COLM
Collection
Accepted papers for COLM (Conference on Language Modeling), one dataset per year. • 2 items • Updated
paper_id stringlengths 10 10 | title stringlengths 21 128 | paper_url stringlengths 42 42 | authors listlengths 1 42 | type stringclasses 0
values | primary_area stringclasses 0
values | abstract large_stringlengths 613 2.09k | keywords listlengths 1 16 | TL;DR large_stringlengths 40 250 | submission_number int64 7 1.95k | arxiv_id stringlengths 10 10 ⌀ | arxiv_id_source stringclasses 2
values |
|---|---|---|---|---|---|---|---|---|---|---|---|
SlRtFwBdzP | Assessing Judging Bias in Large Reasoning Models: An Empirical Study | https://openreview.net/forum?id=SlRtFwBdzP | [
"Qian Wang",
"Zhanzhi Lou",
"Zhenheng Tang",
"Nuo Chen",
"Xuandong Zhao",
"Wenxuan Zhang",
"Dawn Song",
"Bingsheng He"
] | null | null | Large Reasoning Models (LRMs) like DeepSeek-R1 and OpenAI-o1 have demonstrated remarkable reasoning capabilities, raising important questions about their biases in LLM-as-a-judge settings. We present a comprehensive benchmark comparing judging biases between LLMs and LRMs across both subjective preference-alignment dat... | [
"Large Reasoning Models",
"LLM Evaluation"
] | We demonstrate that Large Reasoning Models remain susceptible to judging biases despite their advanced capabilities. | 7 | 2504.09946 | title_snapshot |
eqNItk1sWo | Discovering Knowledge Deficiencies of Language Models on Massive Knowledge Base | https://openreview.net/forum?id=eqNItk1sWo | [
"Linxin Song",
"Xuwei Ding",
"Jieyu Zhang",
"Taiwei Shi",
"Ryotaro Shimizu",
"Rahul Gupta",
"Yang Liu",
"Jian Kang",
"Jieyu Zhao"
] | null | null | Large language models (LLMs) possess impressive linguistic capabilities but often fail to faithfully retain factual knowledge, leading to hallucinations and unreliable outputs. Understanding LLMs' knowledge deficiencies by exhaustively evaluating against full-scale knowledge bases is computationally prohibitive, especi... | [
"Large Language Models",
"Evaluation",
"Misinformation"
] | We propose stochastic error ascend, an optimization-based framework that efficiently identifies and refines failure modes in LLMs, discovering significantly more errors than existing methods while reducing evaluation costs. | 14 | 2503.23361 | title_snapshot |
2Kl8Ztw6wk | PredGen: Accelerated Inference of Large Language Models through Input-Time Speculation for Real-Time Speech Interaction | https://openreview.net/forum?id=2Kl8Ztw6wk | [
"Shufan Li",
"Aditya Grover"
] | null | null | Large Language Models (LLMs) are widely used in real-time voice chat applications, typically in combination with text-to-speech (TTS) systems to generate audio responses. However, their large size often leads to noticeable latency between the end of user input and the start of audio output, resulting in suboptimal user... | [
"Large Language Models",
"Inference",
"Speculative Decoding"
] | We leverage speculative decoding at user input time to reduce the latency of speech interactions. | 16 | 2506.15556 | title_snapshot |
gIqb6zWZoO | KVSink: Understanding and Enhancing the Preservation of Attention Sinks in KV Cache Quantization for LLMs | https://openreview.net/forum?id=gIqb6zWZoO | [
"Zunhai Su",
"Kehong Yuan"
] | null | null | Key-Value (KV) cache quantization has become a widely adopted optimization technique for efficient large language models (LLMs) inference by reducing KV cache memory usage and mitigating memory-bound constraints.
Recent studies have emphasized the importance of preserving the original precision of KVs for the first few... | [
"quantization",
"kv cache",
"transformer",
"llm",
"attention"
] | Understanding and Enhancing the Preservation of Attention Sinks in KV Cache Quantization for LLMs | 22 | 2508.04257 | title_snapshot |
h99hJlU99U | Overflow Prevention Enhances Long-Context Recurrent LLMs | https://openreview.net/forum?id=h99hJlU99U | [
"Assaf Ben-Kish",
"Itamar Zimerman",
"Muhammad Jehanzeb Mirza",
"Lior Wolf",
"James R. Glass",
"Leonid Karlinsky",
"Raja Giryes"
] | null | null | A recent trend in LLMs is developing recurrent sub-quadratic models that improve long-context processing efficiency. We investigate leading large long-context models, focusing on how their fixed-size recurrent memory affects their performance. Our experiments reveal that, even when these models are trained for extended... | [
"Mamba",
"Sub-Quadratic Models",
"Long Context",
"Long-Range Language Modeling",
"RNNs"
] | We identify that recurrent LLMs suffer from recurrent memory overflows that limit their performance in long-context tasks. We propose OPRM, a training-free overflow-prevention mechanism that achieves significant gains in many long-context tasks. | 25 | 2505.07793 | title_snapshot |
NRrXHppaBg | Efficient Self-Improvement in Multimodal Large Language Models: A Model-Level Judge-Free Approach | https://openreview.net/forum?id=NRrXHppaBg | [
"Shijian Deng",
"Wentian Zhao",
"Yu-Jhe Li",
"Kun Wan",
"Daniel Miranda",
"Ajinkya Kale",
"Yapeng Tian"
] | null | null | Self-improvement in multimodal large language models (MLLMs) is crucial for enhancing their reliability and robustness. However, current methods often rely heavily on MLLMs themselves as judges, leading to high computational costs and potential pitfalls like reward hacking and model collapse.
This paper introduces a no... | [
"Self-Improvement",
"Multimodal Large Language Models"
] | A novel judge-free self-improvement framework for multimodal large language models (MLLMs) efficiently enhances reliability by controlling hallucinations without costly model-level verification loops. | 27 | 2411.17760 | title_snapshot |
qMUbhGUFUb | SmolVLM: Redefining small and efficient multimodal models | https://openreview.net/forum?id=qMUbhGUFUb | [
"Andrés Marafioti",
"Orr Zohar",
"Miquel Farré",
"Merve noyan",
"Elie Bakouch",
"Pedro Manuel Cuenca Jiménez",
"Cyril Zakka",
"Loubna Ben allal",
"Anton Lozhkov",
"Nouamane Tazi",
"Vaibhav Srivastav",
"Joshua Lochner",
"Hugo Larcher",
"Mathieu Morlon",
"Lewis Tunstall",
"Leandro Von We... | null | null | Large Vision-Language Models (VLMs) deliver exceptional performance but require significant computational resources, limiting their deployment on mobile and edge devices. Smaller VLMs typically mirror design choices of larger models, such as extensive image tokenization, leading to inefficient GPU memory usage and cons... | [
"Vision Language Models",
"Large Multimodal Models",
"Vision Understanding",
"Video Understanding"
] | We explore extremely efficient VLMs starting at 256M parameters, which run with less than 1GB | 29 | 2504.05299 | title_snapshot |
QTrW2HWNXe | Language Model Uncertainty Quantification with Attention Chain | https://openreview.net/forum?id=QTrW2HWNXe | [
"Yinghao Li",
"Rushi Qiang",
"Lama Moukheiber",
"Chao Zhang"
] | null | null | Accurately quantifying a large language model's (LLM) predictive uncertainty is crucial for judging the reliability of its answers.
While most existing research focuses on short, directly answerable questions with closed-form outputs (e.g., multiple-choice), involving intermediate reasoning steps in LLM responses is in... | [
"Uncertainty Estimation",
"Large Language Model",
"Attention"
] | Investigates large language model uncertainty estimation with critical reasoning token backtracking using attention chain. | 30 | 2503.19168 | title_snapshot |
qQb1JLrwol | Hidden in plain sight: VLMs overlook their visual representations | https://openreview.net/forum?id=qQb1JLrwol | [
"Stephanie Fu",
"tyler bonnen",
"Devin Guillory",
"Trevor Darrell"
] | null | null | Language provides a natural interface to specify and evaluate performance on visual tasks.
To realize this possibility, vision language models (VLMs) must successfully integrate visual and linguistic information.
Our work compares VLMs to a direct readout of their visual encoders to understand their ability to integra... | [
"vision",
"language",
"representation",
"benchmark",
"encoder",
"vlm"
] | VLMs perform worse on vision-centric tasks than their underlying vision models, relying on their language priors instead. Improving their integration of visual data—not just adding stronger vision backbones—is key to unlocking their full potential. | 31 | 2506.08008 | title_snapshot |
gKdhzBiHay | SQuat: Subspace-orthogonal KV Cache Quantization | https://openreview.net/forum?id=gKdhzBiHay | [
"Hao Wang",
"Ligong Han",
"Kai Xu",
"Akash Srivastava"
] | null | null | The key-value (KV) cache accelerates LLMs decoding by storing KV tensors from previously generated tokens. It reduces redundant computation at the cost of increased memory usage. To mitigate this overhead, existing approaches compress KV tensors into lower-bit representations; however, quantization errors can accumulat... | [
"KV cache quantization",
"LLMs"
] | We propose a KV cache quantization method that preserves task-critical information throughout the quantization process. | 34 | 2503.24358 | title_snapshot |
AivRDOFi5H | Language Models Fail to Introspect About Their Knowledge of Language | https://openreview.net/forum?id=AivRDOFi5H | [
"Siyuan Song",
"Jennifer Hu",
"Kyle Mahowald"
] | null | null | There has been recent interest in whether large language models (LLMs) can introspect about their own internal states. Such abilities would make LLMs more interpretable, and also validate the use of standard introspective methods in linguistics to evaluate grammatical knowledge in models (e.g., asking "Is this sentence... | [
"introspection",
"linguistic acceptability judgments",
"syntax",
"grammaticality",
"surprisal",
"metalinguistic",
"metacognition"
] | We study introspection in language models by comparing direct probability measurements with responses to metalinguistic prompts in two domains (grammaticality and word prediction) and find no clear evidence of introspection. | 35 | 2503.07513 | title_snapshot |
bNTrKqqnG9 | The Dual-Route Model of Induction | https://openreview.net/forum?id=bNTrKqqnG9 | [
"Sheridan Feucht",
"Eric Todd",
"Byron C Wallace",
"David Bau"
] | null | null | Prior work on in-context copying has shown the existence of *induction heads*, which attend to and promote individual tokens during copying. In this work we discover a new type of induction head: *concept-level* induction heads, which copy entire lexical units instead of individual tokens. Concept induction heads learn... | [
"interpretability",
"induction heads",
"in-context learning",
"ICL",
"detokenization"
] | We find that LLMs can do in-context copying in two different ways: either by copying individual tokens verbatim, or by copying entire word meanings (which may span multiple tokens). | 36 | 2504.03022 | title_snapshot |
WzGypILLDb | DFRot: Achieving Outlier-Free and Massive Activation-Free for Rotated LLMs with Refined Rotation | https://openreview.net/forum?id=WzGypILLDb | [
"Jingyang Xiang",
"Sai Qian Zhang"
] | null | null | Rotating the activation and weight matrices to reduce the influence of outliers in large language models (LLMs) has recently attracted significant attention, particularly in the context of model quantization. Prior studies have shown that in low-precision quantization scenarios, such as 4-bit weights and 4-bit activati... | [
"W4A4 quantization",
"randomized hadamard transforms",
"randomized orthogonal transforms",
"outlier",
"masssive activation"
] | We explained why random Hadamard is superior to randomized orthogonal transforms in the W4A4 quantization process and proposed an optimization method for the rotation matrix. | 47 | 2412.00648 | title_snapshot |
yGQqTuSJPK | Hypothesis-Driven Theory-of-Mind Reasoning for Large Language Models | https://openreview.net/forum?id=yGQqTuSJPK | [
"Hyunwoo Kim",
"Melanie Sclar",
"Tan Zhi-Xuan",
"Lance Ying",
"Sydney Levine",
"Yang Liu",
"Joshua B. Tenenbaum",
"Yejin Choi"
] | null | null | Existing LLM reasoning methods have shown impressive capabilities across various tasks, such as solving math and coding problems. However, applying these methods to scenarios without ground-truth answers or rule-based verification methods - such as tracking the mental states of an agent - remains challenging. Inspired ... | [
"theory of mind",
"reasoning",
"large language model",
"inference-time algorithm"
] | We introduce a novel inference-time algorithm, ThoughtTracing, which uses LLMs to probabilistically trace and weight hypotheses about agents’ evolving mental states without relying on questions and ground-truth answers in benchmarks. | 48 | 2502.11881 | title_snapshot |
PYHwlyu2fa | VisOnlyQA: Large Vision Language Models Still Struggle with Visual Perception of Geometric Information | https://openreview.net/forum?id=PYHwlyu2fa | [
"Ryo Kamoi",
"Yusen Zhang",
"Sarkar Snigdha Sarathi Das",
"Ranran Haoran Zhang",
"Rui Zhang"
] | null | null | Large Vision Language Models (LVLMs) have achieved remarkable performance in various vision-language tasks. However, it is still unclear how accurately LVLMs can perceive visual information in images. In particular, the capability of LVLMs to perceive geometric information, such as shape, angle, and size, remains insuf... | [
"vision-language models"
] | We introduce VisOnlyQA, a dataset to evaluate the capability of Large Vision Language Models to perceive geometric information, such as lengths, angles, and shapes, and reveal that they still cannot accurately perceive basic geometric information. | 55 | 2412.00947 | title_snapshot |
jSmpq7IRYe | Can Test-Time Scaling Improve World Foundation Model? | https://openreview.net/forum?id=jSmpq7IRYe | [
"Wenyan Cong",
"Hanqing Zhu",
"Peihao Wang",
"Bangya Liu",
"Dejia Xu",
"Kevin Wang",
"David Z. Pan",
"Yan Wang",
"Zhiwen Fan",
"Zhangyang Wang"
] | null | null | World foundation models, which simulate the physical world by predicting future states from current observations and inputs, have become central to many applications in physical intelligence, including autonomous driving and robotics. However, these models require substantial computational resources for pretraining and... | [
"world foundation model",
"test time scaling",
"autoregressive video generation",
"evaluation tookit"
] | Test time scaling could work for world foundation models and shows a clear scaling law. Smaller model with test time scaling is even better than x2 larger pretrained model. A extensible tookit will be opensourced for evaluating performance of WFM. | 59 | 2503.24320 | title_snapshot |
7ZwuGZCopw | FineMedLM-o1: Enhancing Medical Knowledge Reasoning Ability of LLM from Supervised Fine-Tuning to Test-Time Training | https://openreview.net/forum?id=7ZwuGZCopw | [
"hongzhou yu",
"Tianhao Cheng",
"Yingwen Wang",
"Wen He",
"Qing Wang",
"Ying Cheng",
"Yuejie Zhang",
"Rui Feng",
"Xiaobo Zhang"
] | null | null | Recent advancements in large language models (LLMs) have shown promise in medical applications such as disease diagnosis and treatment planning. However, most existing medical LLMs struggle with the deep reasoning required for complex medical problems, such as differential diagnosis and medication recommendations. We p... | [
"LMs on diverse domains and novel applications"
] | We propose a novel synthetic data method to generate thinking data and integrate Test-Time Training during the inference phase to enhance the medical reasoning capabilities of LLMs. | 60 | 2501.09213 | title_snapshot |
p0BwJk3R1p | LLMs as Research Tools: A Large Scale Survey of Researchers’ Usage and Perceptions | https://openreview.net/forum?id=p0BwJk3R1p | [
"Zhehui Liao",
"Maria Antoniak",
"Inyoung Cheong",
"Evie Yu-Yen Cheng",
"Ai-Heng Lee",
"Kyle Lo",
"Joseph Chee Chang",
"Amy X Zhang"
] | null | null | The rise of large language models (LLMs) has led many researchers to consider their usage for scientific work. Some have found benefits using LLMs to augment or automate aspects of their research pipeline, while others have urged caution due to risks and ethical concerns. Yet little work has sought to quantify and char... | [
"survey",
"large language model",
"research",
"societal impact"
] | A large-scale survey of 816 researchers to study usage of LLMs in scientific research and the perception of such usage. | 61 | 2411.05025 | title_snapshot |
1Pmuw08LoM | Modifying Large Language Model Post-Training for Diverse Creative Writing | https://openreview.net/forum?id=1Pmuw08LoM | [
"John Joon Young Chung",
"Vishakh Padmakumar",
"Melissa Roemmele",
"Yuqian Sun",
"Max Kreminski"
] | null | null | As creative writing tasks do not have singular correct answers, large language models (LLMs) trained to perform these tasks should be able to generate diverse valid outputs. However, LLM post-training often focuses on improving generation quality but neglects to facilitate output diversity. Hence, in creative writing g... | [
"creative writing generation",
"diversity",
"post-training"
] | In creative writing generation, we facilitate diversity in LLM outputs by counting in how each training instance differs from other instances with the same prompt. | 69 | 2503.17126 | title_snapshot |
6BGDGKZN7q | ReFeed: Multi-dimensional Summarization Refinement with Reflective Reasoning on Feedback | https://openreview.net/forum?id=6BGDGKZN7q | [
"Taewon Yun",
"Jihwan Oh",
"Hyangsuk Min",
"Yuho Lee",
"Jihwan Bang",
"Jason Cai",
"Hwanjun Song"
] | null | null | Summarization refinement faces challenges when extending to multi-dimension. In this paper, we introduce ReFeed, a powerful summarization refinement pipeline that enhances multiple dimensions through reflective reasoning on feedback. To achieve this, we release SumFeed-CoT, a large-scale Long-CoT-based dataset optimize... | [
"Text Refinement",
"Text Summarization",
"LLMs",
"Reasoning"
] | Reflective Reasoning for Text Refinement | 75 | 2503.21332 | title_snapshot |
oSub7DiyjL | The Devil is in the EOS: Sequence Training for Detailed Image Captioning | https://openreview.net/forum?id=oSub7DiyjL | [
"Abdelrahman Mohamed",
"Yova Kementchedjhieva"
] | null | null | Despite significant advances in vision-language models (VLMs), image captioning often suffers from a lack of detail, with base models producing short, generic captions. This limitation persists even though VLMs are equipped with strong vision and language backbones. While supervised data and complex reward functions ha... | [
"Detailed image captioning; sequence training; reinforcement learning; vision langauge models"
] | Encourging detailed image captioning through end of sequence token debaising | 78 | 2507.20077 | title_snapshot |
EP7mAqx2BO | Extragradient Preference Optimization (EGPO): Beyond Last-Iterate Convergence for Nash Learning from Human Feedback | https://openreview.net/forum?id=EP7mAqx2BO | [
"Runlong Zhou",
"Maryam Fazel",
"Simon Shaolei Du"
] | null | null | Reinforcement learning from human feedback (RLHF) has become essential for improving language model capabilities, but traditional approaches rely on the assumption that human preferences follow a transitive Bradley-Terry model.
This assumption fails to capture the non-transitive nature of populational human preferences... | [
"reinforcement learning",
"reinforcement learning from human feedback",
"large language models"
] | We provide a theoretically strong algorithm for Nash learning from human feedback as well as its equivalent practical implementation using online IPO. | 82 | 2503.08942 | title_snapshot |
EJGlOybbDB | CASCADE Your Datasets for Cross-Mode Knowledge Retrieval of Language Models | https://openreview.net/forum?id=EJGlOybbDB | [
"Runlong Zhou",
"Yi Zhang"
] | null | null | Language models often struggle with cross-mode knowledge retrieval – the ability to access knowledge learned in one format (mode) when queried in another.
We demonstrate that models trained on multiple data sources (e.g., Wikipedia and TinyStories) exhibit significantly reduced accuracy when retrieving knowledge in a f... | [
"large language models",
"pretraining",
"knowledge retrieval",
"spurious correlations"
] | We conduct qualitative and quantitative studies on the cross-mode knowledge retrieval capabilities of LLMs, with a novel approach, CASCADE, to mitigate this issue. | 83 | 2504.01450 | title_snapshot |
b8cW86QcOD | LoRI: Reducing Cross-Task Interference in Multi-Task Low-Rank Adaptation | https://openreview.net/forum?id=b8cW86QcOD | [
"Juzheng Zhang",
"Jiacheng You",
"Ashwinee Panda",
"Tom Goldstein"
] | null | null | Low-Rank Adaptation (LoRA) has emerged as a popular parameter-efficient fine-tuning (PEFT) method for Large Language Models (LLMs), yet it still incurs notable overhead and suffers from parameter interference in multi-task scenarios. We propose LoRA with Reduced Interference (LoRI), a simple yet effective approach that... | [
"Large Language Models",
"Parameter-Efficient Fine-Tuning",
"Model Merging",
"Sparsity"
] | LoRI is a simple yet effective method for parameter-efficient fine-tuning that reduces cross-task interference by freezing projection matrices $A$ and sparsifying $B$. | 86 | 2504.07448 | title_snapshot |
lEpPFmGH3L | Beyond the Reported Cutoff: Where Large Language Models Fall Short on Financial Knowledge | https://openreview.net/forum?id=lEpPFmGH3L | [
"Agam Shah",
"Liqin Ye",
"Sebastian Jaskowski",
"Wei Xu",
"Sudheer Chava"
] | null | null | Large Language Models (LLMs) are frequently utilized as sources of knowledge for question-answering. While it is known that LLMs may lack access to real-time data or newer data produced after the model's cutoff date, it is less clear how their knowledge spans across *historical* information. In this study, we assess th... | [
"Large Language Models",
"Knowledge Cutoff",
"Model Hallucinations"
] | This paper evaluates Large Language Models' knowledge of historical financial data, finding they know more about larger, recent companies but are also prone to hallucinations about these firms. | 104 | 2504.00042 | title_snapshot |
r61s1FNYlj | TRELLIS: Learning to Compress Key-Value Memory in Attention Models | https://openreview.net/forum?id=r61s1FNYlj | [
"Mahdi Karami",
"Ali Behrouz",
"Praneeth Kacham",
"Vahab Mirrokni"
] | null | null | Transformers, while powerful, suffer from quadratic computational complexity and the ever-growing Key-Value (KV) cache of the attention mechanism. This paper introduces Trellis, a novel Transformer architecture with bounded memory that learns how to compress its key-value memory dynamically at test time. Trellis repla... | [
"Sequence Models",
"Language models",
"Recurrent Neural Nets",
"Test Time Training"
] | This paper introduces a novel approach to efficiently compress the K-V cache into a fixed number of slots | 110 | 2512.23852 | title_snapshot |
h5SRsDax8v | Both Direct and Indirect Evidence Contribute to Dative Alternation Preferences in Language Models | https://openreview.net/forum?id=h5SRsDax8v | [
"Qing Yao",
"Kanishka Misra",
"Leonie Weissweiler",
"Kyle Mahowald"
] | null | null | Language models (LMs) tend to show human-like preferences on a number of syntactic phenomena, but the extent to which these are attributable to direct exposure to the phenomena or more general properties of language is unclear. We explore this with the English dative alternation (DO: "gave Y the X" vs. PO: "gave the X ... | [
"linguistics",
"dative alternation",
"indirect evidence",
"language learning",
"cognitive science",
"linguistic constructions"
] | Through manipulating word-order preferences in datives and non-datives in the training sets of language models, we find that they acquire dative alternation preferences from both direct and indirect evidence. | 115 | 2503.20850 | title_snapshot |
tK8GHR62EX | SpectR: Dynamically Composing LM Experts with Spectral Routing | https://openreview.net/forum?id=tK8GHR62EX | [
"William Fleshman",
"Benjamin Van Durme"
] | null | null | Training large, general-purpose language models poses significant challenges. The growing availability of specialized *expert* models, fine-tuned from pretrained models for specific tasks or domains, offers a promising alternative. Leveraging the potential of these existing expert models in real-world applications requ... | [
"MoE",
"routing",
"merging",
"LoRA",
"adapters",
"experts"
] | SpectR is an approach for routing and merging existing LoRA models per-token and per-layer, without any additional training or data. | 118 | 2504.03454 | title_snapshot |
vSMCBUgrQj | SimpleRL-Zoo: Investigating and Taming Zero Reinforcement Learning for Open Base Models in the Wild | https://openreview.net/forum?id=vSMCBUgrQj | [
"Weihao Zeng",
"Yuzhen Huang",
"Qian Liu",
"Wei Liu",
"Keqing He",
"Zejun MA",
"Junxian He"
] | null | null | DeepSeek-R1 has shown that long chain-of-thought (CoT) reasoning can naturally emerge through a simple reinforcement learning (RL) framework with rule-based rewards, where the training may directly start from the base models—a paradigm referred to as zero RL training. Most recent efforts to reproduce zero RL training h... | [
"Reasoning",
"Large Language Model"
] | The paper explores zero training with rule-based rewards for emergent chain-of-thought reasoning in smaller models, producing significant improvements in both reasoning accuracy and CoT length across all settings. | 124 | 2503.18892 | title_snapshot |
klPszYDIRT | SEAL: Steerable Reasoning Calibration of Large Language Models for Free | https://openreview.net/forum?id=klPszYDIRT | [
"Runjin Chen",
"Zhenyu Zhang",
"Junyuan Hong",
"Souvik Kundu",
"Zhangyang Wang"
] | null | null | Large Language Models (LLMs), such as OpenAI’s o1-series have demonstrated compelling capabilities for complex reasoning tasks via the extended chain-of-thought (CoT) reasoning mechanism. However, recent studies reveal substantial redundancy in the CoT reasoning traces, which not only increases inference latency but a... | [
"LLM",
"Reasoning",
"Representation Engineering"
] | We developed a training-free method that improves LLM accuracy and efficiency by calibrating CoT reasoning traces through a learned steering vector, reducing redundant thoughts and enhancing performance across multiple benchmarks. | 127 | 2504.07986 | title_snapshot |
3vxxB3Ar9r | One ruler to measure them all: Benchmarking multilingual long-context language models | https://openreview.net/forum?id=3vxxB3Ar9r | [
"Yekyung Kim",
"Jenna Russell",
"Marzena Karpinska",
"Mohit Iyyer"
] | null | null | We present ONERULER, a multilingual benchmark designed to evaluate long-context language models across 26 languages. ONERULER adapts the English-only RULER benchmark (Hsieh et al., 2024) by including seven synthetic tasks that test both retrieval and aggregation, including new variations of the "needle-in-a-haystack" t... | [
"Multilingual",
"Benchmark",
"Long-context",
"Synthetic dataset"
] | ONERULER is a multilingual benchmark for evaluating long-context LLMs across 26 languages, extending RULER beyond English. It aims to assess model performance in diverse linguistic settings using seven tasks, including detecting absent information. | 132 | 2503.01996 | title_snapshot |
X39dK0SX9W | Agents Are All You Need for LLM Unlearning | https://openreview.net/forum?id=X39dK0SX9W | [
"Debdeep Sanyal",
"Murari Mandal"
] | null | null | Information removal or suppression in large language models (LLMs) is a desired functionality, useful in AI regulation, legal compliance, safety, and privacy. LLM unlearning methods aim to remove information on demand from LLMs. Current LLM unlearning methods struggle to balance the unlearning efficacy and utility due ... | [
"LLM Agents",
"unlearning",
"Safety in AI"
] | LLM agents based unlearning beats all the existing unlearning methods | 135 | 2502.00406 | title_snapshot |
OgWh4J7bkT | Enhancing LLM Reasoning with Iterative DPO: A Comprehensive Empirical Investigation | https://openreview.net/forum?id=OgWh4J7bkT | [
"Songjun Tu",
"Jiahao Lin",
"Xiangyu Tian",
"Qichao Zhang",
"Linjing Li",
"Yuqian Fu",
"Nan Xu",
"Wei He",
"Xiangyuan Lan",
"Dongmei Jiang",
"Dongbin Zhao"
] | null | null | Recent advancements in post-training methodologies for large language models (LLMs) have highlighted reinforcement learning (RL) as a critical component for enhancing reasoning. However, the substantial computational costs associated with RL-based approaches have led to growing interest in alternative paradigms, such a... | [
"LLM Reasoning",
"Iterative Optimization"
] | DPO enables iterative self-improvement for LLMs, achieving RL-level reasoning performance with lower computational cost through preference-based learning and verifiable rewards. | 146 | 2503.12854 | title_snapshot |
vqN8uom4A1 | Base Models Beat Aligned Models at Randomness and Creativity | https://openreview.net/forum?id=vqN8uom4A1 | [
"Peter West",
"Christopher Potts"
] | null | null | Alignment has quickly become a default ingredient in LLM development, with techniques such as reinforcement learning from human feedback making models act safely, follow instructions, and perform ever-better on complex tasks. While these techniques are certainly useful, we propose that they should not be universally ap... | [
"alignment",
"pretrained",
"limitations",
"limits",
"capabilities",
"randomness",
"creativity"
] | Alignment seems to hurt performance on a set of tasks that require randomness or creativity | 149 | 2505.00047 | title_snapshot |
r0AXK5Cnhr | LV-Eval: A Balanced Long-Context Benchmark with 5 Length Levels Up to 256K | https://openreview.net/forum?id=r0AXK5Cnhr | [
"Tao Yuan",
"Xuefei Ning",
"Dong Zhou",
"Zhijie Yang",
"Shiyao Li",
"Minghui Zhuang",
"Zheyue Tan",
"Zhuyu Yao",
"Dahua Lin",
"Boxun Li",
"Guohao Dai",
"Shengen Yan",
"Yu Wang"
] | null | null | State-of-the-art large language models (LLMs) are now claiming remarkable supported context lengths of 256k or even more. In contrast, the average context lengths of mainstream benchmarks are insufficient (5k-21k), and they suffer from potential knowledge leakage and inaccurate metrics, resulting in biased evaluation. ... | [
"large language model",
"long-context benchmark",
"knowledge leakage mitigation"
] | LV-Eval is a long-context benchmark with 5 length levels up to 256K. It's designed to be challenging, suitable for controllable comparison, and mitigates knowledge leakage issue in evaluation. | 156 | 2402.05136 | title_snapshot |
uzauWUW9u3 | News is More than a Collection of Facts: Moral Frame Preserving News Summarization | https://openreview.net/forum?id=uzauWUW9u3 | [
"Enrico Liscio",
"Michela Lorandi",
"Pradeep K. Murukannaiah"
] | null | null | News articles are more than collections of facts; they reflect journalists' framing, shaping how events are presented to the audience. One key aspect of framing is the choice to write in (or quote verbatim) morally charged language as opposed to using neutral terms. This moral framing carries implicit judgments that au... | [
"LLMs",
"news",
"summarization",
"morality",
"framing"
] | The first investigation in how LLMs can summarize news articles while preserving moral framing. | 163 | 2504.00657 | title_snapshot |
f7GG1MbsSM | Inside-Out: Hidden Factual Knowledge in LLMs | https://openreview.net/forum?id=f7GG1MbsSM | [
"Zorik Gekhman",
"Eyal Ben-David",
"Hadas Orgad",
"Eran Ofek",
"Yonatan Belinkov",
"Idan Szpektor",
"Jonathan Herzig",
"Roi Reichart"
] | null | null | This work presents a framework for assessing whether large language models (LLMs) encode more factual knowledge in their parameters than what they express in their outputs. While a few studies hint at this possibility, none has clearly defined or demonstrated this phenomenon. We first propose a formal definition of kno... | [
"LLMs",
"Knowledge"
] | We introduce a framework for evaluating the gap between the knowledge LLMs encode internally and what they express in their outputs, and provide strong evidence of this gap across popular LLMs. | 166 | 2503.15299 | title_snapshot |
fuBrcTH8NM | Efficient Construction of Model Family through Progressive Training Using Model Expansion | https://openreview.net/forum?id=fuBrcTH8NM | [
"Kazuki Yano",
"Sho Takase",
"Sosuke Kobayashi",
"Shun Kiyono",
"Jun Suzuki"
] | null | null | As Large Language Models (LLMs) gain widespread practical applica-
tion, offering model families with varying parameter sizes has become
standard practice to accommodate diverse computational requirements.
Traditionally, each model in the family is trained independently, incurring
computational costs that scale additiv... | [
"pre-training",
"model familly",
"compute efficiency"
] | We propose a progressive training approach that efficiently builds a family of LLMs, reducing total computational requirements while achieving comparable or even better performance. | 173 | 2504.00623 | title_snapshot |
0zxugBcgF5 | Off-Policy Corrected Reward Modeling for Reinforcement Learning from Human Feedback | https://openreview.net/forum?id=0zxugBcgF5 | [
"Johannes Ackermann",
"Takashi Ishida",
"Masashi Sugiyama"
] | null | null | Reinforcement Learning from Human Feedback (RLHF) allows us to train models, such as language models (LMs), to follow complex human preferences. In RLHF for LMs, we first train an LM using supervised fine-tuning, sample pairs of responses, obtain human feedback, and use the resulting data to train a reward model (RM). ... | [
"alignment",
"reinforcement learning from human feedback",
"reinforcement learning",
"reward modeling"
] | We propose to use importance weighting to iteratively retrain an off-policy corrected reward model, resulting in a signficantly better final policy. | 174 | 2507.15507 | title_snapshot |
nSV8Depcpx | Plancraft: an evaluation dataset for planning with LLM agents | https://openreview.net/forum?id=nSV8Depcpx | [
"Gautier Dagan",
"Frank Keller",
"Alex Lascarides"
] | null | null | We present Plancraft, a multi-modal evaluation dataset for LLM agents. Plancraft has both a text-only and multi-modal interface, based on the Minecraft crafting GUI. We include the Minecraft Wiki to evaluate tool use and Retrieval Augmented Generation (RAG), as well as a handcrafted planner and Oracle Retriever, to abl... | [
"planning",
"multi-modal",
"agents",
"LLMs",
"tool use",
"minecraft",
"RAG"
] | We introduce Plancraft, a multi-modal evaluation dataset for LLMs, designed to assess tool use, planning, and decision-making with both solvable and unsolvable examples. | 176 | 2412.21033 | title_snapshot |
5wAfbEs34A | Style over Substance: Distilled Language Models Reason Via Stylistic Replication | https://openreview.net/forum?id=5wAfbEs34A | [
"Philip Lippmann",
"Jie Yang"
] | null | null | Specialized reasoning language models (RLMs) have demonstrated that scaling test-time computation through detailed reasoning traces significantly enhances performance.Although these traces effectively facilitate knowledge distillation into smaller, instruction-tuned models, the precise nature of transferred reasoning r... | [
"Reasoning",
"language models",
"stylistic mimicry",
"pivots",
"synthetic data",
"distillation",
"finetuning",
"metacognition"
] | We show that language models distilled from reasoning models primarily mimic stylistic patterns rather than internalize deeper reasoning capabilities. | 183 | 2504.01738 | title_snapshot |
jdOC24msVq | EuroBERT: Scaling Multilingual Encoders for European Languages | https://openreview.net/forum?id=jdOC24msVq | [
"Nicolas Boizard",
"Hippolyte Gisserot-Boukhlef",
"Duarte Miguel Alves",
"Andre Martins",
"Ayoub Hammal",
"Caio Corro",
"CELINE HUDELOT",
"Emmanuel Malherbe",
"Etienne Malaboeuf",
"Fanny Jourdan",
"Gabriel Hautreux",
"João Alves",
"Kevin El Haddad",
"Manuel Faysse",
"Maxime Peyrard",
"... | null | null | General-purpose multilingual vector representations, used in retrieval, regression and classification, are traditionally obtained from bidirectional encoder models. Despite their wide applicability, encoders have been recently overshadowed by advances in generative decoder-only models. However, many innovations driving... | [
"Encoder",
"Multilingual",
"EuroBERT",
"Training",
"Vector Representations",
"Bidirectional",
"European"
] | We introduce EuroBERT, a family of multilingual encoders leveraging recent architectural advances, achieving state-of-the-art performance across diverse tasks with support for sequences up to 8,192 tokens. | 185 | 2503.05500 | title_snapshot |
ghyyHZYORi | Training Plug-and-Play Knowledge Modules with Deep Context Distillation | https://openreview.net/forum?id=ghyyHZYORi | [
"Lucas Caccia",
"Alan Ansell",
"Edoardo Ponti",
"Ivan Vulić",
"Alessandro Sordoni"
] | null | null | Dynamically integrating new or rapidly evolving information after (Large) Language Model pre-training remains challenging, particularly in low-data scenarios or when dealing with private and specialized documents. In-context learning and retrieval-augmented generation (RAG) face limitations, including their high infere... | [
"knowledge extraction",
"document understanding",
"modular learning"
] | We encapsulate knowledge from a document inside a LoRA adapter via distillation | 187 | 2503.08727 | title_judge |
sX4OoLKSW2 | Supposedly Equivalent Facts That Aren’t? Entity Frequency in Pre-training Induces Asymmetry in LLMs | https://openreview.net/forum?id=sX4OoLKSW2 | [
"Yuan He",
"Bailan He",
"Zifeng Ding",
"Alisia Maria Lupidi",
"Yuqicheng Zhu",
"Shuo Chen",
"Caiqi Zhang",
"Jiaoyan Chen",
"Yunpu Ma",
"Volker Tresp",
"Ian Horrocks"
] | null | null | Understanding and mitigating hallucinations in Large Language Models (LLMs) is crucial for ensuring reliable content generation. While previous research has primarily focused on "when" LLMs hallucinate, our work explains "why" and directly links model behaviour to the pre-training data that forms their prior knowledge.... | [
"Large Language Models",
"Asymmetry",
"Equivalent Facts",
"Entity Frequency",
"Pre-training Bias",
"Knowledge Probing",
"Hallucinations",
"Knowledge Graphs"
] | This work demonstrates that the asymmetry in how large language models recognise equivalent facts stems from inherent biases in their pre-training data, particularly through differences in entity frequency. | 192 | 2503.22362 | title_snapshot |
JMxRn7orEk | CUPID: Evaluating Personalized and Contextualized Alignment of LLMs from Interactions | https://openreview.net/forum?id=JMxRn7orEk | [
"Tae Soo Kim",
"Yoonjoo Lee",
"Yoonah Park",
"Jiho Kim",
"Young-Ho Kim",
"Juho Kim"
] | null | null | Personalization of Large Language Models (LLMs) often assumes users hold static preferences that reflect globally in all tasks. In reality, humans hold dynamic preferences that change depending on the context. As users interact with an LLM in various contexts, they naturally reveal their contextual preferences, which a... | [
"LLM",
"Evaluation",
"Benchmark",
"Personalization",
"Preferences",
"Interactions"
] | We introduce CUPID, a benchmark that evaluates LLMs capability to infer a user's contextual preferences from prior user-LLM interactions. | 196 | 2508.01674 | title_snapshot |
0JzWiigkUy | BEARCUBS: A benchmark for computer-using web agents | https://openreview.net/forum?id=0JzWiigkUy | [
"Yixiao Song",
"Katherine Thai",
"Chau Minh Pham",
"Yapei Chang",
"Mazin Nadaf",
"Mohit Iyyer"
] | null | null | Modern web agents possess computer use abilities that allow them to interact with webpages by sending commands to a virtual keyboard and mouse. While such agents have considerable potential to assist human users with complex tasks, evaluating their capabilities in real-world settings poses a major challenge. To this en... | [
"computer-use agent",
"benchmark",
"multimodal"
] | We introduce BEARCUBS, a benchmark of 111 information-seeking questions designed to evaluate a web agent’s ability to search, browse, and identify factual information from the web. | 200 | 2503.07919 | title_snapshot |
vDr0RV3590 | Do Biased Models Have Biased Thoughts? | https://openreview.net/forum?id=vDr0RV3590 | [
"Swati Rajwal",
"Shivank Garg",
"Reem Abdel-Salam",
"Abdelrahman Zayed"
] | null | null | The impressive performance of language models is undeniable. However, the presence of biases based on gender, race, socio-economic status, physical appearance, and sexual orientation makes the deployment of language models challenging. This paper studies the effect of chain-of-thought prompting, a recent approach that ... | [
"Bias in language models",
"Large Language Models",
"biased thoughts",
"Chain-of-Thought prompting"
] | This paper explores whether biased language models have biased reasoning, finding that their thought processes are not strongly linked to biased outputs. | 203 | 2508.06671 | title_snapshot |
ED5diyzc1C | LLM-based Multi-Agents System Attack via Continuous Optimization with Discrete Efficient Search | https://openreview.net/forum?id=ED5diyzc1C | [
"Weichen Yu",
"Kai Hu",
"Tianyu Pang",
"Chao Du",
"Min Lin",
"Matt Fredrikson"
] | null | null | Large Language Model (LLM)-based Multi-Agent Systems (MAS) have demonstrated remarkable capability in complex tasks. However, emerging evidence indicates significant security vulnerabilities within these systems.
In this paper, we introduce three novel and practical attack scenarios that allow only a single interventi... | [
"multi-agent system",
"adversarial attack",
"LLM-based jailbreak"
] | Attack an LLM-based multi agent system with only one intevention, we propose a token-based optimization method | 204 | null | null |
sy71y74U80 | D3: A Dataset for Training Code LMs to Act Diff-by-Diff | https://openreview.net/forum?id=sy71y74U80 | [
"Ulyana Piterbarg",
"Kanishk Gandhi",
"Lerrel Pinto",
"Noah Goodman",
"Rob Fergus"
] | null | null | We introduce D3 ("Diverse Data for Diff-by-Diff Coding"), a large dataset for training LMs to iteratively synthesize general-purpose Python source code by generating file diffs. D3 frames code synthesis as a goal-conditioned sequential decision-making problem, where goals, states, and actions are represented by token s... | [
"data filtering",
"synthetic data",
"code synthesis",
"code editing",
"file diffs",
"midtraining",
"SFT",
"LM agents"
] | D3 is a dataset of 8 billion tokens of file-diff-sequence examples sampled from 850k Human-written source files, improving LM performance on code synthesis, completion, & editing. | 205 | null | null |
aJDykpJAYF | Shared Global and Local Geometry of Language Model Embeddings | https://openreview.net/forum?id=aJDykpJAYF | [
"Andrew Lee",
"Melanie Weber",
"Fernanda Viégas",
"Martin Wattenberg"
] | null | null | Researchers have recently suggested that models share common representations. In our work, we find numerous geometric similarities across the token embeddings of large language models. First, we find “global” similarities: token embeddings often share similar relative orientations. Next, we characterize local geometry ... | [
"Embeddings",
"Alignment",
"Interpretability"
] | We characterize the global and local geometry of language model token embeddings and find similarities across language models. | 209 | 2503.21073 | title_snapshot |
EFxC34XbDh | $100K or 100 Days: Trade-offs when Pre-Training with Academic Resources | https://openreview.net/forum?id=EFxC34XbDh | [
"Apoorv Khandelwal",
"Tian Yun",
"Nihal V. Nayak",
"Jack Merullo",
"Stephen Bach",
"Chen Sun",
"Ellie Pavlick"
] | null | null | Pre-training is notoriously compute-intensive and academic researchers are notoriously under-resourced. It is, therefore, commonly assumed that academics can't pre-train models. In this paper, we seek to clarify this assumption. We first survey academic researchers to learn about their available compute and then empiri... | [
"pre-training",
"training efficiency",
"benchmarking",
"hardware",
"GPUs"
] | We present insights about pre-training on academic compute and a software benchmark to determine the most efficient training settings. | 210 | 2410.23261 | title_snapshot |
9AFIz0YzD7 | Gating is Weighting: Understanding Gated Linear Attention through In-context Learning | https://openreview.net/forum?id=9AFIz0YzD7 | [
"Yingcong Li",
"Davoud Ataee Tarzanagh",
"Ankit Singh Rawat",
"Maryam Fazel",
"Samet Oymak"
] | null | null | Linear attention methods offer a compelling alternative to softmax attention due to their efficiency in recurrent decoding. Recent research has focused on enhancing standard linear attention by incorporating gating while retaining its computational benefits. Such Gated Linear Attention (GLA) architectures include highl... | [
"linear attention",
"gating",
"in-context learning",
"weighted gradient descent",
"optimization landscape"
] | This work offers theoretical insights into the ICL capabilities of gated linear attention models, demonstrates how gating is crucial for achieving stronger data adaptivity, and characterizes the loss landscape of weighted preconditioned GD. | 211 | 2504.04308 | title_snapshot |
lqC5J7pBP9 | C3PO: Critical-Layer, Core-Expert, Collaborative Pathway Optimization for Test-Time Expert Re-Mixing | https://openreview.net/forum?id=lqC5J7pBP9 | [
"Zhongyang Li",
"Ziyue Li",
"Tianyi Zhou"
] | null | null | Mixture-of-Experts (MoE) Large Language Models (LLMs) suffer from severely sub-optimal expert pathways—our study reveals that naive expert selection learned from pretraining leaves a surprising 10-20% accuracy gap for improvement. Motivated by this observation, we develop a novel class of test-time optimization methods... | [
"Mixture-of-Experts",
"Large Language Models",
"Test-Time Learning"
] | Optimizing MoE LLM pathways at test-time via reference-based expert re-weighting, boosting accuracy by 7-15%. | 223 | 2504.07964 | title_snapshot |
Q5pVZCrrKr | CodeARC: Benchmarking Reasoning Capabilities of LLM Agents for Inductive Program Synthesis | https://openreview.net/forum?id=Q5pVZCrrKr | [
"Anjiang Wei",
"Tarun Suresh",
"Jiannan Cao",
"Naveen Kannan",
"Yuheng Wu",
"Kai Yan",
"Thiago S. F. X. Teixeira",
"Ke Wang",
"Alex Aiken"
] | null | null | Inductive program synthesis, or programming by example, requires synthesizing functions from input-output examples that generalize to unseen inputs. While large language model agents have shown promise in programming tasks guided by natural language, their ability to perform inductive program synthesis is underexplored... | [
"Agent",
"Large Language Model",
"Reasoning",
"Code",
"Program Synthesis"
] | We introduce CodeARC, a benchmark for inductive program synthesis where LLM agents iteratively refine code via oracle feedback, enabling more realistic and challenging evaluation for inductive reasoning. | 225 | 2503.23145 | title_snapshot |
WnZjdQOWiY | Robo-Instruct: Simulator-Augmented Instruction Alignment For Finetuning Code LLMs | https://openreview.net/forum?id=WnZjdQOWiY | [
"Zichao Hu",
"Junyi Jessy Li",
"Arjun Guha",
"Joydeep Biswas"
] | null | null | Code LLMs have shown promising results with converting tasks in natural language to programs that can be executed by service robots. We are interested in finetuning small, specialized LLMs for this purpose, but collecting datasets of task-program pairs specific to each robot is time-consuming and expensive. While appro... | [
"Self-Instruct",
"Fine-tuning Code LLMs for service robot tasks",
"Domain-Specific Program Generation",
"Code LLMs for Robotics"
] | We propose a simulator-augmented approach for generating synthetic training data to fine-tune code LLMs on domain-specific robot tasks. | 226 | 2405.20179 | title_snapshot |
7evvwwdo3z | R2E-Gym: Procedural Environment Generation and Hybrid Verifiers for Scaling Open-Weights SWE Agents | https://openreview.net/forum?id=7evvwwdo3z | [
"Naman Jain",
"Jaskirat Singh",
"Manish Shetty",
"Tianjun Zhang",
"Liang Zheng",
"Koushik Sen",
"Ion Stoica"
] | null | null | Improving open-source models on real-world SWE tasks (solving GITHUB issues) faces two key challenges: 1) scalable curation of execution environments to train these models, and, 2) optimal scaling of test-time compute. We introduce R2EGym, the largest procedurally-curated executable gym environment for training real-wo... | [
"SWE Agents ; Inference Time Scaling ; Test Generation"
] | R2E-Gym: Procedural Environment Generation for Scaling Open-Weights Software Engineering Agents | 231 | 2504.07164 | title_judge |
R7qRUFHGTx | When To Solve, When To Verify: Compute-Optimal Problem Solving and Generative Verification for LLM Reasoning | https://openreview.net/forum?id=R7qRUFHGTx | [
"Nishad Singhi",
"Hritik Bansal",
"Arian Hosseini",
"Aditya Grover",
"Kai-Wei Chang",
"Marcus Rohrbach",
"Anna Rohrbach"
] | null | null | Scaling test-time compute has emerged as a key strategy for enhancing the reasoning capabilities of large language models (LLMs), particularly in tasks like mathematical problem-solving. A traditional approach, Self-Consistency (SC), generates multiple solutions to a problem and selects the most common answer via major... | [
"test-time scaling",
"self-consistency",
"generative reward models",
"compute-matched analysis"
] | We perform test-time compute matched comparison between scaling solutions via self-consistency and verifications via GenRM to find useful insights for the practitioners. | 232 | 2504.01005 | title_snapshot |
hLg2rzBJR2 | Exploring the Limit of Outcome Reward for Learning Mathematical Reasoning | https://openreview.net/forum?id=hLg2rzBJR2 | [
"Chengqi Lyu",
"Songyang Gao",
"Yuzhe Gu",
"Wenwei Zhang",
"Jianfei Gao",
"Kuikun Liu",
"Ziyi Wang",
"Shuaibin Li",
"Qian Zhao",
"Haian Huang",
"Weihan Cao",
"Jiangning Liu",
"Hongwei Liu",
"Junnan Liu",
"Songyang Zhang",
"Dahua Lin",
"Kai Chen"
] | null | null | Reasoning abilities, especially those for solving complex math problems, are crucial components of general intelligence. Recent advances by proprietary companies, such as o-series models of OpenAI, have made remarkable progress on reasoning tasks. However, the complete technical details remain unrevealed, and the techn... | [
"Large Language Model",
"Reinforcement Learning",
"Mathmatical Reasoning"
] | This paper proposes a new RL framework to pursue the performance limit that can be achieved through outcome reward-based reinforcement learning for mathematical reasoning tasks, where only binary outcome rewards are easily accessible. | 234 | 2502.06781 | title_snapshot |
lvQwn8eiRf | How does Watermarking Affect Visual Language Models in Document Understanding? | https://openreview.net/forum?id=lvQwn8eiRf | [
"Chunxue Xu",
"Yiwei Wang",
"Bryan Hooi",
"Yujun Cai",
"Songze Li"
] | null | null | Visual Language Models (VLMs) have become foundational models for document understanding tasks, widely used in the processing of complex multimodal documents across domains such as finance, law, and academia. However, documents often contain noise-like information, such as watermarks, which inevitably leads us to inqui... | [
"VLMs",
"Document Understanding",
"Robustness"
] | Investigation of the robustness of VLM for document understanding task | 236 | 2504.01048 | title_snapshot |
p4wZfBFgyI | Understanding the Uncertainty of LLM Explanations: A Perspective Based on Reasoning Topology | https://openreview.net/forum?id=p4wZfBFgyI | [
"Longchao Da",
"Xiaoou Liu",
"Jiaxin Dai",
"Lu Cheng",
"Yaqing Wang",
"Hua Wei"
] | null | null | Understanding the uncertainty in large language model (LLM) explanations is important for evaluating their faithfulness and reasoning consistency, thus providing insights into the reliability of LLM's output. In this work, we propose a novel framework that quantifies uncertainty in LLM explanations through a formal rea... | [
"Uncertainty Quantification",
"LLM Explanations",
"Graph Mining"
] | A framework quantifies uncertainty in LLM explanations through a formal reasoning topology perspective. | 238 | 2502.17026 | title_snapshot |
n3rZJrWPLE | Mixture of Attention Spans: Optimizing LLM Inference Efficiency with Heterogeneous Sliding-Window Lengths | https://openreview.net/forum?id=n3rZJrWPLE | [
"Tianyu Fu",
"Haofeng Huang",
"Xuefei Ning",
"Genghan Zhang",
"Boju Chen",
"Tianqi Wu",
"Hongyi Wang",
"Zixiao Huang",
"Shiyao Li",
"Shengen Yan",
"Guohao Dai",
"Huazhong Yang",
"Yu Wang"
] | null | null | Sliding-window attention offers a hardware-efficient solution to the memory and throughput challenges of Large Language Models (LLMs) in long-context scenarios. Existing methods typically employ a single window length across all attention heads and input sizes. However, this uniform approach fails to capture the hetero... | [
"Efficient Attention",
"Sparse Attention",
"KV Cache Management",
"Large Language Models",
"Efficiency"
] | We design heterogeneous elastic rules for sliding-window lengths of attention for efficient large language models | 247 | 2406.14909 | title_snapshot |
d9EkgbZZH9 | You Cannot Feed Two Birds with One Score: the Accuracy-Naturalness Tradeoff in Translation | https://openreview.net/forum?id=d9EkgbZZH9 | [
"Gergely Flamich",
"David Vilar",
"Jan-Thorsten Peter",
"Markus Freitag"
] | null | null | The goal of translation, be it by human or by machine, is, given some text in a source language, to produce text in a target language that simultaneously 1) preserves the meaning of the source text and 2) achieves natural expression in the target language. However, researchers in the machine translation community usual... | [
"translation",
"accuracy",
"naturalness",
"tradeoff",
"distortion",
"perception",
"no-reference metric"
] | We prove mathematically and demonstrate empirically that optimizing a single metric for machine translation *cannot* lead to a system that is both accurate and fluent. We also establish a connection between no-reference metrics and our theory. | 251 | 2503.24013 | title_snapshot |
ZYVAtUUNbH | Imagine All The Relevance: Scenario-Profiled Indexing with Knowledge Expansion for Dense Retrieval | https://openreview.net/forum?id=ZYVAtUUNbH | [
"Sangam Lee",
"Ryang Heo",
"SeongKu Kang",
"Dongha Lee"
] | null | null | Existing dense retrieval models struggle with reasoning-intensive retrieval task as they fail to capture implicit relevance that requires reasoning beyond surface-level semantic information.
To address these challenges, we propose Scenario-Profiled Indexing with Knowledge Expansion (SPIKE), a dense retrieval framework ... | [
"Information Retrieval",
"Reasoning Intensive Retrieval",
"Dense Retrieval",
"Reasoning",
"LLM"
] | We propose SPIKE, a dense retrieval framework that decomposes documents into scenarios. | 257 | 2503.23033 | title_snapshot |
63c7hTrUCh | Ensemble Debiasing Across Class and Sample Levels for Fairer Prompting Accuracy | https://openreview.net/forum?id=63c7hTrUCh | [
"Ruixi Lin",
"Ziqiao Wang",
"Yang You"
] | null | null | Language models are strong few-shot learners and achieve good overall accuracy in text classification tasks, masking the fact that their results suffer from great class accuracy imbalance. We believe that the pursuit of overall accuracy should not come from enriching the strong classes, but from raising up the weak one... | [
"ensemble debiasing",
"accuracy imbalance",
"Heaviside step function",
"post-hoc correction"
] | We propose a Heaviside step function based ensemble debiasing method, to flexibly rectify biased ICL output class probabilities across both class and sample levels, achieving fairer prompting accuracy for LLMs. | 260 | 2503.05157 | title_snapshot |
ayB1PACN5j | RWKV-7 "Goose" with Expressive Dynamic State Evolution | https://openreview.net/forum?id=ayB1PACN5j | [
"Bo Peng",
"Ruichong Zhang",
"Daniel Goldstein",
"Eric Alcaide",
"Xingjian Du",
"Haowen Hou",
"Jiaju Lin",
"Jiaxing Liu",
"Janna Lu",
"William Merrill",
"Guangyu Song",
"Kaifeng Tan",
"Saiteja Utpala",
"Nathan Wilce",
"Johan S. Wind",
"Tianyi Wu",
"Daniel Wuttke",
"Christian Zhou-Z... | null | null | We present RWKV-7 "Goose", a new sequence modeling architecture with constant memory usage and constant inference time per token. Despite being trained on dramatically fewer tokens than other top models, our 2.9 billion parameter language model achieves a new 3B SoTA on multilingual tasks and matches the current 3B SoT... | [
"Goose",
"LLM",
"RWKV",
"RWKV-7",
"RWKV7",
"Linear",
"Linear Attention",
"SSM",
"subquadratic"
] | RWKV-7 is a new sequence modeling architecture with constant memory usage and inference time per token, SoTA performance on multilingual tasks, and near SoTA English LLM performance at 3B scale, with dramatically less training than top 3B models. | 266 | 2503.14456 | title_snapshot |
dU4Y2sNfJ2 | Cutting the Root of Hallucination: Structural Trimming for Vulnerability Mitigation in Code LLMs | https://openreview.net/forum?id=dU4Y2sNfJ2 | [
"Yage Zhang"
] | null | null | We introduce a structural perspective on hallucinations in code-generating language models, framing them as causality anchors in syntax graphs that trigger cascading semantic errors and latent security flaws. This work is the first to systematically connect code hallucinations with vulnerability risks, offering a unifi... | [
"LLM hallucinations",
"code generation",
"program repair",
"vulnerability mitigation",
"structural pruning",
"abstract syntax tree",
"hallucination detection",
"CSHS",
"model-agnostic risk estimation",
"generative code safety"
] | LLMs hallucinate code often with security risks. We trace these structurally, prune them surgically, and predict repair effectiveness. Our method patches code and mitigates risk using a transferable score (CSHS). | 269 | null | null |
TiRiDMkTmG | Out-of-Distribution Detection using Synthetic Data Generation | https://openreview.net/forum?id=TiRiDMkTmG | [
"Momin Abbas",
"Muneeza Azmat",
"Raya Horesh",
"Mikhail Yurochkin"
] | null | null | Distinguishing in- and out-of-distribution (OOD) inputs is crucial for reliable deployment of classification systems. However, OOD data is typically unavailable or difficult to collect, posing a significant challenge for accurate OOD detection. In this work, we present a method that harnesses the generative capabilitie... | [
"out-of-distribution detection",
"out-of-distribution generalization",
"synthetic data"
] | This work presents an effective OOD detection method using LLM-generated synthetic proxies, eliminating the need for external OOD data. Experiments show it reduces false positives and outperforms baseline methods in text classification tasks. | 270 | 2502.03323 | title_snapshot |
2H85485yAb | Truth-value judgment in language models: ‘truth directions’ are context sensitive | https://openreview.net/forum?id=2H85485yAb | [
"Stefan F. Schouten",
"Peter Bloem",
"Ilia Markov",
"Piek Vossen"
] | null | null | Recent work has demonstrated that the latent spaces of large language models (LLMs) contain directions predictive of the truth of sentences. Multiple methods recover such directions and build probes that are described as uncovering a model’s “knowledge” or “beliefs”. We investigate this phenomenon, looking closely at t... | [
"mechinterp",
"mechanistic interpretability",
"interpretability",
"truth directions",
"LLM beliefs",
"large language model",
"llm"
] | Investigation of the in-context behaviour of LLM 'truth directions'. | 271 | 2404.18865 | title_snapshot |
IC2WwhUfQg | Short-PHD: Detecting Short LLM-generated Text with Topological Data Analysis After Off-topic Content Insertion | https://openreview.net/forum?id=IC2WwhUfQg | [
"Dongjun Wei",
"Minjia Mao",
"Xiao Fang",
"Michael Chau"
] | null | null | The malicious usage of large language models (LLMs) has motivated the detection of LLM-generated texts. Previous work in topological data analysis shows that the persistent homology dimension (PHD) of text embeddings can serve as a more robust and promising score than other zero-shot methods. However, effectively detec... | [
"large language model",
"zero-shot detection",
"short text",
"topological data analysis"
] | We present Short-PHD, a zero-shot LLM-generated text detection method tailored for short texts. | 273 | 2504.02873 | title_snapshot |
r8nloXtluk | ScholarCopilot: Training Large Language Models for Academic Writing with Accurate Citations | https://openreview.net/forum?id=r8nloXtluk | [
"Yubo Wang",
"Xueguang Ma",
"Ping Nie",
"Huaye Zeng",
"Zhiheng Lyu",
"Yuxuan Zhang",
"Benjamin Schneider",
"Yi Lu",
"Xiang Yue",
"Wenhu Chen"
] | null | null | Academic writing requires both coherent text generation and precise citation of relevant literature. Although recent Retrieval-Augmented Generation (RAG) systems have significantly improved factual accuracy in general-purpose text generation, their ability to support professional academic writing remains limited. In th... | [
"RAG"
] | ScholarCopilot is a language model that combines text generation and citation retrieval for academic writing | 283 | 2504.00824 | title_snapshot |
ruWC5LIMSo | LongProc: Benchmarking Long-Context Language Models on Long Procedural Generation | https://openreview.net/forum?id=ruWC5LIMSo | [
"Xi Ye",
"Fangcong Yin",
"Yinghui He",
"Joie Zhang",
"Howard Yen",
"Tianyu Gao",
"Greg Durrett",
"Danqi Chen"
] | null | null | Existing benchmarks for evaluating long-context language models (LCLMs) primarily focus on long-context recall, requiring models to produce short responses based on a few critical snippets while processing thousands of irrelevant tokens.
We introduce LongProc (Long Procedural Generation), a new benchmark that requires ... | [
"Keywords: Large language models",
"long-context",
"natural language processing"
] | We introduce LongProc (Long Procedural Generation), a new benchmark that requires both the integration of highly dispersed information and long-form generation. | 286 | 2501.05414 | title_snapshot |
vlyl9xZVAL | Improving Table Understanding with LLMs and Entity-Oriented Search | https://openreview.net/forum?id=vlyl9xZVAL | [
"Thi-Nhung Nguyen",
"Hoang Ngo",
"Dinh Phung",
"Thuy-Trang Vu",
"Dat Quoc Nguyen"
] | null | null | Our work addresses the challenges of understanding tables. Existing methods often struggle with the unpredictable nature of table content, leading to a reliance on preprocessing and keyword matching. They also face limitations due to the lack of contextual information, which complicates the reasoning processes of large... | [
"table understanding",
"llm"
] | We introduce an entity-oriented search method to enhance table understanding in LLMs, reducing preprocessing and achieving state-of-the-art results. | 287 | 2508.17028 | title_snapshot |
X5vFauyVWr | DualEdit: Dual Editing for Knowledge Updating in Vision-Language Models | https://openreview.net/forum?id=X5vFauyVWr | [
"Zhiyi Shi",
"Binjie Wang",
"Chongjie Si",
"Yichen Wu",
"Junsik Kim",
"Hanspeter Pfister"
] | null | null | Model editing aims to efficiently update a pre-trained model’s knowledge without the need for time-consuming full retraining. While existing pioneering editing methods achieve promising results, they primarily focus on editing single-modal language models (LLMs). However, for vision-language models (VLMs), which involv... | [
"Model Editing",
"Multimodal Learning",
"VLM"
] | We explore the importance of image and text modalities and propose a novel dual editing method—DualEdit. | 293 | 2506.13638 | title_snapshot |
vBcGnragkr | How do language models learn facts? Dynamics, curricula and hallucinations | https://openreview.net/forum?id=vBcGnragkr | [
"Nicolas Zucchet",
"Jorg Bornschein",
"Stephanie C.Y. Chan",
"Andrew Kyle Lampinen",
"Razvan Pascanu",
"Soham De"
] | null | null | Large language models accumulate vast amounts of knowledge during their pre-training, yet the dynamics governing this acquisition remain poorly understood. This work investigates the learning dynamics of language models on a synthetic factual recall task, uncovering three key findings: First, language models learn in t... | [
"learning dynamics",
"factual recall",
"curricula",
"data distribution",
"hallucinations"
] | We analyze learning dynamics of language models on a synthetic memory task and show that they learn sequentially, that some data distribution properties lead to faster learning, and that hallucinations appear simulataneously to knowledge acquisition. | 294 | 2503.21676 | title_snapshot |
OKvSnV5Ar7 | Limitations of refinement methods for weak to strong generalization | https://openreview.net/forum?id=OKvSnV5Ar7 | [
"Seamus Somerstep",
"Yaacov Ritov",
"Mikhail Yurochkin",
"Subha Maity",
"Yuekai Sun"
] | null | null | Standard techniques for aligning large language models (LLMs) utilize human-produced data, which could limit the capability of any aligned LLM to human level. Label refinement and weak training have emerged as promising strategies to address this *superalignment* problem. In this work, we adopt probabilistic assumption... | [
"Weak to strong generalization",
"superalignment",
"transfer learning"
] | We study label refinement methods for weak to strong generalization. | 299 | 2508.17018 | title_snapshot |
ryTr83DxRq | MLGym: A New Framework and Benchmark for Advancing AI Research Agents | https://openreview.net/forum?id=ryTr83DxRq | [
"Deepak Nathani",
"Lovish Madaan",
"Nicholas Roberts",
"Nikolay Bashlykov",
"Ajay Menon",
"Vincent Moens",
"Mikhail Plekhanov",
"Amar Budhiraja",
"Despoina Magka",
"Vladislav Vorotilov",
"Gaurav Chaurasia",
"Dieuwke Hupkes",
"Ricardo Silveira Cabral",
"Tatiana Shavrina",
"Jakob Nicolaus ... | null | null | We introduce MLGym and MLGym-Bench, a new framework and benchmark for evaluating and developing LLM agents on AI research tasks. This is the first Gym environment for machine learning (ML) tasks, enabling research on reinforcement learning (RL) algorithms for training such agents. MLGym-bench consists of 13 diverse and... | [
"LLM Agents",
"Tool Use",
"Benchmark",
"AI Research Agents"
] | MLGym introduces a framework and benchmark suite for evaluating and developing large language model agents on diverse AI research tasks. | 302 | 2502.14499 | title_snapshot |
gOKTe1KI8K | StagFormer: Time Staggering Decoder only Transformers | https://openreview.net/forum?id=gOKTe1KI8K | [
"Dylan J Cutler",
"Arun Kandoor",
"Nishanth Dikkala",
"Nikunj Saunshi",
"Xin Wang",
"Rina Panigrahy"
] | null | null | Standard decoding in a Transformer based language model is inherently sequential as we wait for a token’s embedding to pass through all the layers in the network before starting the generation of the next token. In this work, we propose anew architecture StagFormer (Staggered Transformer), which staggered execution alo... | [
"staggered execution",
"decoder only language models",
"efficiency",
"novel architectures",
"generative models"
] | We propose a novel variant of the Transformer architecture for decoder only language modeling where we break the causal flow of information along the layers of a model by staggering in the time axis. | 305 | null | null |
CYiXNIQegF | Correctness-Guaranteed Code Generation via Constrained Decoding | https://openreview.net/forum?id=CYiXNIQegF | [
"Lingxiao Li",
"salar rahili",
"Yiwei Zhao"
] | null | null | Language Models (LMs) are increasingly being used for code generation, but ensuring the correctness of generated programs remains a significant challenge. Although imperfect code may be acceptable during software development with human oversight, domains such as video games and robotics require one-shot correctness for... | [
"code generation",
"constrained decoding",
"correctness",
"llm"
] | We present a constrained decoding algorithm that uses context-sensitive parsing with non-extensible regular expressions to generate semantically correct programs that can be extended to runtime guarantees | 306 | 2508.15866 | title_snapshot |
oGO0fNVWrN | Plato: Plan to Efficient Decode for Large Language Model Inference | https://openreview.net/forum?id=oGO0fNVWrN | [
"Shuowei Jin",
"Xueshen Liu",
"Yongji Wu",
"Haizhong Zheng",
"Qingzhao Zhang",
"Atul Prakash",
"Matthew Lentz",
"Danyang Zhuo",
"Feng Qian",
"Zhuoqing Mao"
] | null | null | Large language models (LLMs) have achieved remarkable success in natural language tasks, but their inference incurs substantial computational and memory overhead.
To improve efficiency, parallel decoding methods like Skeleton-of-Thought (SoT) decompose prompts into sub-problems for concurrent processing. However, these... | [
"Efficient LLM Inference"
] | Plan to exploit parallelism structure to break the autoregressive nature of LLM inference. | 308 | 2402.12280 | title_judge |
M7cl4Ldw61 | Can Language Models Falsify? Evaluating Algorithmic Reasoning with Counterexample Creation | https://openreview.net/forum?id=M7cl4Ldw61 | [
"Shiven Sinha",
"Shashwat Goel",
"Ponnurangam Kumaraguru",
"Jonas Geiping",
"Matthias Bethge",
"Ameya Prabhu"
] | null | null | There is growing excitement about the potential of Language Models (LMs) to accelerate scientific discovery. *Falsifying* hypotheses is key to scientific progress, as it allows claims to be iteratively refined over time. This process requires significant researcher effort, reasoning, and ingenuity. Yet current benchmar... | [
"code; self-repair; falsification"
] | We test the abilities of models to find counterexamples automatically using code-execution, and this can be hard for reasoning models. | 309 | 2502.19414 | title_snapshot |
Pg0PAvbhGv | Rank1: Test-Time Compute for Reranking in Information Retrieval | https://openreview.net/forum?id=Pg0PAvbhGv | [
"Orion Weller",
"Kathryn Ricci",
"Eugene Yang",
"Andrew Yates",
"Dawn Lawrie",
"Benjamin Van Durme"
] | null | null | We introduce Rank1, the first reranking model trained to take advantage of test-time compute. Rank1 demonstrates the applicability within retrieval of using a reasoning language model (i.e. OpenAI's o1, Deepseek's R1, etc.) for distillation in order to rapidly improve the performance of a smaller model. We gather and o... | [
"retrieval",
"reranking",
"test-time compute"
] | We train the first reranker using test-time compute in information retrieval | 312 | 2502.18418 | title_snapshot |
B5E3ijlLML | Exposing and Patching the Flaws of Large Language Models in Social Character Simulation | https://openreview.net/forum?id=B5E3ijlLML | [
"Yue Huang",
"Zhengqing Yuan",
"Yujun Zhou",
"Kehan Guo",
"Xiangqi Wang",
"Haomin Zhuang",
"Weixiang Sun",
"Lichao Sun",
"Jindong Wang",
"Yanfang Ye",
"Xiangliang Zhang"
] | null | null | Large Language Models (LLMs) are increasingly used for social character simulations, enabling applications in role-playing agents and Computational Social Science (CSS). However, their inherent flaws—such as inconsistencies in simulated roles—raise concerns about their reliability and trustworthiness. In this paper, we... | [
"Social simulation",
"Large language model",
"reliability"
] | This paper aims to unveil and improve the reliability of LLMs in social simulation scenarios. | 314 | null | null |
u9JXu4L17I | DeepRetrieval: Hacking Real Search Engines and Retrievers with Large Language Models via Reinforcement Learning | https://openreview.net/forum?id=u9JXu4L17I | [
"Pengcheng Jiang",
"Jiacheng Lin",
"Lang Cao",
"Runchu Tian",
"SeongKu Kang",
"Zifeng Wang",
"Jimeng Sun",
"Jiawei Han"
] | null | null | Information retrieval systems are crucial for enabling effective access to large document collections. Recent approaches have leveraged Large Language Models (LLMs) to enhance retrieval performance through query augmentation, but often rely on expensive supervised learning or distillation techniques that require signif... | [
"Large Language Models",
"Information Retrieval",
"Reinforcement Learning"
] | DeepRetrieval trains query generation models through reinforcement learning instead of supervised data, achieving state-of-the-art performance across diverse retrieval tasks while being more efficient than existing approaches. | 315 | 2503.00223 | title_snapshot |
DW8U8ZWa1U | SAT: Dynamic Spatial Aptitude Training for Multimodal Language Models | https://openreview.net/forum?id=DW8U8ZWa1U | [
"Arijit Ray",
"Jiafei Duan",
"Ellis L Brown II",
"Reuben Tan",
"Dina Bashkirova",
"Rose Hendrix",
"Kiana Ehsani",
"Aniruddha Kembhavi",
"Bryan A. Plummer",
"Ranjay Krishna",
"Kuo-Hao Zeng",
"Kate Saenko"
] | null | null | Reasoning about motion and space is a fundamental cognitive capability that is required by multiple real-world applications. While many studies highlight that large multimodal language models (MLMs) struggle to reason about space, they only focus on static spatial relationships and not dynamic awareness of motion and s... | [
"spatial reasoning",
"vqa",
"multimodal language models"
] | dynamic simulations improve spatial reasoning in MLMs - both for static relationships and on our complex tasks that require reasoning about actions | 318 | 2412.07755 | title_snapshot |
zHdSCtNmM4 | Deep Binding of Language Model Virtual Personas: a Study on Approximating Political Partisan Misperceptions | https://openreview.net/forum?id=zHdSCtNmM4 | [
"Minwoo Kang",
"Suhong Moon",
"Seung Hyeong Lee",
"Ayush Raj",
"Joseph Suh",
"David Chan"
] | null | null | Large language models (LLMs) are increasingly capable of simulating human behavior, offering cost-effective ways to estimate user responses during the early phases of survey design. While previous studies have examined whether models can reflect individual opinions or attitudes, we argue that a higher-order binding of ... | [
"user approximation",
"metaperception",
"social psycholog",
"democratic backsliding",
"outgroup hostility"
] | We propose a method to build virtual personas for deeper user binding and demonstrate its superiority in approximating metaperception in political science. | 320 | 2504.11673 | title_snapshot |
s0p9xpORgP | Scalable Zeroth-Order Fine-Tuning for Extremely Large Language Models with Limited GPU Memory | https://openreview.net/forum?id=s0p9xpORgP | [
"Liangyu Wang",
"Jie Ren",
"Hang Xu",
"Junxiao Wang",
"Huanyi Xie",
"David E. Keyes",
"Di Wang"
] | null | null | Fine-tuning large pre-trained LLMs generally demands extensive GPU memory. Traditional first-order optimizers like SGD encounter substantial difficulties due to increased memory requirements from storing activations and gradients during both the forward and backward phases as the model size expands. Alternatively, zero... | [
"Zeroth-Order Optimization",
"LLMs",
"Fine-Tuning"
] | ZO2 is a memory-efficient framework that enables zeroth-order fine-tuning of large language models like OPT-175B on a single 18GB GPU. | 323 | 2503.12668 | title_judge |
wyYL5Jov6e | EnrichIndex: Using LLMs to Enrich Retrieval Indices Offline | https://openreview.net/forum?id=wyYL5Jov6e | [
"Peter Baile Chen",
"Tomer Wolfson",
"Mike Cafarella",
"Dan Roth"
] | null | null | Existing information retrieval systems excel in cases where the language of target documents closely matches that of the user query. However, real-world retrieval systems are often required to *implicitly reason* whether a document is relevant. For example, when retrieving technical texts or tables, their relevance to ... | [
"retrieval",
"offline enrichment",
"implicit reasoning"
] | EnrichIndex enriches documents offline using LLMs, improving retrieval performance on complex retrieval tasks with significantly lower latency and online cost. | 324 | 2504.03598 | title_snapshot |
JloZnCwhmk | Understanding Layer Significance in LLM Alignment | https://openreview.net/forum?id=JloZnCwhmk | [
"Guangyuan SHI",
"ZEXIN LU",
"Xiaoyu DONG",
"Wenlong Zhang",
"Xuanyu Zhang",
"Yujie Feng",
"Xiao-Ming Wu"
] | null | null | Aligning large language models (LLMs) through supervised fine-tuning is essential for tailoring them to specific applications. Recent studies suggest that alignment primarily adjusts a model's presentation style rather than its foundational knowledge, indicating that only certain components of the model are significant... | [
"Layer Significance",
"Language Model Alignment"
] | We propose an algorithm to identify which layers within LLMs are most critical to the alignment process. | 327 | 2410.17875 | title_snapshot |
29jP6OsrIQ | Context-Adaptive Multi-Prompt Embedding with Large Language Models for Vision-Language Alignment | https://openreview.net/forum?id=29jP6OsrIQ | [
"Dahun Kim",
"Anelia Angelova"
] | null | null | We propose Context-Adaptive Multi-Prompt Embedding, a novel approach to enrich semantic representations in vision-language contrastive learning. Unlike standard CLIP-style models that rely on a single text embedding, our method introduces multiple structured prompts, each containing a distinct adaptive token that captu... | [
"CLIP",
"contrastive learning",
"LLM embedding"
] | We propose a method that uses multiple context-adaptive prompts with a pretrained LLM to generate diverse text embeddings for contrastive vision-language learning. | 328 | 2508.02762 | title_snapshot |
S4nTXotasR | Bootstrapping Visual Assistant Modeling with Situated Interaction Simulation | https://openreview.net/forum?id=S4nTXotasR | [
"Yichi Zhang",
"Run Peng",
"Yinpei Dai",
"Lingyun Wu",
"Xuweiyi Chen",
"Qiaozi Gao",
"Joyce Chai"
] | null | null | Visual assistants that can guide humans through complex tasks in physical environments have significant potential, yet their development is hindered by the high cost of human-in-the-loop data collection. We present BASIS (Bootstrapping Assistant modeling with Situated Interaction Simulation), a novel framework that fun... | [
"visual assistant",
"embodied",
"simulation",
"multimodal",
"LLM agent",
"situated dialogue"
] | We show that synthetic interaction data from simulated users and assistants can boost the development of visual assistant models that effectively guide real users to complete complex tasks. | 329 | null | null |
uBg8PClMUu | ProsodyLM: Uncovering the Emerging Prosody Processing Capabilities in Speech Language Models | https://openreview.net/forum?id=uBg8PClMUu | [
"Kaizhi Qian",
"Xulin Fan",
"Junrui Ni",
"Slava Shechtman",
"Mark A. Hasegawa-Johnson",
"Chuang Gan",
"Yang Zhang"
] | null | null | Speech language models refer to language models with speech processing and understanding capabilities. One key desirable capability for speech language models is the ability to capture the intricate interdependency between content and prosody. The existing mainstream paradigm of training speech language models, which c... | [
"Speech LM: Multi-modal LLM"
] | We propose ProsodyLM, a speech language model that demonstrate impressive emerging prosody generation and understand capabilities simply through pre-training on 30k audiobooks. | 332 | 2507.20091 | title_snapshot |
7qhBXq0NLN | IMPersona: Evaluating Individual Level LLM Impersonation | https://openreview.net/forum?id=7qhBXq0NLN | [
"Quan Shi",
"Carlos E Jimenez",
"Stephen Dong",
"Brian Seo",
"Caden Yao",
"Adam Kelch",
"Karthik R Narasimhan"
] | null | null | As language models achieve increasingly human-like capabilities in conversational text generation, a critical question emerges: to what extent can these systems simulate the characteristics of specific individuals? To evaluate this, we introduce IMPersona, a framework for evaluating LMs at impersonating specific indivi... | [
"Impersonation",
"Language Models",
"Personalization",
"Stylistic Mimicry",
"Contextual Knowledge",
"AI Evaluation",
"Social Engineering",
"Ethical AI",
"Memory-Augmented Models",
"Human-AI Interaction"
] | We train LLMs to impersonate individuals by mimicking style and personal knowledge, surpassing prompting methods, while raising safety and alignment concerns. | 333 | 2504.04332 | title_judge |
8Pxdzsqvx9 | JailDAM: Jailbreak Detection with Adaptive Memory for Vision-Language Model | https://openreview.net/forum?id=8Pxdzsqvx9 | [
"Yi Nian",
"Shenzhe Zhu",
"Yuehan Qin",
"Li Li",
"Ziyi Wang",
"Chaowei Xiao",
"Yue Zhao"
] | null | null | Multimodal large language models (MLLMs) excel in vision-language tasks but also pose significant risks of generating harmful content, particularly through jailbreak attacks. Jailbreak attacks refer to intentional manipulations that bypass safety mechanisms in models, leading to the generation of inappropriate or unsaf... | [
"Multi-modality",
"AI Security",
"Jailbreak",
"Attack Defense"
] | A test-time adaptive framework for harmful content detection in VLMs using memory-based unsafe concept representations, eliminating the need for labeled harmful data or model internals, and achieving state-of-the-art accuracy and efficiency. | 336 | 2504.03770 | title_snapshot |
NC6G1KCxlt | Phased Training for LLM-powered Text Retrieval Models Beyond Data Scaling | https://openreview.net/forum?id=NC6G1KCxlt | [
"Xin Zhang",
"Yanzhao Zhang",
"Wen Xie",
"Dingkun Long",
"Mingxin Li",
"Pengjun Xie",
"Meishan Zhang",
"Wenjie Li",
"Min Zhang"
] | null | null | Current efforts in building large language models (LLMs) based general-purpose text retrieval models primarily focus on architectural design and training data scaling. However, significant challenges remain in effectively modeling diverse retrieval tasks and domains, including multi-task conflict, data imbalance, and t... | [
"Text Retrieval",
"Text Embedding",
"Reranking",
"LLM-based Embedding"
] | Training powerful general-purpose text embedding and reranking models by a multi-stage training framework and efficient data synthesis. | 340 | null | null |
0Y2zXLFBji | Impact-driven Context Filtering For Cross-file Code Completion | https://openreview.net/forum?id=0Y2zXLFBji | [
"Yanzhou Li",
"Shangqing Liu",
"Kangjie Chen",
"Tianwei Zhang",
"Yang Liu"
] | null | null | Retrieval-augmented generation (RAG) has recently demonstrated considerable potential for repository-level code completion, as it integrates cross-file knowledge with in-file preceding code to provide comprehensive contexts for generation. To better understand the contribution of the retrieved cross-file contexts, we i... | [
"Code Completion; Adaptive Retreivla-augmented Generation; Large Language Model"
] | We introduce an adaptive retrieval-augmented framework for repository-level code completion, which automatically filters irrelevant context to enhance completion accuracy and efficiency. | 346 | 2508.05970 | title_snapshot |
6vTv9M9ZAA | Building Instruction-Tuning Datasets from Human-Written Instructions with Open-Weight Large Language Models | https://openreview.net/forum?id=6vTv9M9ZAA | [
"Youmi Ma",
"Sakae Mizuki",
"Kazuki Fujii",
"Taishi Nakamura",
"Masanari Ohi",
"Hinari Shimada",
"Taihei Shiotani",
"Koshiro Saito",
"Koki Maeda",
"Kakeru Hattori",
"Takumi Okamoto",
"Shigeki Ishida",
"Rio Yokota",
"Hiroya Takamura",
"Naoaki Okazaki"
] | null | null | Instruction tuning is crucial for enabling Large Language Models (LLMs) to solve real-world tasks.
Prior work has shown the effectiveness of instruction-tuning data synthesized solely from LLMs, raising a fundamental question: Do we still need human-originated signals for instruction tuning?
This work answers the quest... | [
"large language models; instruction tuning; synthetic data generation; cross-lingual datasets"
] | We constructed 4 state-of-the-art datasets in 2 languages for instruction tuning with permissive licenses, by simply appending responses generated by open-weight LLMs to human-written instructions. | 359 | 2503.23714 | title_snapshot |
bdCWK4NkK7 | Hawkeye: Model Collaboration for Efficient Reasoning | https://openreview.net/forum?id=bdCWK4NkK7 | [
"Jianshu She",
"Zhuohao Li",
"Zhemin Huang",
"Qi Li",
"Peiran Xu",
"Haonan Li",
"Qirong Ho"
] | null | null | Chain-of-Thought (CoT) reasoning has demonstrated remarkable effectiveness in enhancing the reasoning abilities of large language models (LLMs). However, its efficiency remains a challenge due to excessive intermediate reasoning tokens, which introduce both semantic redundancy and unnecessarily detailed reasoning steps... | [
"reinforcement learning (with human feedback)",
"fine-tuning",
"compression",
"decoding algorithms",
"reasoning algorithms"
] | We provide an efficient inference pipeline that optimizes Chain-of-Thought (CoT) reasoning by instructing a Large Language Model (LLM) to generate concise yet effective CoTs for a Small Language Model (SLM) to decode through reinforcement learning. | 363 | null | null |
End of preview. Expand in Data Studio
README.md exists but content is empty.
- Downloads last month
- -