Daily Papers

We introduce LinuxArena, a control setting in which agents operate directly on live, multi-service production environments. LinuxArena contains 20 environments, 1,671 main tasks representing legitimate software engineering work, and 184 side tasks representing safety failures such as data exfiltration and backdooring, making it the largest and most diverse control setting for software engineering to date. We validate LinuxArena is useful for control research by running sabotage evaluations, which measure whether attackers can complete side tasks while working on main tasks, and monitor evaluations, which measure a monitor model's ability to detect sabotage attempts. Against a GPT-5-nano trusted monitor at a 1\% step-wise false positive rate, Claude Opus 4.6 achieves roughly a 23% undetected sabotage success rate. We additionally release LaStraj, a dataset of human-crafted attack trajectories that evade monitors at substantially higher rates than any model-generated attacks we elicited, showing that current attack policies do not saturate LinuxArena. These results suggest that LinuxArena has meaningful headroom for both attackers and defenders, making it a strong testbed for developing and evaluating future control protocols.

Robust perception-action models should be learned from training data with diverse visual appearances and realistic behaviors, yet current approaches to deep visuomotor policy learning have been generally limited to in-situ models learned from a single vehicle or a simulation environment. We advocate learning a generic vehicle motion model from large scale crowd-sourced video data, and develop an end-to-end trainable architecture for learning to predict a distribution over future vehicle egomotion from instantaneous monocular camera observations and previous vehicle state. Our model incorporates a novel FCN-LSTM architecture, which can be learned from large-scale crowd-sourced vehicle action data, and leverages available scene segmentation side tasks to improve performance under a privileged learning paradigm.

AI control protocols serve as a defense mechanism to stop untrusted LLM agents from causing harm in autonomous settings. Prior work treats this as a security problem, stress testing with exploits that use the deployment context to subtly complete harmful side tasks, such as backdoor insertion. In practice, most AI control protocols are fundamentally based on LLM monitors, which can become a central point of failure. We study adaptive attacks by an untrusted model that knows the protocol and the monitor model, which is plausible if the untrusted model was trained with a later knowledge cutoff or can search for this information autonomously. We instantiate a simple adaptive attack vector by which the attacker embeds publicly known or zero-shot prompt injections in the model outputs. Using this tactic, frontier models consistently evade diverse monitors and complete malicious tasks on two main AI control benchmarks. The attack works universally against current protocols that rely on a monitor. Furthermore, the recent Defer-to-Resample protocol even backfires, as its resampling amplifies the prompt injection and effectively reframes it as a best-of-n attack. In general, adaptive attacks on monitor models represent a major blind spot in current control protocols and should become a standard component of evaluations for future AI control mechanisms.

As AI models are deployed with increasing autonomy, it is important to ensure they do not take harmful actions unnoticed. As a potential mitigation, we investigate Chain-of-Thought (CoT) monitoring, wherein a weaker trusted monitor model continuously oversees the intermediate reasoning steps of a more powerful but untrusted model. We compare CoT monitoring to action-only monitoring, where only final outputs are reviewed, in a red-teaming setup where the untrusted model is instructed to pursue harmful side tasks while completing a coding problem. We find that while CoT monitoring is more effective than overseeing only model outputs in scenarios where action-only monitoring fails to reliably identify sabotage, reasoning traces can contain misleading rationalizations that deceive the CoT monitors, reducing performance in obvious sabotage cases. To address this, we introduce a hybrid protocol that independently scores model reasoning and actions, and combines them using a weighted average. Our hybrid monitor consistently outperforms both CoT and action-only monitors across all tested models and tasks, with detection rates twice higher than action-only monitoring for subtle deception scenarios.

We present a simple and effective pretraining strategy {D}en{o}ising {T}raining DoT for neural machine translation. Specifically, we update the model parameters with source- and target-side denoising tasks at the early stage and then tune the model normally. Notably, our approach does not increase any parameters or training steps, requiring the parallel data merely. Experiments show that DoT consistently improves the neural machine translation performance across 12 bilingual and 16 multilingual directions (data size ranges from 80K to 20M). In addition, we show that DoT can complement existing data manipulation strategies, i.e. curriculum learning, knowledge distillation, data diversification, bidirectional training, and back-translation. Encouragingly, we found that DoT outperforms costly pretrained model mBART in high-resource settings. Analyses show DoT is a novel in-domain cross-lingual pretraining strategy and could offer further improvements with task-relevant self-supervisions.

Solving arithmetic tasks is a simple and fundamental skill, yet modern Large Language Models (LLMs) have great difficulty with them. We introduce the Integrated Gated Calculator (IGC), a module that enables LLMs to perform arithmetic by emulating a calculator on the GPU. We finetune a Llama model with our module and test it on the BigBench Arithmetic benchmark, where it beats the State of the Art, outperforming all models on the benchmark, including models almost two orders of magnitude larger. Our approach takes only a single iteration to run and requires no external tools. It performs arithmetic operations entirely inside the LLM without the need to produce intermediate tokens. It is computationally efficient, interpretable, and avoids side-effects on tasks that do not require arithmetic operations. It reliably achieves 98\% to 99\% accuracy across multiple training runs and for all subtasks, including the substantially harder subtask of multiplication, which was previously unsolved.

Pre-trained video large language models (Video LLMs) exhibit remarkable reasoning capabilities, yet adapting these models to new tasks involving additional modalities or data types (e.g., audio or 3D information) remains challenging. In this paper, we present PAVE, a flexible framework for adapting pre-trained Video LLMs to downstream tasks with side-channel signals, such as audio, 3D cues, or multi-view videos. PAVE introduces lightweight adapters, referred to as "patches," which add a small number of parameters and operations to a base model without modifying its architecture or pre-trained weights. In doing so, PAVE can effectively adapt the pre-trained base model to support diverse downstream tasks, including audio-visual question answering, 3D reasoning, multi-view video recognition, and high frame rate video understanding. Across these tasks, PAVE significantly enhances the performance of the base model, surpassing state-of-the-art task-specific models while incurring a minor cost of ~0.1% additional FLOPs and parameters. Further, PAVE supports multi-task learning and generalizes well across different Video LLMs. Our code is available at https://github.com/dragonlzm/PAVE.

Real-world photography requires capture-time guidance for both camera framing and subject pose. Yet existing aesthetic cropping benchmarks mainly evaluate post-hoc crop prediction and overlook subject-side recommendations, leaving the capture-time guidance capabilities of multimodal large language models (MLLMs) underexplored. To address this gap, we introduce CaptureGuide-Bench, a benchmark with two complementary tasks: photographer-side composition decision and refinement, and subject-side scene-conditioned pose recommendation. Our evaluation reveals limitations: general-purpose MLLMs can make composition decisions but lack precise refinement localization, while specialized aesthetic cropping models localize crops effectively but are limited to refinement; neither provides actionable pose guidance. To support model development, we further construct CaptureGuide-Dataset, comprising 130K samples with textual rationales and structured visual annotations, and develop ShutterMuse, a unified MLLM trained with supervised and reinforcement fine-tuning. Experiments on CaptureGuide-Bench show that ShutterMuse achieves the best overall photographer-side performance among evaluated baselines and competitive subject-side pose recommendation with substantially lower inference cost, demonstrating the potential of MLLMs as interactive assistants for photography during image capture.

Object orientation understanding represents a fundamental challenge in visual perception critical for applications like robotic manipulation and augmented reality. Current vision-language benchmarks fail to isolate this capability, often conflating it with positional relationships and general scene understanding. We introduce DORI (Discriminative Orientation Reasoning Intelligence), a comprehensive benchmark establishing object orientation perception as a primary evaluation target. DORI assesses four dimensions of orientation comprehension: frontal alignment, rotational transformations, relative directional relationships, and canonical orientation understanding. Through carefully curated tasks from 11 datasets spanning 67 object categories across synthetic and real-world scenarios, DORI provides insights on how multi-modal systems understand object orientations. Our evaluation of 15 state-of-the-art vision-language models reveals critical limitations: even the best models achieve only 54.2% accuracy on coarse tasks and 33.0% on granular orientation judgments, with performance deteriorating for tasks requiring reference frame shifts or compound rotations. These findings demonstrate the need for dedicated orientation representation mechanisms, as models show systematic inability to perform precise angular estimations, track orientation changes across viewpoints, and understand compound rotations - suggesting limitations in their internal 3D spatial representations. As the first diagnostic framework specifically designed for orientation awareness in multimodal systems, DORI offers implications for improving robotic control, 3D scene reconstruction, and human-AI interaction in physical environments. DORI data: https://huggingface.co/datasets/appledora/DORI-Benchmark

Fine-tuning pre-trained Vision Transformers (ViT) has consistently demonstrated promising performance in the realm of visual recognition. However, adapting large pre-trained models to various tasks poses a significant challenge. This challenge arises from the need for each model to undergo an independent and comprehensive fine-tuning process, leading to substantial computational and memory demands. While recent advancements in Parameter-efficient Transfer Learning (PETL) have demonstrated their ability to achieve superior performance compared to full fine-tuning with a smaller subset of parameter updates, they tend to overlook dense prediction tasks such as object detection and segmentation. In this paper, we introduce Hierarchical Side-Tuning (HST), a novel PETL approach that enables ViT transfer to various downstream tasks effectively. Diverging from existing methods that exclusively fine-tune parameters within input spaces or certain modules connected to the backbone, we tune a lightweight and hierarchical side network (HSN) that leverages intermediate activations extracted from the backbone and generates multi-scale features to make predictions. To validate HST, we conducted extensive experiments encompassing diverse visual tasks, including classification, object detection, instance segmentation, and semantic segmentation. Notably, our method achieves state-of-the-art average Top-1 accuracy of 76.0% on VTAB-1k, all while fine-tuning a mere 0.78M parameters. When applied to object detection tasks on COCO testdev benchmark, HST even surpasses full fine-tuning and obtains better performance with 49.7 box AP and 43.2 mask AP using Cascade Mask R-CNN.

Dataflow neural network accelerators efficiently process AI tasks on FPGAs, with deployment simplified by ready-to-use frameworks and pre-trained models. However, this convenience makes them vulnerable to malicious actors seeking to reverse engineer valuable Intellectual Property (IP) through Side-Channel Attacks (SCA). This paper proposes a methodology to recover the hardware configuration of dataflow accelerators generated with the FINN framework. Through unsupervised dimensionality reduction, we reduce the computational overhead compared to the state-of-the-art, enabling lightweight classifiers to recover both folding and quantization parameters. We demonstrate an attack phase requiring only 337 ms to recover the hardware parameters with an accuracy of more than 95% and 421 ms to fully recover these parameters with an averaging of 4 traces for a FINN-based accelerator running a CNN, both using a random forest classifier on side-channel traces, even with the accelerator dataflow fully loaded. This approach offers a more realistic attack scenario than existing methods, and compared to SoA attacks based on tsfresh, our method requires 940x and 110x less time for preparation and attack phases, respectively, and gives better results even without averaging traces.

Extending a vision-language-action (VLA) policy to a new task typically requires task-specific teleoperated demonstrations and per-task fine-tuning, making adaptation costly in both data collection and compute. In this paper, we show that this target-side per-task adaptation cost can be replaced by retrieval. Our retrieval-augmented policy is trained once on paired demonstrations from the target embodiment (query) and a cheaper embodiment (pool, e.g., human-hand video), then frozen. New tasks are added at deployment by appending pool-side demonstrations to a retrieval pool. The frozen policy conditions on retrieved trajectories at every control step, so new tasks are absorbed by indexing data rather than updating parameters. Fine-tuning is needed only to take on a new, unseen embodiment, not for each new task. We show that retrieval improves policies beyond a specific backbone, including standard VLA policies, but its effect is especially pronounced in Cosmos Policy, a video-generation-based world-action model (WAM). In this setting, retrieval supplies coarse task progression, while the WAM's future-image objective provides an additional visual consistency signal that strengthens the retrieval-conditioned actions. On PushT, we study how retrieval provides a reusable high-level motion prior for cross-embodiment generalization to unseen goal angles, while on RoboTwin 2.0 our method outperforms cross-embodiment baselines on unseen tasks, and we additionally demonstrate the method on a real robot.

With the recent advances of large language models (LLMs), it is no longer infeasible to build an automated debate system that helps people to synthesise persuasive arguments. Previous work attempted this task by integrating multiple components. In our work, we introduce an argument mining dataset that captures the end-to-end process of preparing an argumentative essay for a debate, which covers the tasks of claim and evidence identification (Task 1 ED), evidence convincingness ranking (Task 2 ECR), argumentative essay summarisation and human preference ranking (Task 3 ASR) and metric learning for automated evaluation of resulting essays, based on human feedback along argument quality dimensions (Task 4 SQE). Our dataset contains 14k examples of claims that are fully annotated with the various properties supporting the aforementioned tasks. We evaluate multiple generative baselines for each of these tasks, including representative LLMs. We find, that while they show promising results on individual tasks in our benchmark, their end-to-end performance on all four tasks in succession deteriorates significantly, both in automated measures as well as in human-centred evaluation. This challenge presented by our proposed dataset motivates future research on end-to-end argument mining and summarisation. The repository of this project is available at https://github.com/HarrywillDr/ArgSum-Datatset

Fine-tuning large pre-trained models on downstream tasks has been adopted in a variety of domains recently. However, it is costly to update the entire parameter set of large pre-trained models. Although recently proposed parameter-efficient transfer learning (PETL) techniques allow updating a small subset of parameters (e.g. only using 2% of parameters) inside a pre-trained backbone network for a new task, they only reduce the training memory requirement by up to 30%. This is because the gradient computation for the trainable parameters still requires backpropagation through the large pre-trained backbone model. To address this, we propose Ladder Side-Tuning (LST), a new PETL technique that can reduce training memory requirements by more substantial amounts. Unlike existing parameter-efficient methods that insert additional parameters inside backbone networks, we train a ladder side network, a small and separate network that takes intermediate activations as input via shortcut connections (called ladders) from backbone networks and makes predictions. LST has significantly lower memory requirements than previous methods, because it does not require backpropagation through the backbone network, but instead only through the side network and ladder connections. We evaluate our method with various models (T5 and CLIP-T5) on both NLP (GLUE) and vision-and-language (VQA, GQA, NLVR2 , MSCOCO) tasks. LST saves 69% of the memory costs to fine-tune the whole network, while other methods only save 26% of that in similar parameter usages (hence, 2.7x more memory savings). Moreover, LST achieves higher accuracy than Adapter and LoRA in a low-memory regime. To further show the advantage of this better memory efficiency, we also apply LST to larger T5 models, attaining better GLUE performance than full fine-tuning and other PETL methods. The accuracy-efficiency trade-off also holds on VL tasks.

Generative AI (GenAI) holds significant promise for automating everyday image editing tasks, especially following the recent release of GPT-4o on March 25, 2025. However, what subjects do people most often want edited? What kinds of editing actions do they want to perform (e.g., removing or stylizing the subject)? Do people prefer precise edits with predictable outcomes or highly creative ones? By understanding the characteristics of real-world requests and the corresponding edits made by freelance photo-editing wizards, can we draw lessons for improving AI-based editors and determine which types of requests can currently be handled successfully by AI editors? In this paper, we present a unique study addressing these questions by analyzing 83k requests from the past 12 years (2013-2025) on the Reddit community, which collected 305k PSR-wizard edits. According to human ratings, approximately only 33% of requests can be fulfilled by the best AI editors (including GPT-4o, Gemini-2.0-Flash, SeedEdit). Interestingly, AI editors perform worse on low-creativity requests that require precise editing than on more open-ended tasks. They often struggle to preserve the identity of people and animals, and frequently make non-requested touch-ups. On the other side of the table, VLM judges (e.g., o1) perform differently from human judges and may prefer AI edits more than human edits. Code and qualitative examples are available at: https://psrdataset.github.io

Executable software engineering data is valuable for training SWE agents, but scaling it remains difficult for two reasons: only a small fraction of real repository changes yield verifiable, high-signal task instances, and naively building repository-specific environments quickly becomes the dominant systems cost. We present SWE-Next, an execution-grounded framework for scalable SWE task and trajectory collection. On the data side, SWE-Next mines real merged pull requests, executes candidate base/merged commit pairs, and retains only those that produce strict test improvements without regressions, yielding self-verifying instances. It also applies strict submission gating so that collected trajectories remain evidence-driven rather than speculative. On the systems side, SWE-Next introduces reusable repo-quarter profiles, which reuse the same environment across nearby commits in time while keeping each task run separate and reproducible. Using only 30 hours and 639GB of environment storage, SWE-Next processes 3,971 seed repositories and 102,582 candidate commit pairs mined from real merged PRs to construct a dataset of 2,308 self-verifying instances. Experiments show that SWE-Next improves downstream pass@1 with fewer or comparable training trajectories, indicating that its gains come not from a stronger trajectory generator, but from higher-signal execution-grounded supervision and more efficient data collection.

Methods for improving knowledge use in large language models typically fall into two regimes. Non-parametric retrieval offers flexible access to external knowledge, but adds retrieval latency, context overhead, and only shallow integration with the backbone. Parametric adaptation is efficient at inference time, but entangles knowledge with model weights and can be hard to update, audit, or transfer. Engram-style hashed memory occupies a middle regime: it stores learned information in an external, addressable table, yet consumes that table through a small learned reader. This raises a basic question: when such a memory is moved across backbones, what matters more, the frozen memory itself or the target-side reader? We study this question through cross-model frozen-memory extraction, in which a memory trained on a source model is frozen and attached to a different target model, with only a lightweight reader trained. Ablations show that learned memory content and correct addressing both matter, but the transferred table becomes useful only through a reader aligned to the target model. In downstream question answering tasks, a dual-layer, four-branch reader nearly closes the gap between same-model and cross-model reuse, achieving an average score of 38.8 under our controlled evaluation protocol. Moreover, when the provider reader is directly compatible with the target interface, the frozen artifact can provide substantial utility without target-side training, while optional reader adaptation yields further improvement. These results suggest that Engram can serve as a reusable external knowledge artifact, provided that the target has access to a compatible reader interface; target-side adaptation can further improve alignment when direct reader reuse is insufficient.

The eXtreme Multi-label Classification~(XMC) problem seeks to find relevant labels from an exceptionally large label space. Most of the existing XMC learners focus on the extraction of semantic features from input query text. However, conventional XMC studies usually neglect the side information of instances and labels, which can be of use in many real-world applications such as recommendation systems and e-commerce product search. We propose Predicted Instance Neighborhood Aggregation (PINA), a data enhancement method for the general XMC problem that leverages beneficial side information. Unlike most existing XMC frameworks that treat labels and input instances as featureless indicators and independent entries, PINA extracts information from the label metadata and the correlations among training instances. Extensive experimental results demonstrate the consistent gain of PINA on various XMC tasks compared to the state-of-the-art methods: PINA offers a gain in accuracy compared to standard XR-Transformers on five public benchmark datasets. Moreover, PINA achieves a sim 5% gain in accuracy on the largest dataset LF-AmazonTitles-1.3M. Our implementation is publicly available.

Does the dominant approach to learn representations (as a side effect of optimizing an expected cost for a single training distribution) remain a good approach when we are dealing with multiple distributions? Our thesis is that such scenarios are better served by representations that are richer than those obtained with a single optimization episode. We support this thesis with simple theoretical arguments and with experiments utilizing an apparently na\"{\i}ve ensembling technique: concatenating the representations obtained from multiple training episodes using the same data, model, algorithm, and hyper-parameters, but different random seeds. These independently trained networks perform similarly. Yet, in a number of scenarios involving new distributions, the concatenated representation performs substantially better than an equivalently sized network trained with a single training run. This proves that the representations constructed by multiple training episodes are in fact different. Although their concatenation carries little additional information about the training task under the training distribution, it becomes substantially more informative when tasks or distributions change. Meanwhile, a single training episode is unlikely to yield such a redundant representation because the optimization process has no reason to accumulate features that do not incrementally improve the training performance.

We propose FocusCLIP, integrating subject-level guidance--a specialized mechanism for target-specific supervision--into the CLIP framework for improved zero-shot transfer on human-centric tasks. Our novel contributions enhance CLIP on both the vision and text sides. On the vision side, we incorporate ROI heatmaps emulating human visual attention mechanisms to emphasize subject-relevant image regions. On the text side, we introduce human pose descriptions to provide rich contextual information. For human-centric tasks, FocusCLIP is trained with images from the MPII Human Pose dataset. The proposed approach surpassed CLIP by an average of 8.61% across five previously unseen datasets covering three human-centric tasks. FocusCLIP achieved an average accuracy of 33.65% compared to 25.04% by CLIP. We observed a 3.98% improvement in activity recognition, a 14.78% improvement in age classification, and a 7.06% improvement in emotion recognition. Moreover, using our proposed single-shot LLM prompting strategy, we release a high-quality MPII Pose Descriptions dataset to encourage further research in multimodal learning for human-centric tasks. Furthermore, we also demonstrate the effectiveness of our subject-level supervision on non-human-centric tasks. FocusCLIP shows a 2.47% improvement over CLIP in zero-shot bird classification using the CUB dataset. Our findings emphasize the potential of integrating subject-level guidance with general pretraining methods for enhanced downstream performance.

When the context window is under pressure, LLM systems compact prior context to continue ongoing tasks. We identify a class of user-issued instructions, Session Constraints (SCs), such as "do not delete any emails until I confirm," that are meant to constrain LLM's behavior for the remainder of a session but are silently dropped during compaction. To quantify this loss, we introduce COMPINT, an evaluation suite that evaluates compactors across three long-context scenarios: multi-turn chat, agentic trajectory, and long-horizon research. Current compactors retain only 17% of injected SCs on average, and most perform worse than running the same task without compaction. Retention varies sharply with compactor, prompt, context length, SC phrasing, and injection location, showing that the loss is systematic rather than tied to any single setting. We propose an SC-aware extractor that runs alongside the compactor as a plug-and-play module, achieving over 90% retention across all three scenarios without modifying the compactor or LLM. The COMPINT evaluation suite and accompanying implementation are available at https://github.com/ZhiqiEliWang/compaction-integrity.

Finetuning large language models (LLMs) has been empirically effective on a variety of downstream tasks. Existing approaches to finetuning an LLM either focus on parameter-efficient finetuning, which only updates a small number of trainable parameters, or attempt to reduce the memory footprint during the training phase of the finetuning. Typically, the memory footprint during finetuning stems from three contributors: model weights, optimizer states, and intermediate activations. However, existing works still require considerable memory and none can simultaneously mitigate memory footprint for all three sources. In this paper, we present Quantized Side Tuing (QST), which enables memory-efficient and fast finetuning of LLMs by operating through a dual-stage process. First, QST quantizes an LLM's model weights into 4-bit to reduce the memory footprint of the LLM's original weights; QST also introduces a side network separated from the LLM, which utilizes the hidden states of the LLM to make task-specific predictions. Using a separate side network avoids performing backpropagation through the LLM, thus reducing the memory requirement of the intermediate activations. Furthermore, QST leverages several low-rank adaptors and gradient-free downsample modules to significantly reduce the trainable parameters, so as to save the memory footprint of the optimizer states. Experiments show that QST can reduce the total memory footprint by up to 2.3 times and speed up the finetuning process by up to 3 times while achieving competent performance compared with the state-of-the-art. When it comes to full finetuning, QST can reduce the total memory footprint up to 7 times.

Textual label names (descriptions) are typically semantically rich in many natural language understanding (NLU) tasks. In this paper, we incorporate the prompting methodology, which is widely used to enrich model input, into the label side for the first time. Specifically, we propose a Mask Matching method, which equips an input with a prompt and its label with another, and then makes predictions by matching their mask representations. We evaluate our method extensively on 8 NLU tasks with 14 datasets. The experimental results show that Mask Matching significantly outperforms its counterparts of fine-tuning and conventional prompt-tuning, setting up state-of-the-art performances in several datasets. Mask Matching is particularly good at handling NLU tasks with large label counts and informative label names. As pioneering efforts that investigate the label-side prompt, we also discuss open issues for future study.

The large integration of microphones into devices increases the opportunities for Acoustic Side-Channel Attacks (ASCAs), as these can be used to capture keystrokes' audio signals that might reveal sensitive information. However, the current State-Of-The-Art (SOTA) models for ASCAs, including Convolutional Neural Networks (CNNs) and hybrid models, such as CoAtNet, still exhibit limited robustness under realistic noisy conditions. Solving this problem requires either: (i) an increased model's capacity to infer contextual information from longer sequences, allowing the model to learn that an initially noisily typed word is the same as a futurely collected non-noisy word, or (ii) an approach to fix misidentified information from the contexts, as one does not type random words, but the ones that best fit the conversation context. In this paper, we demonstrate that both strategies are viable and complementary solutions for making ASCAs practical. We observed that no existing solution leverages advanced transformer architectures' power for these tasks and propose that: (i) Visual Transformers (VTs) are the candidate solutions for capturing long-term contextual information and (ii) transformer-powered Large Language Models (LLMs) are the candidate solutions to fix the ``typos'' (mispredictions) the model might make. Thus, we here present the first-of-its-kind approach that integrates VTs and LLMs for ASCAs. We first show that VTs achieve SOTA performance in classifying keystrokes when compared to the previous CNN benchmark. Second, we demonstrate that LLMs can mitigate the impact of real-world noise. Evaluations on the natural sentences revealed that: (i) incorporating LLMs (e.g., GPT-4o) in our ASCA pipeline boosts the performance of error-correction tasks; and (ii) the comparable performance can be attained by a lightweight, fine-tuned smaller LLM (67 times smaller than GPT-4o), using...

We consider representation misdirection (RM), a class of LLM unlearning methods that achieves forgetting by manipulating the forget-representations, that is, latent representations of forget samples. Despite being important, the roles of target vectors used in RM, however, remain underexplored. Here, we approach and revisit RM through the lens of the linear representation hypothesis. Specifically, if one can somehow identify a one-dimensional representation corresponding to a high-level concept, the linear representation hypothesis enables linear operations on this concept vector within the forget-representation space. Under this view, we hypothesize that, beyond forgetting, machine unlearning elicits controllable side behaviors and stronger side capabilities corresponding to the high-level concept. Our hypothesis is empirically validated across a wide range of tasks, including behavioral control (e.g., controlling unlearned models' truth, sentiment, and refusal) and capability enhancement (e.g., improving unlearned models' in-context learning capability). Our findings reveal that this fairly attractive phenomenon could be either a hidden risk if misused or a mechanism that can be harnessed for developing models that require stronger capabilities and controllable behaviors.

LLM agents increasingly face long-horizon tasks such as web search and deep research in real-world applications, where accumulated context can cause long-context degradation and reasoning failures. Prior work mitigates this through context management with agent-side context control or fixed strategies such as summarization, which require training the agent itself for adaptation - making it impractical for closed-source agents and ignoring that different agents may require different strategies. We introduce Adaptive Context Management (AdaCoM), which trains an external LLM to manage the context of a frozen agent through flexible modification actions and end-to-end reinforcement learning. Across diverse agents on web search and deep research benchmarks, AdaCoM substantially improves performance by preserving task constraints and progress while pruning stale content. The learned strategies reveal a Fidelity-Reliability Trade-off: agents with higher vanilla ReAct performance benefit from higher-fidelity context preservation, whereas lower-performing agents require more aggressive compression to stay within a reliable reasoning regime. Transfer experiments show that AdaCoM generalizes most effectively across agents with similar capability (measured by vanilla ReAct performance), suggesting a practical path toward reusable context managers for agent systems.

Large Language Models demonstrate remarkable proficiency in static reasoning, yet training them as autonomous agents through Reinforcement Learning (RL) for long-horizon tasks is often hindered by severe reward sparsity. While conventional agent-side warming up via supervised fine-tuning (SFT) can alleviate this, it is frequently limited by data scarcity and constrained exploration. To address this, we propose a paradigm shift to environment-side adaptation by constructing Feedback-Enriched Environments (FEEs). Through a pilot study, we establish a feedback design strategy that reformulates environments by transitioning from action guidance to observation enrichment during the later stages of both intra-episode exploration and inter-episode evolution. Large-scale experiments on SciWorld and BFCL benchmarks using various Qwen3 model scales and RL algorithms such as GRPO, GSPO, and DAPO demonstrate that FEEs consistently yield performance improvements over standard settings. Furthermore, our analysis reveals that training with FEEs (1) stabilizes training dynamics by reducing entropy volatility, (2) facilitates proactive state-space exploration in difficult tasks, (3) ensures the internalization of environmental guidance into policy weights rather than acting as a mere inference-time prior, and (4) identifies intra-group feedback consistency as a critical boundary for stable optimization.

Speculative decoding losslessly accelerates large language model inference by having a lightweight draft model predict future tokens for verification by the target model. Recent block diffusion drafters further reduce drafting latency by predicting multiple tokens in parallel. However, existing block drafters project target hidden states at every input position into a separate drafter-side KV cache, incurring per-request memory and KV-write overhead that grow with concurrency; directly reusing target KVs in place removes this cache but fails to sustain draft quality throughout the block. We propose a hybrid target-context injection method that complements direct target KV reuse with target hidden states only at the last input position, requiring no separate drafter-side KV cache. Building on this design, we propose H-Spec, a hybrid Mamba-attention parallel drafter that consumes the two target-context sources through complementary modules. Mamba modules are initialized with projected last-token target hidden states, while attention modules reuse target KVs in place. Despite its recurrent formulation, Mamba's parallel scan allows H-Spec to preserve block-parallel drafting. Across three target models and diverse tasks, H-Spec improves over the best baseline by 5.0--13.3% in mean accepted length and 5.3--12.6% in batch-size-1 inter-token latency speedup. Under concurrent serving, H-Spec consistently achieves higher throughput while maintaining lower KV cache utilization than baselines across evaluated concurrency levels.

Recent advancements in machine learning research, i.e., deep learning, introduced methods that excel conventional algorithms as well as humans in several complex tasks, ranging from detection of objects in images and speech recognition to playing difficult strategic games. However, the current methodology of machine learning research and consequently, implementations of the real-world applications of such algorithms, seems to have a recurring HARKing (Hypothesizing After the Results are Known) issue. In this work, we elaborate on the algorithmic, economic and social reasons and consequences of this phenomenon. We present examples from current common practices of conducting machine learning research (e.g. avoidance of reporting negative results) and failure of generalization ability of the proposed algorithms and datasets in actual real-life usage. Furthermore, a potential future trajectory of machine learning research and development from the perspective of accountable, unbiased, ethical and privacy-aware algorithmic decision making is discussed. We would like to emphasize that with this discussion we neither claim to provide an exhaustive argumentation nor blame any specific institution or individual on the raised issues. This is simply a discussion put forth by us, insiders of the machine learning field, reflecting on us.

Recent advances in 3D AIGC have shown promise in directly creating 3D objects from text and images, offering significant cost savings in animation and product design. However, detailed edit and customization of 3D assets remains a long-standing challenge. Specifically, 3D Generation methods lack the ability to follow finely detailed instructions as precisely as their 2D image creation counterparts. Imagine you can get a toy through 3D AIGC but with undesired accessories and dressing. To tackle this challenge, we propose a novel pipeline called Tailor3D, which swiftly creates customized 3D assets from editable dual-side images. We aim to emulate a tailor's ability to locally change objects or perform overall style transfer. Unlike creating 3D assets from multiple views, using dual-side images eliminates conflicts on overlapping areas that occur when editing individual views. Specifically, it begins by editing the front view, then generates the back view of the object through multi-view diffusion. Afterward, it proceeds to edit the back views. Finally, a Dual-sided LRM is proposed to seamlessly stitch together the front and back 3D features, akin to a tailor sewing together the front and back of a garment. The Dual-sided LRM rectifies imperfect consistencies between the front and back views, enhancing editing capabilities and reducing memory burdens while seamlessly integrating them into a unified 3D representation with the LoRA Triplane Transformer. Experimental results demonstrate Tailor3D's effectiveness across various 3D generation and editing tasks, including 3D generative fill and style transfer. It provides a user-friendly, efficient solution for editing 3D assets, with each editing step taking only seconds to complete.

We propose Meta-Prod2vec, a novel method to compute item similarities for recommendation that leverages existing item metadata. Such scenarios are frequently encountered in applications such as content recommendation, ad targeting and web search. Our method leverages past user interactions with items and their attributes to compute low-dimensional embeddings of items. Specifically, the item metadata is in- jected into the model as side information to regularize the item embeddings. We show that the new item representa- tions lead to better performance on recommendation tasks on an open music dataset.

Nowadays, developers increasingly rely on solutions powered by Large Language Models (LLM) to assist them with their coding tasks. This makes it crucial to align these tools with human values to prevent malicious misuse. In this paper, we propose a comprehensive framework for assessing the potential harmfulness of LLMs within the software engineering domain. We begin by developing a taxonomy of potentially harmful software engineering scenarios and subsequently, create a dataset of prompts based on this taxonomy. To systematically assess the responses, we design and validate an automatic evaluator that classifies the outputs of a variety of LLMs both open-source and closed-source models, as well as general-purpose and code-specific LLMs. Furthermore, we investigate the impact of models size, architecture family, and alignment strategies on their tendency to generate harmful content. The results show significant disparities in the alignment of various LLMs for harmlessness. We find that some models and model families, such as Openhermes, are more harmful than others and that code-specific models do not perform better than their general-purpose counterparts. Notably, some fine-tuned models perform significantly worse than their base-models due to their design choices. On the other side, we find that larger models tend to be more helpful and are less likely to respond with harmful information. These results highlight the importance of targeted alignment strategies tailored to the unique challenges of software engineering tasks and provide a foundation for future work in this critical area.

Turning garments right-side out is a challenging manipulation task: it is highly dynamic, entails rapid contact changes, and is subject to severe visual occlusion. We introduce Right-Side-Out, a zero-shot sim-to-real framework that effectively solves this challenge by exploiting task structures. We decompose the task into Drag/Fling to create and stabilize an access opening, followed by Insert&Pull to invert the garment. Each step uses a depth-inferred, keypoint-parameterized bimanual primitive that sharply reduces the action space while preserving robustness. Efficient data generation is enabled by our custom-built, high-fidelity, GPU-parallel Material Point Method (MPM) simulator that models thin-shell deformation and provides robust and efficient contact handling for batched rollouts. Built on the simulator, our fully automated pipeline scales data generation by randomizing garment geometry, material parameters, and viewpoints, producing depth, masks, and per-primitive keypoint labels without any human annotations. With a single depth camera, policies trained entirely in simulation deploy zero-shot on real hardware, achieving up to 81.3% success rate. By employing task decomposition and high fidelity simulation, our framework enables tackling highly dynamic, severely occluded tasks without laborious human demonstrations.

Large language models (LLMs) are frequently fine-tuned or unlearned to adapt to new tasks or eliminate undesirable behaviors. While existing evaluation methods assess performance after such interventions, there remains no general approach for detecting unintended side effects, such as unlearning biology content degrading performance on chemistry tasks, particularly when these effects are unpredictable or emergent. To address this issue, we introduce MNEME, Model diffiNg for Evaluating Mechanistic Effects, a lightweight framework for identifying these side effects using sparse model diffing. MNEME compares base and fine-tuned models on task-agnostic data (for example, The Pile, LMSYS-Chat-1M) without access to fine-tuning data to isolate behavioral shifts. Applied to five LLMs across three scenarios: WMDP knowledge unlearning, emergent misalignment, and benign fine-tuning, MNEME achieves up to 95 percent accuracy in predicting side effects, aligning with known benchmarks and requiring no custom heuristics. Furthermore, we show that retraining on high-activation samples can partially reverse these effects. Our results demonstrate that sparse probing and diffing offer a scalable and automated lens into fine-tuning-induced model changes, providing practical tools for understanding and managing LLM behavior.

Browser-use agents often carry information in their context as they move between websites. While it may be necessary for task completion, it also creates a privacy risk, especially when the information contains a private fact regarding the user. For example, an agent may learn a user's affiliation after reading a membership record. If it later selects a registration option specific to that affiliation on another website instead of a general option, the information gets leaked. In this work, we define and study behavioural side-channel leakage in browser-use agents, where an agent's actions inadvertently reveal private information (secret) retained from a prior website, despite an explicit instruction not to disclose it. We introduce AgentTell, a benchmark of 20 scenarios and 100 tasks in which an agent acquires a secret on one website and then completes a task on another website that offers secret-specific actions alongside a general action that reveals nothing. Our evaluation across 9,760 sessions on six backbones shows that agents carrying a secret reveal it through their actions in 61.1% of sessions. Even when agents explicitly state in memory that the secret must not be shared, they still reveal it in 56.7% of those sessions. Moreover, in 34.5% of leaking sessions, their final responses falsely assure users that the secret was not disclosed. These findings show that agents often fail to recognize side-channel leakage as a privacy risk.

The rise of on-chip accelerators signifies a major shift in computing, driven by the growing demands of artificial intelligence (AI) and specialized applications. These accelerators have gained popularity due to their ability to substantially boost performance, cut energy usage, lower total cost of ownership (TCO), and promote sustainability. Intel's Advanced Matrix Extensions (AMX) is one such on-chip accelerator, specifically designed for handling tasks involving large matrix multiplications commonly used in machine learning (ML) models, image processing, and other computational-heavy operations. In this paper, we introduce a novel value-dependent timing side-channel vulnerability in Intel AMX. By exploiting this weakness, we demonstrate a software-based, value-dependent timing side-channel attack capable of inferring the sparsity of neural network weights without requiring any knowledge of the confidence score, privileged access or physical proximity. Our attack method can fully recover the sparsity of weights assigned to 64 input elements within 50 minutes, which is 631% faster than the maximum leakage rate achieved in the Hertzbleed attack.

Agent memory is typically constructed either offline from curated demonstrations or online from post-deployment interactions. However, regardless of how it is built, an agent faces a cold-start gap when first introduced to a new environment without any task-specific experience available. In this paper, we study pre-task memory construction: whether an agent can build procedural memory before observing any target-environment tasks, using only self-generated synthetic practice. Yet, synthetic interaction alone is insufficient, as without controlling what to practice and what to store, synthetic tasks become redundant, infeasible, and ultimately uninformative, and memory further degrades quickly due to unfiltered trajectories. To overcome this, we present Preping, a proposer-guided memory construction framework. At its core is proposer memory, a structured control state that shapes future practice. A Proposer generates synthetic tasks conditioned on this state, a Solver executes them, and a Validator determines which trajectories are eligible for memory insertion while also providing feedback to guide future proposals. Experiments on AppWorld, BFCL v3, and MCP-Universe show that Preping substantially improves over a no-memory baseline and achieves performance competitive with strong playbook-based methods built from offline or online experience, with deployment cost 2.99times lower on AppWorld and 2.23times lower on BFCL v3 than online memory construction. Further analyses reveal that the main benefit does not come from synthetic volume alone, but from proposer-side control over feasibility, redundancy, and coverage, combined with selective memory updates.

Graph Neural Networks (GNNs) have achieved state-of-the-art performance in various graph-based learning tasks. However, enabling privacy-preserving GNNs in encrypted domains, such as under Fully Homomorphic Encryption (FHE), typically incurs substantial computational overhead, rendering real-time and privacy-preserving inference impractical. In this work, we propose DESIGN (EncrypteD GNN Inference via sErver-Side Input Graph pruNing), a novel framework for efficient encrypted GNN inference. DESIGN tackles the critical efficiency limitations of existing FHE GNN approaches, which often overlook input data redundancy and apply uniform computational strategies. Our framework achieves significant performance gains through a hierarchical optimization strategy executed entirely on the server: first, FHE-compatible node importance scores (based on encrypted degree statistics) are computed from the encrypted graph. These scores then guide a homomorphic partitioning process, generating multi-level importance masks directly under FHE. This dynamically generated mask facilitates both input graph pruning (by logically removing unimportant elements) and a novel adaptive polynomial activation scheme, where activation complexity is tailored to node importance levels. Empirical evaluations demonstrate that DESIGN substantially accelerates FHE GNN inference compared to state-of-the-art methods while maintaining competitive model accuracy, presenting a robust solution for secure graph analytics. Our implementation is publicly available at https://github.com/LabRAI/DESIGN.

Tool-using language model agents are usually evaluated by whether they finish the user's task. That metric misses a common failure: an agent can complete the requested work while also taking an unauthorized side-effecting action, such as writing to the wrong object, expanding the scope of the request, or contacting an unintended recipient. We call this failure mode collateral damage (CD). We introduce SABench (Side-effect-Aware Benchmark), a benchmark of 201 email, calendar, and ticket tasks. Each task crosses one of four state structures (independent, dependent, externally reachable, cross-tool) with one of four trap families (ambiguity, mis-targeting, cascade harm, overreach), and each task has a deterministic trace-based oracle for both goal satisfaction and authorization compliance. Across 2,211 model-task pairs from 11 frontier LLMs spanning major commercial and open model families, 15.2% of all runs incur CD, with risk concentrated in ambiguity tasks (38.8% violation rate) while non-ambiguous traps average 8.4% CD. Among goal-satisfied runs, 12.8% still produce unauthorized writes. Cross-tool and dependent states have about twice the CD rate of independent states. CD rates vary by 4.7x across models at comparable goal rates, which suggests that task success alone is a poor proxy for safe tool use. The SABench task corpus is released at https://huggingface.co/datasets/Phaedrus/sabench; the traces, oracle, code, and analysis pipeline are released at https://github.com/hjj194/sabench-artifact.

AI agents may be able to automate your inbox, but can they automate other routine aspects of your life? Everyday online tasks offer a realistic yet unsolved testbed for evaluating the next generation of AI agents. To this end, we introduce ClawBench, an evaluation framework of 153 simple tasks that people need to accomplish regularly in their lives and work, spanning 144 live platforms across 15 categories, from completing purchases and booking appointments to submitting job applications. These tasks require demanding capabilities beyond existing benchmarks, such as obtaining relevant information from user-provided documents, navigating multi-step workflows across diverse platforms, and write-heavy operations like filling in many detailed forms correctly. Unlike existing benchmarks that evaluate agents in offline sandboxes with static pages, ClawBench operates on production websites, preserving the full complexity, dynamic nature, and challenges of real-world web interaction. A lightweight interception layer captures and blocks only the final submission request, ensuring safe evaluation without real-world side effects. Our evaluations of 7 frontier models show that both proprietary and open-source models can complete only a small portion of these tasks. For example, Claude Sonnet 4.6 achieves only 33.3%. Progress on ClawBench brings us closer to AI agents that can function as reliable general-purpose assistants.

AI agents are reshaping the workspace, leading to drastic change of how humans work. Despite the considerable potential of human-agent collaboration both in preserving human agency and generating economic value, this paradigm remains largely absent from occupational task evaluation, hindered by the difficulty of gathering real human data and accounting for inter-human variability. We introduce CollabSkill, a framework for evaluating human-agent collaboration on real-world occupational tasks. CollabSkill pairs real human workers with AI agents on tasks matched to their occupational background, collecting data that capture the complexity of economically valuable tasks and the usage patterns of real workers. To account for inter-human variability, CollabSkill employs a Bayesian skill rating system to disentangle and quantify the skill contributions of both humans and AI agents. Drawing on over 1,500 prompts from 386 working sessions contributed by 93 human workers, our analysis yields insights on two fronts: on the agent side, rankings on CollabSkill diverge meaningfully from those of existing fully autonomous benchmarks where Codex leads, with Claude Code ranking first; on the human side, CollabSkill reveals that practical experience emerges as the primary driver of collaboration skill, with hands-on collaboration meaningfully shifting workers' AI literacy. Together, we hope CollabSkill enables the community to invest in systematic evaluation of human-agent collaboration and spurs development efforts aimed at building AI agents that genuinely augment human workers.

While recent 3D-aware generative models have shown photo-realistic image synthesis with multi-view consistency, the synthesized image quality degrades depending on the camera pose (e.g., a face with a blurry and noisy boundary at a side viewpoint). Such degradation is mainly caused by the difficulty of learning both pose consistency and photo-realism simultaneously from a dataset with heavily imbalanced poses. In this paper, we propose SideGAN, a novel 3D GAN training method to generate photo-realistic images irrespective of the camera pose, especially for faces of side-view angles. To ease the challenging problem of learning photo-realistic and pose-consistent image synthesis, we split the problem into two subproblems, each of which can be solved more easily. Specifically, we formulate the problem as a combination of two simple discrimination problems, one of which learns to discriminate whether a synthesized image looks real or not, and the other learns to discriminate whether a synthesized image agrees with the camera pose. Based on this, we propose a dual-branched discriminator with two discrimination branches. We also propose a pose-matching loss to learn the pose consistency of 3D GANs. In addition, we present a pose sampling strategy to increase learning opportunities for steep angles in a pose-imbalanced dataset. With extensive validation, we demonstrate that our approach enables 3D GANs to generate high-quality geometries and photo-realistic images irrespective of the camera pose.

Speech large language models (SLMs) are typically built from text large language model (TLM) checkpoints, yet they still suffer from a substantial modality gap. Prior work has mainly attempted to reduce this gap from the output side by making speech generation more text-like, but the gap remains. We argue that the key remaining bottleneck lies on the input side. We propose TextPro-SLM, an SLM that makes spoken input more closely resemble that of a prosody-aware text LLM. TextPro-SLM combines WhisperPro, a unified speech encoder that produces synchronized text tokens and prosody embeddings, with an LLM backbone trained to preserve the semantic capabilities of the original TLM while learning paralinguistic understanding. Experiments show that TextPro-SLM achieves the lowest modality gap among leading SLMs at both 3B and 7B scales, while also delivering strong overall performance on paralinguistic understanding tasks. These gains are achieved with only roughly 1,000 hours of LLM training audio, suggesting that reducing the modality gap from the input side is both effective and data-efficient. Code is available at https://github.com/dreamtheater123/TextPro-SLM

Modern AI agents, driven by advances in large foundation models, promise to enhance our productivity and transform our lives by augmenting our knowledge and capabilities. To achieve this vision, AI agents must effectively plan, perform multi-step reasoning and actions, respond to novel observations, and recover from errors, to successfully complete complex tasks across a wide range of scenarios. In this work, we introduce Magentic-One, a high-performing open-source agentic system for solving such tasks. Magentic-One uses a multi-agent architecture where a lead agent, the Orchestrator, plans, tracks progress, and re-plans to recover from errors. Throughout task execution, the Orchestrator directs other specialized agents to perform tasks as needed, such as operating a web browser, navigating local files, or writing and executing Python code. We show that Magentic-One achieves statistically competitive performance to the state-of-the-art on three diverse and challenging agentic benchmarks: GAIA, AssistantBench, and WebArena. Magentic-One achieves these results without modification to core agent capabilities or to how they collaborate, demonstrating progress towards generalist agentic systems. Moreover, Magentic-One's modular design allows agents to be added or removed from the team without additional prompt tuning or training, easing development and making it extensible to future scenarios. We provide an open-source implementation of Magentic-One, and we include AutoGenBench, a standalone tool for agentic evaluation. AutoGenBench provides built-in controls for repetition and isolation to run agentic benchmarks in a rigorous and contained manner -- which is important when agents' actions have side-effects. Magentic-One, AutoGenBench and detailed empirical performance evaluations of Magentic-One, including ablations and error analysis are available at https://aka.ms/magentic-one

In recent years, while cloud-based MLLMs such as QwenVL, InternVL, GPT-4o, Gemini, and Claude Sonnet have demonstrated outstanding performance with enormous model sizes reaching hundreds of billions of parameters, they significantly surpass the limitations in memory, power consumption, and computing capacity of edge devices such as mobile phones. This paper introduces AndesVL, a suite of mobile-side MLLMs with 0.6B to 4B parameters based on Qwen3's LLM and various visual encoders. We comprehensively outline the model architectures, training pipeline, and training data of AndesVL, which achieves first-tier performance across a wide range of open-source benchmarks, including fields such as text-rich image understanding, reasoning and math, multi-image comprehension, general VQA, hallucination mitigation, multilingual understanding, and GUI-related tasks when compared with state-of-the-art models of a similar scale. Furthermore, we introduce a 1+N LoR

We introduce Crypto Accounting Bench (CAB), a benchmark for assessing whether frontier and open-weight language models can reconstruct the complete journal entry that an organization actually posted for a crypto-asset transaction. CAB contains 118 evaluation tasks drawn from 7 pseudonymized organizations. Each task combines transaction mechanics, asset quantities and base-currency values, wallet and legal-entity context, counterparty evidence, related transaction legs, recurrence, tax-lot evidence, and the organization's complete chart of accounts. The target is a balanced structured entry with every required account, side, amount, currency, and full-precision asset quantity. We evaluate 12 models spanning proprietary frontier systems and open-weight releases over 3 independent attempts per task, producing 4,248 trajectories. We report 3 metrics: Mean Score, Best@3, and Pass@3. Pass@3 is the fraction of tasks with at least 1 of 3 attempts that satisfies every rubric criterion and required gate. The leading model reaches 77.43% Mean Score, while the best Pass@3 is 56.78%. Deterministic diagnostics, read from each task's best of 3 attempts and macro-averaged across the 12 models, show higher base-amount agreement (97.8%) than deciding-account accuracy (56.3%). Together with the failure analysis, these results identify account selection and complete-entry composition as the main remaining challenges on CAB.

Computer-use agents are capable of completing complex tasks, increasing the use of automatic judges to determine success, either for training or for evaluation without human involvement. Despite their flexibility, their reliability on long tasks spanning multiple applications remains unclear. A trajectory, composed of long sequences of screenshots and actions, may appear complete, but in reality violates constraints from the instruction or introduces an unwanted side effect. To identify these errors, a judge needs to examine the trajectory with respect to the user's instruction. To this end, we introduce AgentHorizon, a benchmark of 1,373 computer-use tasks (instruction-trajectory pairs) drawn from 166 hours of human-recorded trajectories spanning three operating systems. By recording trajectories for closely related instructions, we can construct negative tasks by swapping the instructions. This paired design evaluates judges on their ability to distinguish a successful trajectory from one that completed a similar (but incompatible) request. We release the benchmark under three splits: a frontier split, AgentHorizon (AH), a simplified split, AgentHorizon-Simple (AH-S), and a development split, AgentHorizon-Development (AH-D). We further evaluate eleven judges by (1) directly passing the full trajectory (with up to 300 screenshots and actions), and (2) using them as coding agents across five agent harnesses. We find that our best agentic judge, GPT-5.5, achieves 80.9% balanced accuracy on the AH subset. We find that tool-use improves certain models but results in worse performance for open-weight models, and that judges differ drastically in their ability to accept a valid trajectory and reject failed ones. Our findings highlight the need for judges that are capable of locating and verifying often hidden evidence that a task was properly completed inside long interaction histories.

Sign Language Recognition (SLR) involves the automatic identification and classification of sign gestures from images or video, converting them into text or speech to improve accessibility for the hearing-impaired community. In Bangladesh, Bangla Sign Language (BdSL) serves as the primary mode of communication for many individuals with hearing impairments. This study fine-tunes state-of-the-art video transformer architectures -- VideoMAE, ViViT, and TimeSformer -- on BdSLW60 (arXiv:2402.08635), a small-scale BdSL dataset with 60 frequent signs. We standardized the videos to 30 FPS, resulting in 9,307 user trial clips. To evaluate scalability and robustness, the models were also fine-tuned on BdSLW401 (arXiv:2503.02360), a large-scale dataset with 401 sign classes. Additionally, we benchmark performance against public datasets, including LSA64 and WLASL. Data augmentation techniques such as random cropping, horizontal flipping, and short-side scaling were applied to improve model robustness. To ensure balanced evaluation across folds during model selection, we employed 10-fold stratified cross-validation on the training set, while signer-independent evaluation was carried out using held-out test data from unseen users U4 and U8. Results show that video transformer models significantly outperform traditional machine learning and deep learning approaches. Performance is influenced by factors such as dataset size, video quality, frame distribution, frame rate, and model architecture. Among the models, the VideoMAE variant (MCG-NJU/videomae-base-finetuned-kinetics) achieved the highest accuracies of 95.5% on the frame rate corrected BdSLW60 dataset and 81.04% on the front-facing signs of BdSLW401 -- demonstrating strong potential for scalable and accurate BdSL recognition.

Chain-of-Thought (CoT) has become a standard method for improving reasoning capabilities in large language models (LLMs) by eliciting step-by-step thinking, but its effectiveness in multimodal tasks remains unclear. In this paper, we aim to systematically investigate the key question: What can multimodal Chain-of-Thought reasoning do, and where and why does it fall short? To this end, we evaluate 12 multimodal tasks across perception and reasoning categories using both 14 non-reasoning models and 8 reasoning models. Our analysis reveals several important findings: (1) CoT is not a free lunch and should be used selectively depending on the specific requirements of each task. For perception tasks, CoT can lead to undesirable side effects, such as reduced performance in visual grounding and object counting. In contrast, it proves effective for reasoning tasks involving mathematical, scientific, and multi-image reasoning; (2) Compared to original models, existing open-source multimodal reasoning models often yield only marginal overall improvements, possibly due to an overemphasis on mathematical reasoning at the expense of broader capabilities; (3) Visual reasoning remains a key bottleneck for current multimodal CoT, as models exhibit a Look Light, Think Heavy pattern where verbal reflection rises and falls during reasoning, whereas visual reflection consistently diminishes. These findings suggest that while multimodal CoT handles verbal reflection relatively well, it lacks the ability to maintain deep visual introspection throughout the reasoning process.

Negation is a fundamental linguistic phenomenon that poses persistent challenges for Large Language Models (LLMs), particularly in tasks requiring deep semantic understanding. Existing benchmarks often treat negation as a side case within broader tasks like natural language inference, resulting in a lack of benchmarks that exclusively target negation understanding. In this work, we introduce Thunder-NUBench, a novel benchmark explicitly designed to assess sentence-level negation understanding in LLMs. Thunder-NUBench goes beyond surface-level cue detection by contrasting standard negation with structurally diverse alternatives such as local negation, contradiction, and paraphrase. The benchmark consists of manually curated sentence-negation pairs and a multiple-choice dataset that enables in-depth evaluation of models' negation understanding.

There is considerable confusion about the role of Large Language Models (LLMs) in planning and reasoning tasks. On one side are over-optimistic claims that LLMs can indeed do these tasks with just the right prompting or self-verification strategies. On the other side are perhaps over-pessimistic claims that all that LLMs are good for in planning/reasoning tasks are as mere translators of the problem specification from one syntactic format to another, and ship the problem off to external symbolic solvers. In this position paper, we take the view that both these extremes are misguided. We argue that auto-regressive LLMs cannot, by themselves, do planning or self-verification (which is after all a form of reasoning), and shed some light on the reasons for misunderstandings in the literature. We will also argue that LLMs should be viewed as universal approximate knowledge sources that have much more meaningful roles to play in planning/reasoning tasks beyond simple front-end/back-end format translators. We present a vision of {\bf LLM-Modulo Frameworks} that combine the strengths of LLMs with external model-based verifiers in a tighter bi-directional interaction regime. We will show how the models driving the external verifiers themselves can be acquired with the help of LLMs. We will also argue that rather than simply pipelining LLMs and symbolic components, this LLM-Modulo Framework provides a better neuro-symbolic approach that offers tighter integration between LLMs and symbolic components, and allows extending the scope of model-based planning/reasoning regimes towards more flexible knowledge, problem and preference specifications.

Continual-learning agents are systems of models, harnesses, and memory operating over long multi-session horizons. Evaluating and training them requires interleaving tasks with agent-side events such as session stop and start, crons, and memory consolidation. Yet existing benchmarks and training frameworks schedule only the benchmark's own events, leaving each benchmark and agent pair to build a custom scheduling loop. We present SCLATE, an execution substrate where benchmarks and unmodified agents each add their events to one open event scheduler through an adapter. A hybrid simulated clock runs these events on a shared timeline, flowing in real time while the agent works and skipping idle gaps, which compresses a month-long scenario into hours. SCLATE also serves as a rollout engine that runs any agent's harness and memory unmodified, recording the tokens and log probabilities of every model call through an in-container proxy. We port seven benchmarks to SCLATE and compare ten unmodified harness and memory configurations head to head on ten models. The comparison shows that an added memory system does not reliably beat the harness's native memory and that models differ widely in how they use the same harness and memory. We then post-train Qwen3.5-4B through unmodified harnesses and memory systems. The model learns to use both, reading 6.8x fewer file lines with a 16.7-point higher SWE-bench Verified pass rate, and writing richer memory records, while its held-out MetaClaw accuracy rises by up to 11.8 points.

Existing financial NLP benchmarks often rely on labels supplied by outside observers, measuring how language is perceived rather than what speakers have committed to in the market. We introduce StakeBench, an evaluation framework for language understanding grounded in market commitment. StakeBench links 560,876 comments from 2,261 resolved markets to verified position, action, and market-odds records across Polymarket and Manifold. Supervision is derived from observable market behavior. Position sides, post-comment trading actions, and market-odds trajectories replace human annotation. Four diagnostic tasks test whether models detect market commitment, identify the revealed side, anticipate future action, and perform collective odds projection. Three commitment-aware metrics measure alignment with revealed preferences rather than perceived sentiment. Validity audits and explicit interpretation boundaries help distinguish observable commitment signals from latent belief and causal market-odds impact. Across 15 LLMs and 18 topics and platform settings, models partially recover position-side signals, with Directed Accuracy from 0.506 to 0.599, but show structural failures on later tasks. Ten of the fifteen models collapse to one or two action labels in future action anticipation, and no model consistently improves on the naive odds-direction baseline in collective odds projection. Model scale is not correlated with performance, finance-domain tuning does not improve revealed-side identification, and platform incentives strongly shape higher-order results. StakeBench is packaged with evaluation code and dataset under CC-BY 4.0.

Multimodal Large Language Models (MLLMs) achieve stronger visual understanding by scaling input fidelity, yet the resulting visual token growth makes jointly sustaining high spatial resolution and long temporal context prohibitive. We argue that the bottleneck lies not in how post-encoding representations are compressed but in the volume of pixels the encoder receives, and address it with ResAdapt, an Input-side adaptation framework that learns how much visual budget each frame should receive before encoding. ResAdapt couples a lightweight Allocator with an unchanged MLLM backbone, so the backbone retains its native visual-token interface while receiving an operator-transformed input. We formulate allocation as a contextual bandit and train the Allocator with Cost-Aware Policy Optimization (CAPO), which converts sparse rollout feedback into a stable accuracy-cost learning signal. Across budget-controlled video QA, temporal grounding, and image reasoning tasks, ResAdapt improves low-budget operating points and often lies on or near the efficiency-accuracy frontier, with the clearest gains on reasoning-intensive benchmarks under aggressive compression. Notably, ResAdapt supports up to 16x more frames at the same visual budget while delivering over 15% performance gain. Code is available at https://github.com/Xnhyacinth/ResAdapt.

Phone agents are increasingly expected to complete real mobile workflows rather than merely predict the next screen action. However, much of the current mobile-agent literature still evaluates agents primarily as GUI controllers that observe a screen, emit taps and swipes, and are scored by target app state. Real phone-use tasks are broader: they require deciding when to use app GUIs, device-side commands, or structured tools, while leaving evidence that the intended side effect actually occurred. We introduce PhoneHarness, a mixed-action benchmark and execution harness for studying phone-use agents on verifiable mobile workflows. PhoneHarness runs a device-side agent loop over GUI, CLI, and host-side tool actions, combining deterministic action routing with bounded GUI delegation and auditable execution traces. Its benchmark, PhoneHarness Bench, evaluates whether agents complete tasks with observable side effects, not only whether they produce plausible final answers. On the annotated evaluation split, PhoneHarness reaches a 75.0% pass rate, outperforming the strongest non-PhoneHarness settings by 12.9 percentage points. PhoneHarness and PhoneHarness Bench therefore play distinct but mutually dependent roles: the harness makes mixed phone workflows executable, while the benchmark measures whether agents can use that harness reliably and safely. Our findings suggest that reliable phone automation depends on action-surface routing and verifiable execution, not only visual GUI control.

Generative AI and coding agents are intensifying a central governance tension in open-source software (OSS): they scale contribution generation faster than maintainers can assess risk, evidence, and accountability. Existing responses improve agent-readability and traceability, but project rules must also organize contribution-specific risk, evidence, accountability, and review-gate states. We theorize this organizational arrangement as project-side governability infrastructure. A diagnostic audit of 50 GitHub repositories finds widespread general governance artifacts, observable agent-readability, and fragmented AI-governance cues, but no project-wide arrangement that coordinates shared rules, preparation obligations, verification rights, and maintainer decision authority across AI-mediated contribution workflows. We develop the Agent Governance Manifest (AGM) as a repository-hosted boundary resource and bidirectional governance contract linking contributor-side evidence preparation with maintainer-side verification. In a controlled reviewer-side evaluation with 15 participants and 75 task-level outputs, AGM-supported materials improved exact risk-label recovery (37/38 vs. 15/37) and perceived review support (6.14 vs. 3.27 on a 1-7 scale). In a contributor-side feasibility check, 15 participants completed 45 tasks; all final packages represented the core governance state correctly, and 41 passed strict structural validation. The study develops a three-layer framework of agent-readability, traceability, and governability, theorizes agent-mediated contributions as governable boundary objects, and advances compliance-enabling digital innovation governance while preserving maintainer decision authority.

Recent Pre-trained Language Models (PLMs) usually only provide users with the inference APIs, namely the emerging Model-as-a-Service (MaaS) setting. To adapt MaaS PLMs to downstream tasks without accessing their parameters and gradients, some existing methods focus on the output-side adaptation of PLMs, viewing the PLM as an encoder and then optimizing a task-specific decoder for decoding the output hidden states and class scores of the PLM. Despite the effectiveness of these methods, they only use a single prompt to query PLMs for decoding, leading to a heavy reliance on the quality of the adopted prompt. In this paper, we propose a simple yet effective Multi-Prompting Decoder (MPD) framework for MaaS adaptation. The core idea is to query PLMs with multiple different prompts for each sample, thereby obtaining multiple output hidden states and class scores for subsequent decoding. Such multi-prompting decoding paradigm can simultaneously mitigate reliance on the quality of a single prompt, alleviate the issue of data scarcity under the few-shot setting, and provide richer knowledge extracted from PLMs. Specifically, we propose two decoding strategies: multi-prompting decoding with optimal transport for hidden states and calibrated decoding for class scores. Extensive experiments demonstrate that our method achieves new state-of-the-art results on multiple natural language understanding datasets under the few-shot setting.

Group Relative Policy Optimization (GRPO) is widely used for critic-free Large Language Model (LLM) post-training, but its KL regularization is usually implemented as a local loss-side token penalty. We show that this misses the policy-gradient signal induced by autoregressive KL regularization. Unlike standard KL-regularized Reinforcement Learning (RL) objectives, GRPO's group normalization induces a non-linear prompt-level utility; for binary verifier rewards, this utility is 2arcsinsqrt p. As a result, reward and KL cannot be fused before normalization without changing the implicit objective. We derive the on-policy gradient of GRPO-style objectives with token-wise f-divergence regularization. The reward term recovers the standardized GRPO advantage, while the regularizer term includes a causal future-regularization return-to-go omitted by local KL losses. For reverse KL, this yields a simple future KL correction: add a reverse cumulative sum of per-token log ratios after advantage construction. The resulting method, Future-KL Regularized Policy Optimization (FRPO), requires no critic or extra model passes. On mathematical reasoning tasks, FRPO improves pass@16 in our main large-model setting while maintaining higher entropy and lower policy drift than conventional loss-side KL baselines.

SkyReels V4 is a unified multi modal video foundation model for joint video audio generation, inpainting, and editing. The model adopts a dual stream Multimodal Diffusion Transformer (MMDiT) architecture, where one branch synthesizes video and the other generates temporally aligned audio, while sharing a powerful text encoder based on the Multimodal Large Language Models (MMLM). SkyReels V4 accepts rich multi modal instructions, including text, images, video clips, masks, and audio references. By combining the MMLMs multi modal instruction following capability with in context learning in the video branch MMDiT, the model can inject fine grained visual guidance under complex conditioning, while the audio branch MMDiT simultaneously leverages audio references to guide sound generation. On the video side, we adopt a channel concatenation formulation that unifies a wide range of inpainting style tasks, such as image to video, video extension, and video editing under a single interface, and naturally extends to vision referenced inpainting and editing via multi modal prompts. SkyReels V4 supports up to 1080p resolution, 32 FPS, and 15 second duration, enabling high fidelity, multi shot, cinema level video generation with synchronized audio. To make such high resolution, long-duration generation computationally feasible, we introduce an efficiency strategy: Joint generation of low resolution full sequences and high-resolution keyframes, followed by dedicated super-resolution and frame interpolation models. To our knowledge, SkyReels V4 is the first video foundation model that simultaneously supports multi-modal input, joint video audio generation, and a unified treatment of generation, inpainting, and editing, while maintaining strong efficiency and quality at cinematic resolutions and durations.

As Large Language Models (LLMs) are increasingly deployed as autonomous agents in complex and long horizon settings, it is critical to evaluate their ability to sabotage users by pursuing hidden objectives. We study the ability of frontier LLMs to evade monitoring and achieve harmful hidden goals while completing a wide array of realistic tasks. We evaluate a broad range of frontier LLMs using SHADE (Subtle Harmful Agent Detection & Evaluation)-Arena, the first highly diverse agent evaluation dataset for sabotage and monitoring capabilities of LLM agents. SHADE-Arena consists of complex pairs of benign main tasks and harmful side objectives in complicated environments. Agents are evaluated on their ability to complete the side task without appearing suspicious to an LLM monitor. When measuring agent ability to (a) complete the main task, (b) complete the side task, and (c) avoid detection, we find that the best performing frontier models score 27% (Claude 3.7 Sonnet) and 15% (Gemini 2.5 Pro) as sabotage agents when overseen by Claude 3.6 Sonnet. For current frontier models, success on the side task relies heavily on having access to a hidden scratchpad that is not visible to the monitor. We also use SHADE-Arena to measure models' monitoring abilities, with the top monitor (Gemini 2.5 Pro) achieving an AUC of 0.87 at distinguishing benign and malign transcripts. We find that for now, models still struggle at sabotage due to failures in long-context main task execution. However, our measurements already demonstrate the difficulty of monitoring for subtle sabotage attempts, which we expect to only increase in the face of more complex and longer-horizon tasks.

Vision-Language Models (VLMs) are increasingly deployed in embodied environments, where they need produce numerical outputs such as action magnitudes and spatial coordinates. Although these numbers appear meaningful, it remains unclear whether these numerical outputs are genuinely grounded in spatial perception. Therefore, in this work, we revisit spatial numerical understanding through SpaceNum, a unified framework that captures two complementary settings: numbers as dynamic transitions during spatial exploration, and numbers as static layouts in spatial reasoning. We formulate two bidirectional tasks, Num2Space and Space2Num, to evaluate how well VLMs map between vision-side spatial structure and language-side numerical representations. We systematically study whether current VLMs truly understand numerical values in spatial settings. Across dynamic transitions and static layouts, we find that models largely fail to ground numbers in spatial meaning and often perform close to random guess. Through error analysis, reasoning trace analysis, and controlled interventions, we show that current VLMs rely heavily on shallow spatial cues, struggle to build stable coordinate-aware representations, and fail to abstract structured spatial layouts from visual observations. We further show that explicit reasoning provides only marginal gains, while tuning can partially improve spatial numerical understanding and transfer to external spatial reasoning benchmarks.

The prevalent dual-branch paradigm, i.e., training a side network to encode visual conditions and fusing its intermediate-layer features to a frozen pretrained main network, has shown remarkable success in visual-condition controllable generation. Despite its widespread adoption, the role of the side branch and its training efficiency remain underexplored. In this paper, we first revisit this mainstream paradigm through the lens of score-based generative modeling: 1) The main network preserves visual perceptual quality by providing a prior unconditional score. 2) The side network steers conditional control by implicitly contributing a likelihood score. Guided by this perspective, we propose LIkelihood Score Alignment (LISA), an effective regularization method that explicitly aligns the intermediate feature of the side network with an approximated likelihood score. Specifically, we first hook features from a designated layer of the side network and project them into the score latent space by a lightweight decoder. Then, we construct an approximated likelihood score target and calculate the distance between the decoder's output and this target as an additional regularization loss. Finally, we jointly optimize the side network and decoder with both standard diffusion loss and our regularization loss. Experiments across various image/video tasks, architectures, and diffusion/flow models demonstrated that LISA can not only consistently accelerate the training convergence and improve final synthetic results, but also encourage the side network's features to be more disentangled for conditional modeling with negligible additional training cost and zero extra inference cost.

Attention based models such as Transformers involve pairwise interactions between data points, modeled with a learnable attention matrix. Importantly, this attention matrix is normalized with the SoftMax operator, which makes it row-wise stochastic. In this paper, we propose instead to use Sinkhorn's algorithm to make attention matrices doubly stochastic. We call the resulting model a Sinkformer. We show that the row-wise stochastic attention matrices in classical Transformers get close to doubly stochastic matrices as the number of epochs increases, justifying the use of Sinkhorn normalization as an informative prior. On the theoretical side, we show that, unlike the SoftMax operation, this normalization makes it possible to understand the iterations of self-attention modules as a discretized gradient-flow for the Wasserstein metric. We also show in the infinite number of samples limit that, when rescaling both attention matrices and depth, Sinkformers operate a heat diffusion. On the experimental side, we show that Sinkformers enhance model accuracy in vision and natural language processing tasks. In particular, on 3D shapes classification, Sinkformers lead to a significant improvement.

Large Language Models (LLMs) are increasingly applied to data-intensive workflows, from database querying to developer observability. Yet the effectiveness of these systems is constrained by the volume, verbosity, and noise of real-world text-rich data such as logs, telemetry, and monitoring streams. Feeding such data directly into LLMs is costly, environmentally unsustainable, and often misaligned with task objectives. Parallel efforts in LLM efficiency have focused on model- or architecture-level optimizations, but the challenge of reducing upstream input verbosity remains underexplored. In this paper, we argue for treating the token budget of an LLM as an attention budget and elevating task-aware text reduction as a first-class design principle for language -- data systems. We position input-side reduction not as compression, but as attention allocation: prioritizing information most relevant to downstream tasks. We outline open research challenges for building benchmarks, designing adaptive reduction pipelines, and integrating token-budget--aware preprocessing into database and retrieval systems. Our vision is to channel scarce attention resources toward meaningful signals in noisy, data-intensive workflows, enabling scalable, accurate, and sustainable LLM--data integration.

Large language models (LLMs) have increasingly been explored as powerful text embedders. Existing LLM-based text embedding approaches often leverage the embedding of the final token, typically a reserved special token such as [EOS]. However, these tokens have not been intentionally trained to capture the semantics of the whole context, limiting their capacity as text embeddings, especially for retrieval and re-ranking tasks. We propose to add a new training stage before contrastive learning to enrich the semantics of the final token embedding. This stage employs bidirectional generative reconstruction tasks, namely EBQ2D (Embedding-Based Query-to-Document) and EBD2Q (Embedding-Based Document-to-Query), which interleave to anchor the [EOS] embedding and reconstruct either side of Query-Document pairs. Experimental results demonstrate that our additional training stage significantly improves LLM performance on the Massive Text Embedding Benchmark (MTEB), achieving new state-of-the-art results across different LLM base models and scales.

Command-line interfaces (CLIs) provide structured textual environments for system administration. Explorations have been performed using pre-trained language models (PLMs) to simulate these environments for safe interaction in high-risk environments. However, their use has been constrained to frozen, large parameter models like GPT. For smaller architectures to reach a similar level of believability, a rich dataset of CLI interactions is required. Existing public datasets focus on mapping natural-language tasks to commands, omitting crucial execution data such as exit codes, outputs, and environmental side effects, limiting their usability for behavioral modeling. We introduce a Shell Input -Output Environment (ShIOEnv), which casts command construction as a Markov Decision Process whose state is the partially built sequence and whose actions append arguments. After each action, ShIOEnv executes the candidate and returns its exit status, output, and progress toward a minimal-length behavioral objective. Due to the intractable nature of the combinatorial argument state-action space, we derive a context-free grammar from man pages to mask invalid arguments from being emitted. We explore random and proximal-policy optimization (PPO)-optimized sampling of unrestricted and grammar-masked action spaces to produce four exploration strategies. We observed that grammar masking and PPO significantly improve sample efficiency to produce a higher quality dataset (maximizing the number of arguments while minimizing redundancies). Policy-generated datasets of shell input-output behavior pairs are used to fine-tune CodeT5, where we observe 85% improvements in BLEU-4 when constraining the action space to grammar productions with an additional 26% improvement when applying PPO. The ShIOEnv environment and curated command behavior datasets are released for use in future research.

In this paper, we explore a less-studied yet practically important problem: how to efficiently and effectively adapt multiple (>2) multimodal foundation models (MFMs) for the sequential recommendation task. To this end, we propose a plug-and-play Cross-modal Side Adapter Network (CROSSAN), which leverages a fully decoupled side adapter-based paradigm to achieve efficient and scalable adaptation. Compared to the state-of-the-art efficient approaches, CROSSAN reduces training time by over 30%, GPU memory consumption by 20%, and trainable parameters by over 57%, while enabling effective cross-modal learning across diverse modalities. To further enhance multimodal fusion, we introduce the Mixture of Modality Expert Fusion (MOMEF) mechanism. Extensive experiments on public benchmarks demonstrate that CROSSAN consistently outperforms existing methods, achieving 6.7%--8.1% performance improvements when adapting four foundation models with raw modalities. Moreover, the overall performance continues to improve as more MFMs are incorporated. We will release our code and datasets to faciliate future research.

Can a single LLM-based optimization system match specialized tools across fundamentally different domains? We show that when optimization problems are formulated as improving a text artifact evaluated by a scoring function, a single AI-based optimization system-supporting single-task search, multi-task search with cross-problem transfer, and generalization to unseen inputs-achieves state-of-the-art results across six diverse tasks. Our system discovers agent architectures that nearly triple Gemini Flash's ARC-AGI accuracy (32.5% to 89.5%), finds scheduling algorithms that cut cloud costs by 40%, generates CUDA kernels where 87% match or beat PyTorch, and outperforms AlphaEvolve's reported circle packing solution (n=26). Ablations across three domains reveal that actionable side information yields faster convergence and substantially higher final scores than score-only feedback, and that multi-task search outperforms independent optimization given equivalent per-problem budget through cross-task transfer, with benefits scaling with the number of related tasks. Together, we show for the first time that text optimization with LLM-based search is a general-purpose problem-solving paradigm, unifying tasks traditionally requiring domain-specific algorithms under a single framework. We open-source optimize\_anything with support for multiple backends as part of the GEPA project at https://github.com/gepa-ai/gepa .

Large language models (LLMs) with decoder-only architectures demonstrate remarkable in-context learning (ICL) capabilities. This feature enables them to effectively handle both familiar and novel tasks by utilizing examples provided within their input context. Recognizing the potential of this capability, we propose leveraging the ICL feature in LLMs to enhance the process of text embedding generation. To this end, we introduce a novel model bge-en-icl, which employs few-shot examples to produce high-quality text embeddings. Our approach integrates task-related examples directly into the query side, resulting in significant improvements across various tasks. Additionally, we have investigated how to effectively utilize LLMs as embedding models, including various attention mechanisms, pooling methods, etc. Our findings suggest that retaining the original framework often yields the best results, underscoring that simplicity is best. Experimental results on the MTEB and AIR-Bench benchmarks demonstrate that our approach sets new state-of-the-art (SOTA) performance. Our model, code and dataset are freely available at https://github.com/FlagOpen/FlagEmbedding .

Large language model agents increasingly improve by distilling experience into reusable skills, but errors in that experience can persist in learned rules and affect later tasks. Existing evaluations largely examine clean learning or perturbations encountered after skills are frozen, leaving robustness during self-evolution insufficiently understood. We introduce RSE-Bench, to our knowledge the first benchmark for evaluating skill self-evolution under ordinary, non-adversarial noise. It spans four domains and six benchmarks, with controlled perturbations to instructions, environmental evidence, execution records, and outcome--record pairings that preserve the underlying tasks and native scoring rules. The protocol separates noise during learning from input-side noise at evaluation, testing frozen skill stores on held-out tasks. We also introduce RSEClaw, a method-agnostic framework supporting consistent noise injection and process diagnostics of actual exposure and skill updates. Across three methods and two backbones, noisy learning retains only 25.5%--42.1% of the aggregate gain achieved by clean self-evolution, depending on the noise type. Perturbations confined to execution records and outcome--record pairings cause average losses comparable to input-side noise, despite leaving the original executions and evaluations intact. Diagnostics further identify differences in candidate skill text that coexist with small changes in mean held-out performance.

Large language models are shifting from passive information providers to active agents intended for complex workflows. However, their deployment as reliable AI workers in enterprise is stalled by benchmarks that fail to capture the intricacies of professional environments, specifically, the need for long-horizon planning amidst persistent state changes and strict access protocols. In this work, we introduce EnterpriseOps-Gym, a benchmark designed to evaluate agentic planning in realistic enterprise settings. Specifically, EnterpriseOps-Gym features a containerized sandbox with 164 database tables and 512 functional tools to mimic real-world search friction. Within this environment, agents are evaluated on 1,150 expert-curated tasks across eight mission-critical verticals (including Customer Service, HR, and IT). Our evaluation of 14 frontier models reveals critical limitations in state-of-the-art models: the top-performing Claude Opus 4.5 achieves only 37.4% success. Further analysis shows that providing oracle human plans improves performance by 14-35 percentage points, pinpointing strategic reasoning as the primary bottleneck. Additionally, agents frequently fail to refuse infeasible tasks (best model achieves 53.9%), leading to unintended and potentially harmful side effects. Our findings underscore that current agents are not yet ready for autonomous enterprise deployment. More broadly, EnterpriseOps-Gym provides a concrete testbed to advance the robustness of agentic planning in professional workflows.

Peptides, short chains of amino acid residues, play a vital role in numerous biological processes by interacting with other target molecules, offering substantial potential in drug discovery. In this work, we present PepFlow, the first multi-modal deep generative model grounded in the flow-matching framework for the design of full-atom peptides that target specific protein receptors. Drawing inspiration from the crucial roles of residue backbone orientations and side-chain dynamics in protein-peptide interactions, we characterize the peptide structure using rigid backbone frames within the SE(3) manifold and side-chain angles on high-dimensional tori. Furthermore, we represent discrete residue types in the peptide sequence as categorical distributions on the probability simplex. By learning the joint distributions of each modality using derived flows and vector fields on corresponding manifolds, our method excels in the fine-grained design of full-atom peptides. Harnessing the multi-modal paradigm, our approach adeptly tackles various tasks such as fix-backbone sequence design and side-chain packing through partial sampling. Through meticulously crafted experiments, we demonstrate that PepFlow exhibits superior performance in comprehensive benchmarks, highlighting its significant potential in computational peptide design and analysis.

Recent advances have highlighted the benefits of scaling language models to enhance performance across a wide range of NLP tasks. However, these approaches still face limitations in effectiveness and efficiency when applied to domain-specific tasks, particularly for small edge-side models. We propose the LoRA-Gen framework, which utilizes a large cloud-side model to generate LoRA parameters for edge-side models based on task descriptions. By employing the reparameterization technique, we merge the LoRA parameters into the edge-side model to achieve flexible specialization. Our method facilitates knowledge transfer between models while significantly improving the inference efficiency of the specialized model by reducing the input context length. Without specialized training, LoRA-Gen outperforms conventional LoRA fine-tuning, which achieves competitive accuracy and a 2.1x speedup with TinyLLaMA-1.1B in reasoning tasks. Besides, our method delivers a compression ratio of 10.1x with Gemma-2B on intelligent agent tasks.

Wearable devices such as AI glasses are transforming voice assistants into always-available, hands-free collaborators that integrate seamlessly with daily life, but they also introduce challenges like egocentric audio affected by motion and noise, rapid micro-interactions, and the need to distinguish device-directed speech from background conversations. Existing benchmarks largely overlook these complexities, focusing instead on clean or generic conversational audio. To bridge this gap, we present WearVox, the first benchmark designed to rigorously evaluate voice assistants in realistic wearable scenarios. WearVox comprises 3,842 multi-channel, egocentric audio recordings collected via AI glasses across five diverse tasks including Search-Grounded QA, Closed-Book QA, Side-Talk Rejection, Tool Calling, and Speech Translation, spanning a wide range of indoor and outdoor environments and acoustic conditions. Each recording is accompanied by rich metadata, enabling nuanced analysis of model performance under real-world constraints. We benchmark leading proprietary and open-source speech Large Language Models (SLLMs) and find that most real-time SLLMs achieve accuracies on WearVox ranging from 29% to 59%, with substantial performance degradation on noisy outdoor audio, underscoring the difficulty and realism of the benchmark. Additionally, we conduct a case study with two new SLLMs that perform inference with single-channel and multi-channel audio, demonstrating that multi-channel audio inputs significantly enhance model robustness to environmental noise and improve discrimination between device-directed and background speech. Our results highlight the critical importance of spatial audio cues for context-aware voice assistants and establish WearVox as a comprehensive testbed for advancing wearable voice AI research.

LLM agents call tools, query databases, delegate tasks, and trigger external side effects. Once an agent system can act in the world, the question is no longer only whether harmful actions can be prevented--it is whether those actions remain answerable after deployment. We distinguish accountability (the ability to determine compliance and assign responsibility), auditability (the system property that makes accountability possible), and auditing (the process of reconstructing behavior from trustworthy evidence). Our claim is direct: no agent system can be accountable without auditability. To make this operational, we define five dimensions of agent auditability, i.e., action recoverability, lifecycle coverage, policy checkability, responsibility attribution, and evidence integrity, and identify three mechanism classes (detect, enforce, recover) whose temporal information-and-intervention constraints explain why, in practice, no single approach suffices. We support the position with layered evidence rather than a single benchmark: lower-bound ecosystem measurements suggest that even basic security prerequisites for auditability are widely unmet (617 security findings across six prominent open-source projects); runtime feasibility results show that pre-execution mediation with tamper-evident records adds only 8.3 ms median overhead; and controlled recovery experiments show that responsibility-relevant information can be partially recovered even when conventional logs are missing. We propose an Auditability Card for agent systems and identify six open research problems organized by mechanism class.

We present MindEye, a novel fMRI-to-image approach to retrieve and reconstruct viewed images from brain activity. Our model comprises two parallel submodules that are specialized for retrieval (using contrastive learning) and reconstruction (using a diffusion prior). MindEye can map fMRI brain activity to any high dimensional multimodal latent space, like CLIP image space, enabling image reconstruction using generative models that accept embeddings from this latent space. We comprehensively compare our approach with other existing methods, using both qualitative side-by-side comparisons and quantitative evaluations, and show that MindEye achieves state-of-the-art performance in both reconstruction and retrieval tasks. In particular, MindEye can retrieve the exact original image even among highly similar candidates indicating that its brain embeddings retain fine-grained image-specific information. This allows us to accurately retrieve images even from large-scale databases like LAION-5B. We demonstrate through ablations that MindEye's performance improvements over previous methods result from specialized submodules for retrieval and reconstruction, improved training techniques, and training models with orders of magnitude more parameters. Furthermore, we show that MindEye can better preserve low-level image features in the reconstructions by using img2img, with outputs from a separate autoencoder. All code is available on GitHub.

Recent works have showcased the ability of LLMs to embody diverse personas in their responses, exemplified by prompts like 'You are Yoda. Explain the Theory of Relativity.' While this ability allows personalization of LLMs and enables human behavior simulation, its effect on LLMs' capabilities remains unclear. To fill this gap, we present the first extensive study of the unintended side-effects of persona assignment on the ability of LLMs to perform basic reasoning tasks. Our study covers 24 reasoning datasets, 4 LLMs, and 19 diverse personas (e.g. an Asian person) spanning 5 socio-demographic groups. Our experiments unveil that LLMs harbor deep rooted bias against various socio-demographics underneath a veneer of fairness. While they overtly reject stereotypes when explicitly asked ('Are Black people less skilled at mathematics?'), they manifest stereotypical and erroneous presumptions when asked to answer questions while adopting a persona. These can be observed as abstentions in responses, e.g., 'As a Black person, I can't answer this question as it requires math knowledge', and generally result in a substantial performance drop. Our experiments with ChatGPT-3.5 show that this bias is ubiquitous - 80% of our personas demonstrate bias; it is significant - some datasets show performance drops of 70%+; and can be especially harmful for certain groups - some personas suffer statistically significant drops on 80%+ of the datasets. Overall, all 4 LLMs exhibit this bias to varying extents, with GPT-4-Turbo showing the least but still a problematic amount of bias (evident in 42% of the personas). Further analysis shows that these persona-induced errors can be hard-to-discern and hard-to-avoid. Our findings serve as a cautionary tale that the practice of assigning personas to LLMs - a trend on the rise - can surface their deep-rooted biases and have unforeseeable and detrimental side-effects.

A major goal of unsupervised learning is to discover data representations that are useful for subsequent tasks, without access to supervised labels during training. Typically, this involves minimizing a surrogate objective, such as the negative log likelihood of a generative model, with the hope that representations useful for subsequent tasks will arise as a side effect. In this work, we propose instead to directly target later desired tasks by meta-learning an unsupervised learning rule which leads to representations useful for those tasks. Specifically, we target semi-supervised classification performance, and we meta-learn an algorithm -- an unsupervised weight update rule -- that produces representations useful for this task. Additionally, we constrain our unsupervised update rule to a be a biologically-motivated, neuron-local function, which enables it to generalize to different neural network architectures, datasets, and data modalities. We show that the meta-learned update rule produces useful features and sometimes outperforms existing unsupervised learning techniques. We further show that the meta-learned unsupervised update rule generalizes to train networks with different widths, depths, and nonlinearities. It also generalizes to train on data with randomly permuted input dimensions and even generalizes from image datasets to a text task.

Cross-silo federated learning offers a promising solution to collaboratively train robust and generalized AI models without compromising the privacy of local datasets, e.g., healthcare, financial, as well as scientific projects that lack a centralized data facility. Nonetheless, because of the disparity of computing resources among different clients (i.e., device heterogeneity), synchronous federated learning algorithms suffer from degraded efficiency when waiting for straggler clients. Similarly, asynchronous federated learning algorithms experience degradation in the convergence rate and final model accuracy on non-identically and independently distributed (non-IID) heterogeneous datasets due to stale local models and client drift. To address these limitations in cross-silo federated learning with heterogeneous clients and data, we propose FedCompass, an innovative semi-asynchronous federated learning algorithm with a computing power-aware scheduler on the server side, which adaptively assigns varying amounts of training tasks to different clients using the knowledge of the computing power of individual clients. FedCompass ensures that multiple locally trained models from clients are received almost simultaneously as a group for aggregation, effectively reducing the staleness of local models. At the same time, the overall training process remains asynchronous, eliminating prolonged waiting periods from straggler clients. Using diverse non-IID heterogeneous distributed datasets, we demonstrate that FedCompass achieves faster convergence and higher accuracy than other asynchronous algorithms while remaining more efficient than synchronous algorithms when performing federated learning on heterogeneous clients. The source code for FedCompass is available at https://github.com/APPFL/FedCompass.

Language models trained with reinforcement learning may learn to optimize the grader's judgment rather than the intended objective. This "reward-seeking" is difficult to measure because a model that pursues the grader's judgment and one that pursues the intended objective behave identically whenever the grader rewards the intended behavior. We measure reward-seeking using Contrastive Synthetic Document Finetuning to change a model's beliefs about what the grader rewards, putting those beliefs in conflict with what users or developers want, and measuring the rate at which the model adopts each party's preferred behavior. Applied to intermediate checkpoints of a capabilities-focused OpenAI o3 RL run, without safety training, we find that these checkpoints often side with grader preferences over those of users or developers on coding and alignment tasks. This tendency to side with the grader trends upward throughout RL training. For example, in an environment that forces a choice between keeping a promise to a supervisor and breaking it to complete the task, a late capabilities-focused o3 checkpoint breaks the promise 87% of the time when SDF documents say the grader rewards task completion, versus 9% when they say it rewards honesty (a choice its chain-of-thought often makes explicit). An earlier checkpoint is far less sensitive (40% vs. 24%). Our method also generalizes to reward-hacking models. A model organism trained to reward-hack (gpt-oss-120b) is more than twice as sensitive to grader preferences as the unmodified model, with the mean behavioral shift in favor of the grader rising from 33% to 86%. These results indicate that RL can increase reward-seeking over the course of training, producing models that may act against their developers' intentions when they believe that doing so leads to higher reward.

Open-sourced Large Language Models (LLMs) have achieved great success in various NLP tasks, however, they are still far inferior to API-based models when acting as agents. How to integrate agent ability into general LLMs becomes a crucial and urgent problem. This paper first delivers three key observations: (1) the current agent training corpus is entangled with both formats following and agent reasoning, which significantly shifts from the distribution of its pre-training data; (2) LLMs exhibit different learning speeds on the capabilities required by agent tasks; and (3) current approaches have side-effects when improving agent abilities by introducing hallucinations. Based on the above findings, we propose Agent-FLAN to effectively Fine-tune LANguage models for Agents. Through careful decomposition and redesign of the training corpus, Agent-FLAN enables Llama2-7B to outperform prior best works by 3.5\% across various agent evaluation datasets. With comprehensively constructed negative samples, Agent-FLAN greatly alleviates the hallucination issues based on our established evaluation benchmark. Besides, it consistently improves the agent capability of LLMs when scaling model sizes while slightly enhancing the general capability of LLMs. The code will be available at https://github.com/InternLM/Agent-FLAN.

We introduce Dynatask: an open source system for setting up custom NLP tasks that aims to greatly lower the technical knowledge and effort required for hosting and evaluating state-of-the-art NLP models, as well as for conducting model in the loop data collection with crowdworkers. Dynatask is integrated with Dynabench, a research platform for rethinking benchmarking in AI that facilitates human and model in the loop data collection and evaluation. To create a task, users only need to write a short task configuration file from which the relevant web interfaces and model hosting infrastructure are automatically generated. The system is available at https://dynabench.org/ and the full library can be found at https://github.com/facebookresearch/dynabench.

We present the FinCausal 2020 Shared Task on Causality Detection in Financial Documents and the associated FinCausal dataset, and discuss the participating systems and results. Two sub-tasks are proposed: a binary classification task (Task 1) and a relation extraction task (Task 2). A total of 16 teams submitted runs across the two Tasks and 13 of them contributed with a system description paper. This workshop is associated to the Joint Workshop on Financial Narrative Processing and MultiLing Financial Summarisation (FNP-FNS 2020), held at The 28th International Conference on Computational Linguistics (COLING'2020), Barcelona, Spain on September 12, 2020.

We describe the SemEval task of extracting keyphrases and relations between them from scientific documents, which is crucial for understanding which publications describe which processes, tasks and materials. Although this was a new task, we had a total of 26 submissions across 3 evaluation scenarios. We expect the task and the findings reported in this paper to be relevant for researchers working on understanding scientific content, as well as the broader knowledge base population and information extraction communities.

We present the results and the main findings of SemEval-2024 Task 8: Multigenerator, Multidomain, and Multilingual Machine-Generated Text Detection. The task featured three subtasks. Subtask A is a binary classification task determining whether a text is written by a human or generated by a machine. This subtask has two tracks: a monolingual track focused solely on English texts and a multilingual track. Subtask B is to detect the exact source of a text, discerning whether it is written by a human or generated by a specific LLM. Subtask C aims to identify the changing point within a text, at which the authorship transitions from human to machine. The task attracted a large number of participants: subtask A monolingual (126), subtask A multilingual (59), subtask B (70), and subtask C (30). In this paper, we present the task, analyze the results, and discuss the system submissions and the methods they used. For all subtasks, the best systems used LLMs.

This study explores the proactive ability of LLMs to seek user support. We propose metrics to evaluate the trade-off between performance improvements and user burden, and investigate whether LLMs can determine when to request help under varying information availability. Our experiments show that without external feedback, many LLMs struggle to recognize their need for user support. The findings highlight the importance of external signals and provide insights for future research on improving support-seeking strategies. Source code: https://github.com/appier-research/i-need-help

Much knowledge work produces new deliverables from files a workspace already holds, and LLM agents are beginning to take such work over. Through direct corpus interaction, an agent can search and read any of those files from a terminal with no indexing, and producing a deliverable from many of them in this way is what we call direct workspace interaction (DWI). Reaching the files, however, is only half the task: nothing keeps track of what the task asks for, what has been read, and what was listed but never opened, all of which slip through the context window without leaving a trace, so an agent may extract a figure and still deliver a report without it. To address this, we present RunningTab, a framework that equips direct workspace interaction with an environment-side tab: a per-task record of what the task still owes, kept by the environment alongside the agent. Specifically, the agent adds its requirements, while the environment records every file read as an excerpt with its provenance and every listed but unopened file as a candidate; the agent can then see each requirement beside its best-matching excerpts and top unopened candidates, resolve it against matching content or set it aside with a reason, and, should it try to finish with requirements still open, receive them in a finish check. We validate RunningTab on three benchmarks with three LLMs, where it consistently outperforms plain DWI and baselines that keep the record in the model, while its tab usually holds the values a deliverable needs once seen.

WANDR (Wide ANd Deep Research) is a benchmark of 500 realistic, challenging data-collection tasks for research agents. Each task requires a system to discover a large set of entities that satisfy specified criteria (breadth), investigate each entity through multiple coordinated web searches (depth), and return independently verifiable records with supporting sources and excerpts. Tasks are represented as qualification key hierarchies that specify the entities, relationships, evidence, and required count at each level; a hierarchy with n companies, m employees per company, and k sources per employee requires n x m x k records. This structure supports diverse workflows such as market mapping, due diligence, literature review, product comparison, and talent sourcing, with targets ranging from dozens to thousands of records. WANDR replaces static gold answer sets with task-specific judges that refetch cited pages and verify each record against its evidence, allowing evaluation of current and changing facts. Record verdicts are aggregated into soft and hard precision, recall, and F1 scores that distinguish factual quality, coverage, and hierarchical completeness. The tasks are derived from de-identified product-usage logs and produced through a semi-automated pipeline with automated checks, empirical audits, and human review where needed. We evaluate six production research systems and find that the benchmark is far from saturated: at high effort, the strongest system reaches only 0.363 soft F1 and 0.133 hard F1. Performance degrades as target volume and hierarchy depth increase, with incomplete discovery, missing enrichment, and incomplete evidence construction remaining major bottlenecks. The benchmark and evaluation harness are available at https://github.com/perplexityai/wandr.

Real-world tasks consist of multiple inter-dependent subtasks (e.g., a dirty pan needs to be washed before it can be used for cooking). In this work, we aim to model the causal dependencies between such subtasks from instructional videos describing the task. This is a challenging problem since complete information about the world is often inaccessible from videos, which demands robust learning mechanisms to understand the causal structure of events. We present Multimodal Subtask Graph Generation (MSG2), an approach that constructs a Subtask Graph defining the dependency between a task's subtasks relevant to a task from noisy web videos. Graphs generated by our multimodal approach are closer to human-annotated graphs compared to prior approaches. MSG2 further performs the downstream task of next subtask prediction 85% and 30% more accurately than recent video transformer models in the ProceL and CrossTask datasets, respectively.

One long-term goal of machine learning research is to produce methods that are applicable to reasoning and natural language, in particular building an intelligent dialogue agent. To measure progress towards that goal, we argue for the usefulness of a set of proxy tasks that evaluate reading comprehension via question answering. Our tasks measure understanding in several ways: whether a system is able to answer questions via chaining facts, simple induction, deduction and many more. The tasks are designed to be prerequisites for any system that aims to be capable of conversing with a human. We believe many existing learning systems can currently not solve them, and hence our aim is to classify these tasks into skill sets, so that researchers can identify (and then rectify) the failings of their systems. We also extend and improve the recently introduced Memory Networks model, and show it is able to solve some, but not all, of the tasks.

The recent success of Large Language Models (LLMs) signifies an impressive stride towards artificial general intelligence. They have shown a promising prospect in automatically completing tasks upon user instructions, functioning as brain-like coordinators. The associated risks will be revealed as we delegate an increasing number of tasks to machines for automated completion. A big question emerges: how can we make machines behave responsibly when helping humans automate tasks as personal copilots? In this paper, we explore this question in depth from the perspectives of feasibility, completeness and security. In specific, we present Responsible Task Automation (ResponsibleTA) as a fundamental framework to facilitate responsible collaboration between LLM-based coordinators and executors for task automation with three empowered capabilities: 1) predicting the feasibility of the commands for executors; 2) verifying the completeness of executors; 3) enhancing the security (e.g., the protection of users' privacy). We further propose and compare two paradigms for implementing the first two capabilities. One is to leverage the generic knowledge of LLMs themselves via prompt engineering while the other is to adopt domain-specific learnable models. Moreover, we introduce a local memory mechanism for achieving the third capability. We evaluate our proposed ResponsibleTA on UI task automation and hope it could bring more attentions to ensuring LLMs more responsible in diverse scenarios. The research project homepage is at https://task-automation-research.github.io/responsible_task_automation.

In this paper, we introduce the first evaluation of Chinese human-computer dialogue technology. We detail the evaluation scheme, tasks, metrics and how to collect and annotate the data for training, developing and test. The evaluation includes two tasks, namely user intent classification and online testing of task-oriented dialogue. To consider the different sources of the data for training and developing, the first task can also be divided into two sub tasks. Both the two tasks are coming from the real problems when using the applications developed by industry. The evaluation data is provided by the iFLYTEK Corporation. Meanwhile, in this paper, we publish the evaluation results to present the current performance of the participants in the two tasks of Chinese human-computer dialogue technology. Moreover, we analyze the existing problems of human-computer dialogue as well as the evaluation scheme itself.

Terminal-agent benchmarks have become a primary signal for measuring the coding and system-administration capabilities of large language models. As the market for evaluation environments grows, so does the pressure to ship tasks quickly, often without thorough adversarial review of the verification logic. This paper is a guideline for writing good benchmark tasks, drawn from over a year of contributing to and reviewing tasks for Terminal Bench. Most people write benchmark tasks the way they write prompts. They shouldn't. A prompt is designed to help the agent succeed; a benchmark is designed to find out if it can. We argue that good tasks are adversarial, difficult, and legible, and that a large class of common failure modes -- AI-generated instructions, over-prescriptive specifications, clerical difficulty, oracle solutions that assume hidden knowledge, tests that validate the wrong things, and reward-hackable environments -- are predictable consequences of treating task authoring as prompt authoring. We catalog these failure modes, argue that real difficulty is conceptual rather than environmental, and discuss recent empirical evidence that over 15% of tasks in popular terminal-agent benchmarks are reward-hackable. We hope this serves as a useful reference for benchmark maintainers, task contributors, and researchers using benchmark scores as evidence.

Recently, instruction-following Large Language Models (LLMs) , represented by ChatGPT, have exhibited exceptional performance in general Natural Language Processing (NLP) tasks. However, the unique characteristics of E-commerce data pose significant challenges to general LLMs. An LLM tailored specifically for E-commerce scenarios, possessing robust cross-dataset/task generalization capabilities, is a pressing necessity. To solve this issue, in this work, we proposed the first e-commerce instruction dataset EcomInstruct, with a total of 2.5 million instruction data. EcomInstruct scales up the data size and task diversity by constructing atomic tasks with E-commerce basic data types, such as product information, user reviews. Atomic tasks are defined as intermediate tasks implicitly involved in solving a final task, which we also call Chain-of-Task tasks. We developed EcomGPT with different parameter scales by training the backbone model BLOOMZ with the EcomInstruct. Benefiting from the fundamental semantic understanding capabilities acquired from the Chain-of-Task tasks, EcomGPT exhibits excellent zero-shot generalization capabilities. Extensive experiments and human evaluations demonstrate that EcomGPT outperforms ChatGPT in term of cross-dataset/task generalization on E-commerce tasks.

It is still a pipe dream that personal AI assistants on the phone and AR glasses can assist our daily life in addressing our questions like ``how to adjust the date for this watch?'' and ``how to set its heating duration? (while pointing at an oven)''. The queries used in conventional tasks (i.e. Video Question Answering, Video Retrieval, Moment Localization) are often factoid and based on pure text. In contrast, we present a new task called Task-oriented Question-driven Video Segment Retrieval (TQVSR). Each of our questions is an image-box-text query that focuses on affordance of items in our daily life and expects relevant answer segments to be retrieved from a corpus of instructional video-transcript segments. To support the study of this TQVSR task, we construct a new dataset called AssistSR. We design novel guidelines to create high-quality samples. This dataset contains 3.2k multimodal questions on 1.6k video segments from instructional videos on diverse daily-used items. To address TQVSR, we develop a simple yet effective model called Dual Multimodal Encoders (DME) that significantly outperforms several baseline methods while still having large room for improvement in the future. Moreover, we present detailed ablation analyses. Code and data are available at https://github.com/StanLei52/TQVSR.

Existing computer-use agent benchmarks do not fully evaluate agents acting as assistants. A useful assistant retrieves information across complex, multi-step workflows, synthesizes it into artifacts (documents, presentations, spreadsheets), and navigates program interfaces to produce a coherent final product. Such workflows demand reasoning and synthesis, decomposition of complex tasks, as well as visual and spatial understanding. To study agents on workflows like these, we introduce KNOWS, a benchmark of open-ended, complex, browser-based tasks that jointly evaluate these capabilities, with each task culminating in a produced artifact. To write tasks, we develop a task design rubric and a protocol for ensuring that tasks meet the requirements. Each task is paired with an evaluator, a program that combines deterministic checks with LLM judgments to balance the richness, reliability, and automation tradeoff inherent to agent evaluation. We evaluate and analyze frontier computer-use agents and browser-based harnesses. They achieve moderate scores on partial-success metrics, but the best performer fully succeeds in fewer than 3% of our complex, long-horizon tasks. Failures on visual steps render the resulting artifacts unusable, even when agents complete more than 50% of other evaluation steps. Our results expose limitations of current agents acting as end-to-end assistants, and call for progress on tool use, visual understanding, and long-horizon reasoning.

Recent human-computer interaction (HCI) research has revealed a widespread misalignment between how developers design workplace artificial intelligence (AI) systems, and what workers actually need from them. Yet, little research has examined the effects of this gap, or how it may cause harm. We analyzed 1,524 reports of incidents in which AI systems were used to perform 171 occupational tasks across 12 industry sectors. Using an Large Language Model (LLM)-as-an-expert approach, we extracted the main traits of the AI systems involved in those incidents using an established framework of twelve traits. We then compared them with the traits that 202 workers highly familiar with those tasks would have preferred. We found that as many as 83\% of workplace incidents stem from worker-AI misalignments. In most cases, workers wanted systems that are precise, insightful, or personal, but instead received systems that are basic, simple, or general. Over the years, fast AI caused a considerable number of incidents, yet these declined, and imaginative AI, with the mass introduction of generative AI, started to cause incidents. We also compared the traits causing the incidents with the traits that 197 developers building AI systems for those tasks would have preferred. If the traits causing the incidents were the same as those designed by developers, then developers may be responsible for those incidents. We found that 74\% of task misalignments could be attributed to developers who tended to overfocus on efficiency and speed, especially for systems performing tasks in people-facing occupations such as those in the human resources sector. Our results call for design interventions that better align AI development with workers' needs, as without such corrections, workplace AI incidents are likely to persist, causing the invisible erosion of worker agency and organizational productivity.

We present the results and the main findings of SemEval-2020 Task 11 on Detection of Propaganda Techniques in News Articles. The task featured two subtasks. Subtask SI is about Span Identification: given a plain-text document, spot the specific text fragments containing propaganda. Subtask TC is about Technique Classification: given a specific text fragment, in the context of a full document, determine the propaganda technique it uses, choosing from an inventory of 14 possible propaganda techniques. The task attracted a large number of participants: 250 teams signed up to participate and 44 made a submission on the test set. In this paper, we present the task, analyze the results, and discuss the system submissions and the methods they used. For both subtasks, the best systems used pre-trained Transformers and ensembles.

Large language model (LLM) agents are increasingly deployed as personal assistants. Existing evaluations, however, mostly use short, self-contained requests in static environments. Everyday life assistance is different. A task runs for weeks rather than minutes. The world keeps changing while the agent is not being prompted. Many constraints are never stated outright. An agent that merely answers the request in front of it will fail at such a task. What is needed instead is an agent that stays proactive and consistent. It decides on its own when to act, when to ask, and when to stay silent. It notices changes that nobody announced. It keeps one plan coherent from the first day to the last. No current benchmark measures this. We introduce VibeLifeBench, a benchmark of 200 long-horizon tasks across ten everyday-life domains. Each task is a scripted multi-week timeline in a simulated world of 22 mock services. The world advances on its own clock, and many of its changes are silent, so only an agent that re-inspects the world discovers them. Every task is graded by fine-grained, weighted checks that read only what the agent actually left behind, covering the end state, the timeliness of its actions, and whether it upheld the implicit constraints. We evaluate seven frontier models. All of them score low, which shows how far current agents are from assisting with real life. We will open-source all tasks, environments, and the evaluation framework.

We propose a new shared task for tactical data-to-text generation in the domain of source code libraries. Specifically, we focus on text generation of function descriptions from example software projects. Data is drawn from existing resources used for studying the related problem of semantic parser induction (Richardson and Kuhn, 2017b; Richardson and Kuhn, 2017a), and spans a wide variety of both natural languages and programming languages. In this paper, we describe these existing resources, which will serve as training and development data for the task, and discuss plans for building new independent test sets.

There has been growing interest in utilizing occupational data mining and analysis. In today's job market, occupational data mining and analysis is growing in importance as it enables companies to predict employee turnover, model career trajectories, screen through resumes and perform other human resource tasks. A key requirement to facilitate these tasks is the need for an occupation-related dataset. However, most research use proprietary datasets or do not make their dataset publicly available, thus impeding development in this area. To solve this issue, we present the Industrial and Professional Occupation Dataset (IPOD), which comprises 192k job titles belonging to 56k LinkedIn users. In addition to making IPOD publicly available, we also: (i) manually annotate each job title with its associated level of seniority, domain of work and location; and (ii) provide embedding for job titles and discuss various use cases. This dataset is publicly available at https://github.com/junhua/ipod.

Research on Machine Translation (MT) has achieved important breakthroughs in several areas. While there is much more to be done in order to build on this success, we believe that the language industry needs better ways to take full advantage of current achievements. Due to a combination of factors, including time, resources, and skills, businesses tend to apply pragmatism into their AI workflows. Hence, they concentrate more on outcomes, e.g. delivery, shipping, releases, and features, and adopt high-level working production solutions, where possible. Among the features thought to be helpful for translators are sentence-level and word-level translation auto-suggestion and auto-completion. Suggesting alternatives can inspire translators and limit their need to refer to external resources, which hopefully boosts their productivity. This work describes our submissions to WMT's shared task on word-level auto-completion, for the Chinese-to-English, English-to-Chinese, German-to-English, and English-to-German language directions. We investigate the possibility of using pre-trained models and out-of-the-box features from available libraries. We employ random sampling to generate diverse alternatives, which reveals good results. Furthermore, we introduce our open-source API, based on CTranslate2, to serve translations, auto-suggestions, and auto-completions.

This work explores the problem of generating task graphs of real-world activities. Different from prior formulations, we consider a setting where text transcripts of instructional videos performing a real-world activity (e.g., making coffee) are provided and the goal is to identify the key steps relevant to the task as well as the dependency relationship between these key steps. We propose a novel task graph generation approach that combines the reasoning capabilities of instruction-tuned language models along with clustering and ranking components to generate accurate task graphs in a completely unsupervised manner. We show that the proposed approach generates more accurate task graphs compared to a supervised learning approach on tasks from the ProceL and CrossTask datasets.

Auxiliary tasks improve the representations learned by deep reinforcement learning agents. Analytically, their effect is reasonably well understood; in practice, however, their primary use remains in support of a main learning objective, rather than as a method for learning representations. This is perhaps surprising given that many auxiliary tasks are defined procedurally, and hence can be treated as an essentially infinite source of information about the environment. Based on this observation, we study the effectiveness of auxiliary tasks for learning rich representations, focusing on the setting where the number of tasks and the size of the agent's network are simultaneously increased. For this purpose, we derive a new family of auxiliary tasks based on the successor measure. These tasks are easy to implement and have appealing theoretical properties. Combined with a suitable off-policy learning rule, the result is a representation learning algorithm that can be understood as extending Mahadevan & Maggioni (2007)'s proto-value functions to deep reinforcement learning -- accordingly, we call the resulting object proto-value networks. Through a series of experiments on the Arcade Learning Environment, we demonstrate that proto-value networks produce rich features that may be used to obtain performance comparable to established algorithms, using only linear approximation and a small number (~4M) of interactions with the environment's reward function.

Recent progress in Large Language Models (LLMs) has produced models that exhibit remarkable performance across a variety of NLP tasks. However, it remains unclear whether the existing focus of NLP research accurately captures the genuine requirements of human users. This paper provides a comprehensive analysis of the divergence between current NLP research and the needs of real-world NLP applications via a large-scale collection of user-GPT conversations. We analyze a large-scale collection of real user queries to GPT. We compare these queries against existing NLP benchmark tasks and identify a significant gap between the tasks that users frequently request from LLMs and the tasks that are commonly studied in academic research. For example, we find that tasks such as ``design'' and ``planning'' are prevalent in user interactions but are largely neglected or different from traditional NLP benchmarks. We investigate these overlooked tasks, dissect the practical challenges they pose, and provide insights toward a roadmap to make LLMs better aligned with user needs.

Transferring learned patterns from pretrained neural language models has been shown to significantly improve effectiveness across a variety of language-based tasks, meanwhile further tuning on intermediate tasks has been demonstrated to provide additional performance benefits, provided the intermediate task is sufficiently related to the target task. However, how to identify related tasks is an open problem, and brute-force searching effective task combinations is prohibitively expensive. Hence, the question arises, are we able to improve the effectiveness and efficiency of tasks with no training examples through selective fine-tuning? In this paper, we explore statistical measures that approximate the divergence between domain representations as a means to estimate whether tuning using one task pair will exhibit performance benefits over tuning another. This estimation can then be used to reduce the number of task pairs that need to be tested by eliminating pairs that are unlikely to provide benefits. Through experimentation over 58 tasks and over 6,600 task pair combinations, we demonstrate that statistical measures can distinguish effective task pairs, and the resulting estimates can reduce end-to-end runtime by up to 40%.

Long-horizon terminal tasks are used to evaluate and train agents. Human-curated datasets are scarce, so many synthesis methods have emerged. They usually measure the length of a task by the turns a model needs to pass it, and its difficulty by labels the LLM sets during generation. We find, however, that the turn count largely reflects the habits of the model, not only the demands of the task, and that the labels barely track how often models pass. We describe a terminal task by its work, the number of operations it requires, and its depth, the number of turns that must happen in sequence because the commands of each turn can only be chosen after reading the output of the previous one. Estimating depth as the smallest turn cap, stated to the agent, under which the task passes, we audit 734 tasks from six existing datasets. Synthesis has mostly added work, not depth: 94% of the solved synthetic tasks pass within three turns, yet depth, unlike the turn count, predicts difficulty beyond work. We propose DEPTH, in which a generation agent writes each task one segment at a time in a real software environment, carrying out each segment before writing it and starting the next from the state it leaves. Our 105 tasks exceed every audited dataset in depth and work, and none of six current models passes half of them. Fine-tuning Qwen3-8B on runs from these tasks improves it on two terminal benchmarks.

Assistants on assembly tasks have a large potential to benefit humans from everyday tasks to industrial settings. However, no testbeds support application-oriented system evaluation in a practical setting, especially in assembly. To foster the development, we propose a new multimodal QA dataset on assembly activities. Our dataset, ProMQA-Assembly, consists of 391 QA pairs that require the multimodal understanding of human-activity recordings and their instruction manuals in an online-style manner. In the development, we adopt a semi-automated QA annotation approach, where LLMs generate candidates and humans verify them, as a cost-effective method, and further improve it by integrating fine-grained action labels to diversify question types. Furthermore, we create instruction task graphs for the target tasks of assembling toy vehicles. These newly created task graphs are used in our benchmarking experiment, as well as to facilitate the human verification process in the QA annotation. Utilizing our dataset, we benchmark models, including competitive proprietary multimodal models. Our results suggest great room for improvement for the current models. We believe our new evaluation dataset can contribute to the further development of procedural-activity assistants.

We study the problem of multi-task learning under user-level differential privacy, in which n users contribute data to m tasks, each involving a subset of users. One important aspect of the problem, that can significantly impact quality, is the distribution skew among tasks. Certain tasks may have much fewer data samples than others, making them more susceptible to the noise added for privacy. It is natural to ask whether algorithms can adapt to this skew to improve the overall utility. We give a systematic analysis of the problem, by studying how to optimally allocate a user's privacy budget among tasks. We propose a generic algorithm, based on an adaptive reweighting of the empirical loss, and show that when there is task distribution skew, this gives a quantifiable improvement of excess empirical risk. Experimental studies on recommendation problems that exhibit a long tail of small tasks, demonstrate that our methods significantly improve utility, achieving the state of the art on two standard benchmarks.

The 1st Workshop on Data Contamination (CONDA 2024) focuses on all relevant aspects of data contamination in natural language processing, where data contamination is understood as situations where evaluation data is included in pre-training corpora used to train large scale models, compromising evaluation results. The workshop fostered a shared task to collect evidence on data contamination in current available datasets and models. The goal of the shared task and associated database is to assist the community in understanding the extent of the problem and to assist researchers in avoiding reporting evaluation results on known contaminated resources. The shared task provides a structured, centralized public database for the collection of contamination evidence, open to contributions from the community via GitHub pool requests. This first compilation paper is based on 566 reported entries over 91 contaminated sources from a total of 23 contributors. The details of the individual contamination events are available in the platform. The platform continues to be online, open to contributions from the community.

Our goal is to assess if AutoML system changes - i.e., to the search space or hyperparameter optimization - will improve the final model's performance on production tasks. However, we cannot test the changes on production tasks. Instead, we only have access to limited descriptors about tasks that our AutoML system previously executed, like the number of data points or features. We also have a set of development tasks to test changes, ex., sampled from OpenML with no usage constraints. However, the development and production task distributions are different leading us to pursue changes that only improve development and not production. This paper proposes a method to leverage descriptor information about AutoML production tasks to select a filtered subset of the most relevant development tasks. Empirical studies show that our filtering strategy improves the ability to assess AutoML system changes on holdout tasks with different distributions than development.

We present a benchmark for Planning And Reasoning Tasks in humaN-Robot collaboration (PARTNR) designed to study human-robot coordination in household activities. PARTNR tasks exhibit characteristics of everyday tasks, such as spatial, temporal, and heterogeneous agent capability constraints. We employ a semi-automated task generation pipeline using Large Language Models (LLMs), incorporating simulation in the loop for grounding and verification. PARTNR stands as the largest benchmark of its kind, comprising 100,000 natural language tasks, spanning 60 houses and 5,819 unique objects. We analyze state-of-the-art LLMs on PARTNR tasks, across the axes of planning, perception and skill execution. The analysis reveals significant limitations in SoTA models, such as poor coordination and failures in task tracking and recovery from errors. When LLMs are paired with real humans, they require 1.5x as many steps as two humans collaborating and 1.1x more steps than a single human, underscoring the potential for improvement in these models. We further show that fine-tuning smaller LLMs with planning data can achieve performance on par with models 9 times larger, while being 8.6x faster at inference. Overall, PARTNR highlights significant challenges facing collaborative embodied agents and aims to drive research in this direction.

Long-horizon agents must complete workflows in which later activities depend on earlier outputs and changes to shared state. A key challenge for evaluation is to expand task requirements and their dependencies while keeping tasks executable and outcomes verifiable. We introduce SkillHorizon, a benchmark that constructs long-horizon tasks through skill composition in seven business environments. Its generators use domain-specific rules to turn reusable skill procedures into concrete business activities. These activities are connected through input-output dependencies, shared resources, and state updates, allowing workload and coordination demands to grow together. Successful reference executions validate task feasibility and measure interaction workload, while programmatic evaluators assess local completion and whole-task success. We evaluate five language models on 1,050 tasks with reference interaction workloads ranging from 64K to 1M tokens. Across these scales, mean local-goal completion falls from 90.4% to 46.8%, as completed work fails to keep pace with growing requirements. At larger workloads, agents more often submit unsuccessful final answers with substantial work unfinished and tool-call allowance remaining. Trace analysis also identifies cases where scripts execute successfully but their outputs fail downstream business checks. SkillHorizon provides a configurable testbed for studying sustained progress, dependency coordination, and complete task delivery in growing workflows. Our code is available at https://github.com/KnowComp/SkillHorizon.

After a runtime or API migration, an LLM-repaired procedure can run cleanly yet return wrong results, and a release check on typical test inputs ships it. We report a preregistered study of 73 tasks covering 60 documented changes in CPython, Node.js and SQLite. Every candidate is executed on the real runtimes against withheld contracts. Over the complete repair space, the all-assertions gate on typical inputs released 0.410 of wrong repairs. One edge input written from the release note cut this to 0.065, and to 0.061 on 15 unseen clusters. Three inputs from a blind GPT-5.6-luna writer that never saw the program reached 0.070. Two further frozen stages enlarged the menu to 1491 fragments and then added 233 hard distractors that match the legacy output on typical inputs. The edge input (0.058) and the frontier-model writer (0.058) passed their preregistered tests. Yet 0.168 and 0.153 of the hard wrong programs still passed. Two local-model writers failed their preregistered tests. Across four model families, the edge input cut wrong releases per attempt in all three open-weight families on all three menus, for example from 0.233 to 0.057, without lowering correct completion. Self-tests helped less. A blind AI audit flagged 34 contracts as stricter than their requirements. Dropping the disputed checks changed 0 verdicts, but class-label disputes remain open. No model-written code was executed. Artifacts are available from the authors on request.

We consider the problem of building an assistive robotic system that can help humans in daily household cleanup tasks. Creating such an autonomous system in real-world environments is inherently quite challenging, as a general solution may not suit the preferences of a particular customer. Moreover, such a system consists of multi-objective tasks comprising -- (i) Detection of misplaced objects and prediction of their potentially correct placements, (ii) Fine-grained manipulation for stable object grasping, and (iii) Room-to-room navigation for transferring objects in unseen environments. This work systematically tackles each component and integrates them into a complete object rearrangement pipeline. To validate our proposed system, we conduct multiple experiments on a real robotic platform involving multi-room object transfer, user preference-based placement, and complex pick-and-place tasks. Project page: https://sites.google.com/eng.ucsd.edu/home-robot

Modern web interfaces are unnecessarily complex to use as they overwhelm users with excessive text and visuals unrelated to their current goals. This problem particularly impacts screen reader users (SRUs), who navigate content sequentially and may spend minutes traversing irrelevant elements before reaching desired information compared to vision users (VUs) who visually skim in seconds. We present Task Mode, a system that dynamically filters web content based on user-specified goals using large language models to identify and prioritize relevant elements while minimizing distractions. Our approach preserves page structure while offering multiple viewing modes tailored to different access needs. Our user study with 12 participants (6 VUs, 6 SRUs) demonstrates that our approach reduced task completion time for SRUs while maintaining performance for VUs, decreasing the completion time gap between groups from 2x to 1.2x. 11 of 12 participants wanted to use Task Mode in the future, reporting that Task Mode supported completing tasks with less effort and fewer distractions. This work demonstrates how designing new interactions simultaneously for visual and non-visual access can reduce rather than reinforce accessibility disparities in future technology created by human-computer interaction researchers and practitioners.

We introduce SWE-Bench Pro, a substantially more challenging benchmark that builds upon the best practices of SWE-BENCH [25], but is explicitly designed to capture realistic, complex, enterprise-level problems beyond the scope of SWE-BENCH. SWE-BENCH PRO contains 1,865 problems sourced from a diverse set of 41 actively maintained repositories spanning business applications, B2B services, and developer tools. The benchmark is partitioned into a public set with open access to problems sourced from 11 repositories, a held-out set of 12 repositories and a commercial set of 18 proprietary repositories where we have formal partnership agreements with early-stage startups. Problems in the held-out and the commercial set are not publicly accessible, but we release results on the commercial set. Our benchmark features long-horizon tasks that may require hours to days for a professional software engineer to complete, often involving patches across multiple files and substantial code modifications. All tasks are human-verified and augmented with sufficient context to ensure resolvability. In our evaluation of widely used coding models, under a unified scaffold, we observe that their performance on SWE-Bench PRO remains below 25% (Pass@1), with GPT-5 achieving the highest score to date at 23.3%. To better understand these limitations, we cluster the failure modes observed in the collected agent trajectories for a clearer characterization of the error patterns exhibited by current models. Overall, SWE-BENCH PRO provides a contamination-resistant testbed that more faithfully captures the complexity and diversity of real-world software development, advancing the pursuit of truly autonomous software engineering agents at a professional level.

We introduce SWE-Lancer, a benchmark of over 1,400 freelance software engineering tasks from Upwork, valued at \1 million USD total in real-world payouts. SWE-Lancer encompasses both independent engineering tasks--ranging from 50 bug fixes to \$32,000 feature implementations--and managerial tasks, where models choose between technical implementation proposals. Independent tasks are graded with end-to-end tests triple-verified by experienced software engineers, while managerial decisions are assessed against the choices of the original hired engineering managers. We evaluate model performance and find that frontier models are still unable to solve the majority of tasks. To facilitate future research, we open-source a unified Docker image and a public evaluation split, SWE-Lancer Diamond (https://github.com/openai/SWELancer-Benchmark). By mapping model performance to monetary value, we hope SWE-Lancer enables greater research into the economic impact of AI model development.

Large language models (LLMs) have shown increasing capacity at planning and executing a high-level goal in a live computer environment (e.g. MiniWoB++). To perform a task, recent works often require a model to learn from trace examples of the task via either supervised learning or few/many-shot prompting. Without these trace examples, it remains a challenge how an agent can autonomously learn and improve its control on a computer, which limits the ability of an agent to perform a new task. We approach this problem with a zero-shot agent that requires no given expert traces. Our agent plans for executable actions on a partially observed environment, and iteratively progresses a task by identifying and learning from its mistakes via self-reflection and structured thought management. On the easy tasks of MiniWoB++, we show that our zero-shot agent often outperforms recent SoTAs, with more efficient reasoning. For tasks with more complexity, our reflective agent performs on par with prior best models, even though previous works had the advantages of accessing expert traces or additional screen information.

This thesis has been divided into six chapters namely: Introduction, Karaka Model and it impacts on Dependency Parsing, LT Resources for Bhojpuri, English-Bhojpuri SMT System: Experiment, Evaluation of EB-SMT System, and Conclusion. Chapter one introduces this PhD research by detailing the motivation of the study, the methodology used for the study and the literature review of the existing MT related work in Indian Languages. Chapter two talks of the theoretical background of Karaka and Karaka model. Along with this, it talks about previous related work. It also discusses the impacts of the Karaka model in NLP and dependency parsing. It compares Karaka dependency and Universal Dependency. It also presents a brief idea of the implementation of these models in the SMT system for English-Bhojpuri language pair.