben Bogin", "Uri Katz", "Daniel Deutch", "Jonathan Berant"], "categories": ["cs.CL", "cs.AI"], "fields_of_study": ["Computer Science"], "published_date": "2023-04-25", "url": "https://arxiv.org/abs/2304.13007", "pdf_url": "https://arxiv.org/pdf/2304.13007v4", "arxiv_id": "2304.13007", "doi": "10.48550/arXiv.2304.13007", "citation_count": 133, "influential_citation_count": 6, "has_code": false, "code_url": null, "venue": "Conference on Empirical Methods in Natural Language Processing", "quality_score": 0.5318}
{"id": "c2805de89e758756ace71b5b0bf503f51d1a33cb8fe6ced0d7610a5e4b670e73", "sources": ["arxiv", "semantic_scholar"], "title": "Chain of Thought Prompt Tuning in Vision Language Models", "abstract": "Language-Image Pre-training has demonstrated promising results on zero-shot and few-shot downstream tasks by prompting visual models with natural language prompts. However, most recent studies only use a single prompt for tuning, neglecting the inherent step-to-step cognitive reasoning process that humans conduct in complex task settings, for example, when processing images from unfamiliar domains. Chain of Thought is a simple and effective approximation to human reasoning process and has been proven useful for natural language processing (NLP) tasks. Based on this cognitive intuition, we believe that conducting effective reasoning is also an important problem in visual tasks, and a chain of thought could be a solution to this problem. In this work, we propose a novel chain of thought prompt tuning for vision-language modeling. Extensive experiments show that our method not only generalizes better in image classification tasks, has greater transferability beyond a single dataset, and has stronger domain generalization performance, but also performs much better in imagetext retrieval and visual question answering, which require more reasoning capabilities. We are the first to successfully adapt chain-of-thought prompting that combines visual and textual embeddings. We will release our codes", "authors": ["Jiaxin Ge", "Hongyin Luo", "Siyuan Qian", "Yulu Gan", "Jie Fu", "Shanghang Zhang"], "categories": ["cs.CV", "cs.AI"], "fields_of_study": ["Computer Science"], "published_date": "2023-04-16", "url": "https://arxiv.org/abs/2304.07919", "pdf_url": "https://arxiv.org/pdf/2304.07919v2", "arxiv_id": "2304.07919", "doi": null, "citation_count": 37, "influential_citation_count": 1, "has_code": false, "code_url": null, "venue": null, "quality_score": 0.3949}
{"id": "be6dc34ec716459d4073c03ccb1632ef764eb5b4eaac19f6bc1e761d9bb4d5b6", "sources": ["arxiv", "semantic_scholar"], "title": "A Comprehensive Survey on Test-Time Adaptation under Distribution Shifts", "abstract": "Machine learning methods strive to acquire a robust model during the training process that can effectively generalize to test samples, even in the presence of distribution shifts. However, these methods often suffer from performance degradation due to unknown test distributions. Test-time adaptation (TTA), an emerging paradigm, has the potential to adapt a pre-trained model to unlabeled data during testing, before making predictions. Recent progress in this paradigm has highlighted the significant benefits of using unlabeled data to train self-adapted models prior to inference. In this survey, we categorize TTA into several distinct groups based on the form of test data, namely, test-time domain adaptation, test-time batch adaptation, and online test-time adaptation. For each category, we provide a comprehensive taxonomy of advanced algorithms and discuss various learning scenarios. Furthermore, we analyze relevant applications of TTA and discuss open challenges and promising areas for future research. For a comprehensive list of TTA methods, kindly refer to \\url{https://github.com/tim-learn/awesome-test-time-adaptation}.", "authors": ["Jian Liang", "Ran He", "Tieniu Tan"], "categories": ["cs.LG", "cs.AI", "cs.CV"], "fields_of_study": ["Computer Science"], "published_date": "2023-03-27", "url": "https://arxiv.org/abs/2303.15361", "pdf_url": "https://arxiv.org/pdf/2303.15361v2", "arxiv_id": "2303.15361", "doi": "10.1007/s11263-024-02181-w", "citation_count": 522, "influential_citation_count": 23, "has_code": true, "code_url": "https://github.com/tim-learn/awesome-test-time-adaptation}", "venue": "International Journal of Computer Vision", "quality_score": 0.6901}
{"id": "0d4815e0b535b725719263f80e4488d9be6e2ed02532185cfb0594075fd6a525", "sources": ["arxiv", "semantic_scholar"], "title": "Revisiting Realistic Test-Time Training: Sequential Inference and Adaptation by Anchored Clustering Regularized Self-Training", "abstract": "Deploying models on target domain data subject to distribution shift requires adaptation. Test-time training (TTT) emerges as a solution to this adaptation under a realistic scenario where access to full source domain data is not available, and instant inference on the target domain is required. Despite many efforts into TTT, there is a confusion over the experimental settings, thus leading to unfair comparisons. In this work, we first revisit TTT assumptions and categorize TTT protocols by two key factors. Among the multiple protocols, we adopt a realistic sequential test-time training (sTTT) protocol, under which we develop a test-time anchored clustering (TTAC) approach to enable stronger test-time feature learning. TTAC discovers clusters in both source and target domains and matches the target clusters to the source ones to improve adaptation. When source domain information is strictly absent (i.e. source-free) we further develop an efficient method to infer source domain distributions for anchored clustering. Finally, self-training~(ST) has demonstrated great success in learning from unlabeled data and we empirically figure out that applying ST alone to TTT is prone to confirmation bias. Therefore, a more effective TTT approach is introduced by regularizing self-training with anchored clustering, and the improved model is referred to as TTAC++. We demonstrate that, under all TTT protocols, TTAC++ consistently outperforms the state-of-the-art methods on five TTT datasets, including corrupted target domain, selected hard samples, synthetic-to-real adaptation and adversarially attacked target domain. We hope this work will provide a fair benchmarking of TTT methods, and future research should be compared within respective protocols.", "authors": ["Yongyi Su", "Xun Xu", "Tianrui Li", "Kui Jia"], "categories": ["cs.LG", "cs.CV"], "fields_of_study": ["Computer Science", "Medicine"], "published_date": "2023-03-20", "url": "https://arxiv.org/abs/2303.10856", "pdf_url": "https://arxiv.org/pdf/2303.10856v1", "arxiv_id": "2303.10856", "doi": "10.1109/TPAMI.2024.3370963", "citation_count": 22, "influential_citation_count": 1, "has_code": false, "code_url": null, "venue": "IEEE Transactions on Pattern Analysis and Machine Intelligence", "quality_score": 0.3404}
{"id": "982de00539d0413eabb348de37acb852f2d232694b50364c3f879a9ef1687092", "sources": ["arxiv", "semantic_scholar"], "title": "Multimodal Chain-of-Thought Reasoning in Language Models", "abstract": "Large language models (LLMs) have shown impressive performance on complex reasoning by leveraging chain-of-thought (CoT) prompting to generate intermediate reasoning chains as the rationale to infer the answer. However, existing CoT studies have primarily focused on the language modality. We propose Multimodal-CoT that incorporates language (text) and vision (images) modalities into a two-stage framework that separates rationale generation and answer inference. In this way, answer inference can leverage better generated rationales that are based on multimodal information. Experimental results on ScienceQA and A-OKVQA benchmark datasets show the effectiveness of our proposed approach. With Multimodal-CoT, our model under 1 billion parameters achieves state-of-the-art performance on the ScienceQA benchmark. Our analysis indicates that Multimodal-CoT offers the advantages of mitigating hallucination and enhancing convergence speed. Code is publicly available at https://github.com/amazon-science/mm-cot.", "authors": ["Zhuosheng Zhang", "Aston Zhang", "Mu Li", "Hai Zhao", "George Karypis", "Alex Smola"], "categories": ["cs.CL", "cs.AI", "cs.CV"], "fields_of_study": ["Computer Science"], "published_date": "2023-02-02", "url": "https://arxiv.org/abs/2302.00923", "pdf_url": "https://arxiv.org/pdf/2302.00923v5", "arxiv_id": "2302.00923", "doi": "10.48550/arXiv.2302.00923", "citation_count": 875, "influential_citation_count": 46, "has_code": true, "code_url": "https://github.com/amazon-science/mm-cot", "venue": null, "quality_score": 0.836}
{"id": "64ccfb70229c7d1af898fdb9cf9bd98cebfc0c6a4a6528a5b49283599896f7fc", "sources": ["arxiv", "semantic_scholar"], "title": "Faithful Chain-of-Thought Reasoning", "abstract": "While Chain-of-Thought (CoT) prompting boosts Language Models' (LM) performance on a gamut of complex reasoning tasks, the generated reasoning chain does not necessarily reflect how the model arrives at the answer (aka. faithfulness). We propose Faithful CoT, a reasoning framework involving two stages: Translation (Natural Language query $\\rightarrow$ symbolic reasoning chain) and Problem Solving (reasoning chain $\\rightarrow$ answer), using an LM and a deterministic solver respectively. This guarantees that the reasoning chain provides a faithful explanation of the final answer. Aside from interpretability, Faithful CoT also improves empirical performance: it outperforms standard CoT on 9 of 10 benchmarks from 4 diverse domains, with a relative accuracy gain of 6.3% on Math Word Problems (MWP), 3.4% on Planning, 5.5% on Multi-hop Question Answering (QA), and 21.4% on Relational Inference. Furthermore, with GPT-4 and Codex, it sets the new state-of-the-art few-shot performance on 7 datasets (with 95.0+ accuracy on 6 of them), showing a strong synergy between faithfulness and accuracy.", "authors": ["Qing Lyu", "Shreya Havaldar", "Adam Stein", "Li Zhang", "Delip Rao", "Eric Wong", "Marianna Apidianaki", "Chris Callison-Burch"], "categories": ["cs.CL"], "fields_of_study": ["Computer Science"], "published_date": "2023-01-31", "url": "https://arxiv.org/abs/2301.13379", "pdf_url": "https://arxiv.org/pdf/2301.13379v3", "arxiv_id": "2301.13379", "doi": "10.48550/arXiv.2301.13379", "citation_count": 406, "influential_citation_count": 17, "has_code": false, "code_url": null, "venue": "International Joint Conference on Natural Language Processing", "quality_score": 0.6524}
{"id": "b5f91c641d0e4642b14bc3a1bedb90259ee51f590df25518d14ed2f9a930a656", "sources": ["arxiv", "semantic_scholar"], "title": "Chaining Simultaneous Thoughts for Numerical Reasoning", "abstract": "Given that rich information is hidden behind ubiquitous numbers in text, numerical reasoning over text should be an essential skill of AI systems. To derive precise equations to solve numerical reasoning problems, previous work focused on modeling the structures of equations, and has proposed various structured decoders. Though structure modeling proves to be effective, these structured decoders construct a single equation in a pre-defined autoregressive order, potentially placing an unnecessary restriction on how a model should grasp the reasoning process. Intuitively, humans may have numerous pieces of thoughts popping up in no pre-defined order; thoughts are not limited to the problem at hand, and can even be concerned with other related problems. By comparing diverse thoughts and chaining relevant pieces, humans are less prone to errors. In this paper, we take this inspiration and propose CANTOR, a numerical reasoner that models reasoning steps using a directed acyclic graph where we produce diverse reasoning steps simultaneously without pre-defined decoding dependencies, and compare and chain relevant ones to reach a solution. Extensive experiments demonstrated the effectiveness of CANTOR under both fully-supervised and weakly-supervised settings.", "authors": ["Zhihong Shao", "Fei Huang", "Minlie Huang"], "categories": ["cs.CL", "cs.AI"], "fields_of_study": ["Computer Science"], "published_date": "2022-11-29", "url": "https://arxiv.org/abs/2211.16482", "pdf_url": "https://arxiv.org/pdf/2211.16482v1", "arxiv_id": "2211.16482", "doi": "10.48550/arXiv.2211.16482", "citation_count": 21, "influential_citation_count": 0, "has_code": false, "code_url": null, "venue": "Conference on Empirical Methods in Natural Language Processing", "quality_score": 0.3356}
{"id": "4eeb0f40bd9b0c64fe3fe5fc898534ab7c2ef7884a93917fb5d7b10b8605fb8c", "sources": ["arxiv", "semantic_scholar"], "title": "Program of Thoughts Prompting: Disentangling Computation from Reasoning for Numerical Reasoning Tasks", "abstract": "Recently, there has been significant progress in teaching language models to perform step-by-step reasoning to solve complex numerical reasoning tasks. Chain-of-thoughts prompting (CoT) is by far the state-of-art method for these tasks. CoT uses language models to perform both reasoning and computation in the multi-step `thought' process. To disentangle computation from reasoning, we propose `Program of Thoughts' (PoT), which uses language models (mainly Codex) to express the reasoning process as a program. The computation is relegated to an external computer, which executes the generated programs to derive the answer. We evaluate PoT on five math word problem datasets (GSM, AQuA, SVAMP, TabMWP, MultiArith) and three financial-QA datasets (FinQA, ConvFinQA, TATQA) for both few-shot and zero-shot setups. Under both few-shot and zero-shot settings, PoT can show an average performance gain over CoT by around 12\\% across all the evaluated datasets. By combining PoT with self-consistency decoding, we can achieve SoTA performance on all math problem datasets and near-SoTA performance on financial datasets. All of our data and code are released in Github https://github.com/wenhuchen/Program-of-Thoughts", "authors": ["Wenhu Chen", "Xueguang Ma", "Xinyi Wang", "William W. Cohen"], "categories": ["cs.CL", "cs.AI"], "fields_of_study": ["Computer Science"], "published_date": "2022-11-22", "url": "https://arxiv.org/abs/2211.12588", "pdf_url": "https://arxiv.org/pdf/2211.12588v4", "arxiv_id": "2211.12588", "doi": null, "citation_count": 1357, "influential_citation_count": 129, "has_code": true, "code_url": "https://github.com/wenhuchen/Program-of-Thoughts", "venue": null, "quality_score": 1.0}
{"id": "1abbcb5e1135aa8c802da791b14306f546b5b2f899d5aff4cc31605ce318ba07", "sources": ["arxiv", "semantic_scholar"], "title": "Exploitation of material consolidation trade-offs in multi-tier complex supply networks", "abstract": "While consolidation strategies form the backbone of many supply chain optimisation problems, exploitation of multi-tier material relationships through consolidation remains an understudied area, despite being a prominent feature of industries that produce complex made-to-order products. In this paper, we propose an optimisation framework for exploiting multi-to-multi relationship between tiers of a supply chain. The resulting formulation is flexible such that quantity discounts, inventory holding, and transport costs can be included. The framework introduces a new trade-off between tiers, leading to cost reductions in one tier but increased costs in the other, which helps to reduce the overall procurement cost in the supply chain. A mixed integer linear programming model is developed and tested with a range of small to large-scale test problems from aerospace manufacturing. Our comparison to benchmark results shows that there is indeed a cost trade-off between two tiers, and that its reduction can be achieved using a holistic approach to reconfiguration. Costs are decreased when second tier fixed ordering costs and the number of machining options increase. Consolidation results in reduced inventory holding costs in all scenarios. Several secondary effects such as simplified supplier selection may also be observed.", "authors": ["Vinod Kumar Chauhan", "Muhannad Alomari", "James Arney", "Ajith Kumar Parlikad", "Alexandra Brintrup"], "categories": ["cs.CE"], "fields_of_study": ["Computer Science"], "published_date": "2022-10-19", "url": "https://arxiv.org/abs/2210.11479", "pdf_url": "https://arxiv.org/pdf/2210.11479v3", "arxiv_id": "2210.11479", "doi": "10.1016/j.sca.2023.100050", "citation_count": 5, "influential_citation_count": 0, "has_code": false, "code_url": null, "venue": "Supply Chain Analytics, Volume 4, December 2023, 100050", "quality_score": 0.1945}
{"id": "9a4252a531082aa575b7229db40d5ff066b6ea5761b01dc001834d4577fa475f", "sources": ["arxiv", "semantic_scholar"], "title": "Language Models are Multilingual Chain-of-Thought Reasoners", "abstract": "We evaluate the reasoning abilities of large language models in multilingual settings. We introduce the Multilingual Grade School Math (MGSM) benchmark, by manually translating 250 grade-school math problems from the GSM8K dataset (Cobbe et al., 2021) into ten typologically diverse languages. We find that the ability to solve MGSM problems via chain-of-thought prompting emerges with increasing model scale, and that models have strikingly strong multilingual reasoning abilities, even in underrepresented languages such as Bengali and Swahili. Finally, we show that the multilingual reasoning abilities of language models extend to other tasks such as commonsense reasoning and word-in-context semantic judgment. The MGSM benchmark is publicly available at https://github.com/google-research/url-nlp.", "authors": ["Freda Shi", "Mirac Suzgun", "Markus Freitag", "Xuezhi Wang", "Suraj Srivats", "Soroush Vosoughi", "Hyung Won Chung", "Yi Tay", "Sebastian Ruder", "Denny Zhou", "Dipanjan Das", "Jason Wei"], "categories": ["cs.CL", "cs.AI", "cs.LG"], "fields_of_study": ["Computer Science"], "published_date": "2022-10-06", "url": "https://arxiv.org/abs/2210.03057", "pdf_url": "https://arxiv.org/pdf/2210.03057v1", "arxiv_id": "2210.03057", "doi": "10.48550/arXiv.2210.03057", "citation_count": 603, "influential_citation_count": 112, "has_code": true, "code_url": "https://github.com/google-research/url-nlp", "venue": "International Conference on Learning Representations", "quality_score": 1.0}
{"id": "8c0f5203a0adc1cb0afb686f7f4f48566ff38c7ba592a72540bc7f92f8b5d500", "sources": ["arxiv", "semantic_scholar"], "title": "Language Models Are Greedy Reasoners: A Systematic Formal Analysis of Chain-of-Thought", "abstract": "Large language models (LLMs) have shown remarkable reasoning capabilities given chain-of-thought prompts (examples with intermediate reasoning steps). Existing benchmarks measure reasoning ability indirectly, by evaluating accuracy on downstream tasks such as mathematical reasoning. However, it is unclear how these models obtain the answers and whether they rely on simple heuristics rather than the generated chain-of-thought. To enable systematic exploration of the reasoning ability of LLMs, we present a new synthetic question-answering dataset called PrOntoQA, where each example is generated from a synthetic world model represented in first-order logic. This allows us to parse the generated chain-of-thought into symbolic proofs for formal analysis. Our analysis on InstructGPT and GPT-3 shows that LLMs are quite capable of making correct individual deduction steps, and so are generally capable of reasoning, even in fictional contexts. However, they have difficulty with proof planning: When multiple valid deduction steps are available, they are not able to systematically explore the different options.", "authors": ["Abulhair Saparov", "He He"], "categories": ["cs.CL"], "fields_of_study": ["Computer Science"], "published_date": "2022-10-03", "url": "https://arxiv.org/abs/2210.01240", "pdf_url": "https://arxiv.org/pdf/2210.01240v4", "arxiv_id": "2210.01240", "doi": "10.48550/arXiv.2210.01240", "citation_count": 496, "influential_citation_count": 77, "has_code": false, "code_url": null, "venue": "International Conference on Learning Representations", "quality_score": 0.946}
{"id": "79e099bcdcfa2aa4b7c0cc7c3f2dd92b45c50ea455fb92e731555b98146f3333", "sources": ["arxiv", "semantic_scholar"], "title": "Learn to Explain: Multimodal Reasoning via Thought Chains for Science Question Answering", "abstract": "When answering a question, humans utilize the information available across different modalities to synthesize a consistent and complete chain of thought (CoT). This process is normally a black box in the case of deep learning models like large-scale language models. Recently, science question benchmarks have been used to diagnose the multi-hop reasoning ability and interpretability of an AI system. However, existing datasets fail to provide annotations for the answers, or are restricted to the textual-only modality, small scales, and limited domain diversity. To this end, we present Science Question Answering (ScienceQA), a new benchmark that consists of ~21k multimodal multiple choice questions with a diverse set of science topics and annotations of their answers with corresponding lectures and explanations. We further design language models to learn to generate lectures and explanations as the chain of thought (CoT) to mimic the multi-hop reasoning process when answering ScienceQA questions. ScienceQA demonstrates the utility of CoT in language models, as CoT improves the question answering performance by 1.20% in few-shot GPT-3 and 3.99% in fine-tuned UnifiedQA. We also explore the upper bound for models to leverage explanations by feeding those in the input; we observe that it improves the few-shot performance of GPT-3 by 18.96%. Our analysis further shows that language models, similar to humans, benefit from explanations to learn from fewer data and achieve the same performance with just 40% of the data. The data and code are available at https://scienceqa.github.io.", "authors": ["Pan Lu", "Swaroop Mishra", "Tony Xia", "Liang Qiu", "Kai-Wei Chang", "Song-Chun Zhu", "Oyvind Tafjord", "Peter Clark", "Ashwin Kalyan"], "categories": ["cs.CL", "cs.AI", "cs.CV", "cs.LG", "cs.MM"], "fields_of_study": ["Computer Science"], "published_date": "2022-09-20", "url": "https://arxiv.org/abs/2209.09513", "pdf_url": "https://arxiv.org/pdf/2209.09513v2", "arxiv_id": "2209.09513", "doi": "10.48550/arXiv.2209.09513", "citation_count": 2395, "influential_citation_count": 266, "has_code": false, "code_url": null, "venue": "Neural Information Processing Systems", "quality_score": 1.0}
{"id": "cb37175ec5d6baccb756c615ca345e9b22d8c83a659b6b51c9045c14b412e8bd", "sources": ["arxiv", "semantic_scholar"], "title": "Classicals versus Keynesians: Fifty Distinctions between Two Major Schools of Economic Thought", "abstract": "Macroeconomics essentially discusses macroeconomic phenomena from the perspectives of various schools of economic thought, each of which takes different views on how macroeconomic agents make decisions and how the corresponding markets operate. Therefore, developing a clear, comprehensive understanding of how and in what ways these schools of economic thought differ is a key and a prerequisite for economics students to prosper academically and professionally in the discipline. This becomes even more crucial as economics students pursue their studies toward higher levels of education and graduate school, during which students are expected to attain higher levels of Bloom's taxonomy, including analysis, synthesis, evaluation, and creation. Teaching the distinctions and similarities of the two major schools of economic thought has never been an easy task to undertake in the classroom. Although the reason for such a hardship can be multi-fold, one reason has undoubtedly been students' lack of a holistic view on how the two mainstream economic schools of thought differ. There is strong evidence that students make smoother transition to higher levels of education after building up such groundwork, on which they can build further later on (e.g. Didia and Hasnat, 1998; Marcal and Roberts, 2001; Islam, et al., 2008; Green, et al., 2009; White, 2016). The paper starts with a visual spectrum of various schools of economic thought, and then narrows down the scope to the classical and Keynesian schools, i.e. the backbone of modern macroeconomics. Afterwards, a holistic table contrasts the two schools in terms of 50 aspects. Not only does this table help economics students enhance their comprehension, retention, and critical-thinking capability, it also benefits macroeconomic instructors to ...", "authors": ["Seyyed Ali Zeytoon Nejad Moosavian"], "categories": ["econ.GN"], "fields_of_study": ["Economics"], "published_date": "2022-09-06", "url": "https://arxiv.org/abs/2209.02683", "pdf_url": "https://arxiv.org/pdf/2209.02683v1", "arxiv_id": "2209.02683", "doi": "10.1453/jest.v9i2.2325", "citation_count": 1, "influential_citation_count": 0, "has_code": false, "code_url": null, "venue": "Journal of Economic and Social Thought, 9(2), 63-79 (2022)", "quality_score": 0.0753}
{"id": "ec26d0073422d3ea40723d11a1cc1ac091edfaab2ea9cae01f956485aab54941", "sources": ["arxiv", "semantic_scholar"], "title": "Multi-Step Deductive Reasoning Over Natural Language: An Empirical Study on Out-of-Distribution Generalisation", "abstract": "Combining deep learning with symbolic logic reasoning aims to capitalize on the success of both fields and is drawing increasing attention. Inspired by DeepLogic, an end-to-end model trained to perform inference on logic programs, we introduce IMA-GloVe-GA, an iterative neural inference network for multi-step reasoning expressed in natural language. In our model, reasoning is performed using an iterative memory neural network based on RNN with a gated attention mechanism. We evaluate IMA-GloVe-GA on three datasets: PARARULES, CONCEPTRULES V1 and CONCEPTRULES V2. Experimental results show DeepLogic with gated attention can achieve higher test accuracy than DeepLogic and other RNN baseline models. Our model achieves better out-of-distribution generalisation than RoBERTa-Large when the rules have been shuffled. Furthermore, to address the issue of unbalanced distribution of reasoning depths in the current multi-step reasoning datasets, we develop PARARULE-Plus, a large dataset with more examples that require deeper reasoning steps. Experimental results show that the addition of PARARULE-Plus can increase the model's performance on examples requiring deeper reasoning depths. The source code and data are available at https://github.com/Strong-AI-Lab/Multi-Step-Deductive-Reasoning-Over-Natural-Language.", "authors": ["Qiming Bao", "Alex Yuxuan Peng", "Tim Hartill", "Neset Tan", "Zhenyun Deng", "Michael Witbrock", "Jiamou Liu"], "categories": ["cs.CL", "cs.AI", "cs.LG", "cs.LO"], "fields_of_study": ["Computer Science"], "published_date": "2022-07-28", "url": "https://arxiv.org/abs/2207.14000", "pdf_url": "https://arxiv.org/pdf/2207.14000v4", "arxiv_id": "2207.14000", "doi": "10.48550/arXiv.2207.14000", "citation_count": 21, "influential_citation_count": 3, "has_code": true, "code_url": "https://github.com/Strong-AI-Lab/Multi-Step-Deductive-Reasoning-Over-Natural-Language", "venue": "International Workshop on Neural-Symbolic Learning and Reasoning", "quality_score": 0.3356}
{"id": "1ee33f85ad219442364f0315f1ec81ebd8a43f405eb55f28a8631f6b237cc618", "sources": ["arxiv", "semantic_scholar"], "title": "Self-Consistency Improves Chain of Thought Reasoning in Language Models", "abstract": "Chain-of-thought prompting combined with pre-trained large language models has achieved encouraging results on complex reasoning tasks. In this paper, we propose a new decoding strategy, self-consistency, to replace the naive greedy decoding used in chain-of-thought prompting. It first samples a diverse set of reasoning paths instead of only taking the greedy one, and then selects the most consistent answer by marginalizing out the sampled reasoning paths. Self-consistency leverages the intuition that a complex reasoning problem typically admits multiple different ways of thinking leading to its unique correct answer. Our extensive empirical evaluation shows that self-consistency boosts the performance of chain-of-thought prompting with a striking margin on a range of popular arithmetic and commonsense reasoning benchmarks, including GSM8K (+17.9%), SVAMP (+11.0%), AQuA (+12.2%), StrategyQA (+6.4%) and ARC-challenge (+3.9%).", "authors": ["Xuezhi Wang", "Jason Wei", "Dale Schuurmans", "Quoc Le", "Ed Chi", "Sharan Narang", "Aakanksha Chowdhery", "Denny Zhou"], "categories": ["cs.CL", "cs.AI"], "fields_of_study": ["Computer Science"], "published_date": "2022-03-21", "url": "https://arxiv.org/abs/2203.11171", "pdf_url": "https://arxiv.org/pdf/2203.11171v4", "arxiv_id": "2203.11171", "doi": null, "citation_count": 6930, "influential_citation_count": 920, "has_code": false, "code_url": null, "venue": "International Conference on Learning Representations", "quality_score": 1.0}
{"id": "b68ae3068e9b87113d6c365d64ba4ef55e8d4748b2924ca0386b7378b9119e8a", "sources": ["arxiv", "semantic_scholar"], "title": "IID Time Series Testing", "abstract": "Traditional white noise testing, for example the Ljung-Box test, studies only the autocorrelation function (ACF). Time series can be heteroscedastic and therefore not i.i.d. but still white noise (that is, with zero ACF). An example of heteroscedasticity is financial time series: times of high variance (financial crises) can alternate with times of low variance (calm times). Here, absolute values of time series terms are not white noise. We could test for white noise separately for original and absolute values, for example using Ljung-Box tests for both. In this article, we create an omnibus test which combines these two tests. Moreover, we create a general framework to create various i.i.d. tests. We apply tests to simulated data, both autoregressive linear and heteroscedastic.", "authors": ["Andrey Sarantsev"], "categories": ["math.ST"], "fields_of_study": ["Mathematics"], "published_date": "2022-03-19", "url": "https://arxiv.org/abs/2203.10405", "pdf_url": "https://arxiv.org/pdf/2203.10405v2", "arxiv_id": "2203.10405", "doi": "10.3842/tsp-8836211480-29", "citation_count": 0, "influential_citation_count": 0, "has_code": false, "code_url": null, "venue": "Theory of Stochastic Processes", "quality_score": 0.0}
{"id": "1c6206bf3f3b00fc2222bf8cd9639b2bb7d9fa4e6cf0f3f4936f14df34e41bf7", "sources": ["arxiv", "semantic_scholar"], "title": "Chain-of-Thought Prompting Elicits Reasoning in Large Language Models", "abstract": "We explore how generating a chain of thought -- a series of intermediate reasoning steps -- significantly improves the ability of large language models to perform complex reasoning. In particular, we show how such reasoning abilities emerge naturally in sufficiently large language models via a simple method called chain of thought prompting, where a few chain of thought demonstrations are provided as exemplars in prompting. Experiments on three large language models show that chain of thought prompting improves performance on a range of arithmetic, commonsense, and symbolic reasoning tasks. The empirical gains can be striking. For instance, prompting a 540B-parameter language model with just eight chain of thought exemplars achieves state of the art accuracy on the GSM8K benchmark of math word problems, surpassing even finetuned GPT-3 with a verifier.", "authors": ["Jason Wei", "Xuezhi Wang", "Dale Schuurmans", "Maarten Bosma", "Brian Ichter", "Fei Xia", "Ed Chi", "Quoc Le", "Denny Zhou"], "categories": ["cs.CL", "cs.AI"], "fields_of_study": ["Computer Science"], "published_date": "2022-01-28", "url": "https://arxiv.org/abs/2201.11903", "pdf_url": "https://arxiv.org/pdf/2201.11903v6", "arxiv_id": "2201.11903", "doi": "10.52202/068431-1800", "citation_count": 19220, "influential_citation_count": 1309, "has_code": false, "code_url": null, "venue": "Neural Information Processing Systems", "quality_score": 1.0}