ICLR 2026 论文列表

Mistake detection in procedural tasks is essential for building intelligent systems that support learning and task execution. Existing approaches primarily analyze how an action is performed, while overlooking what it produces, i.e., the \textbf{action effect}. Yet many errors manifest not in the execution itself but in the resulting outcome, such as an unintended object state or incorrect spatial arrangement. To address this gap, we propose Action Effect Modeling (AEM), a unified framework that jointly captures action execution and its outcomes through a probabilistic formulation. AEM first identifies the outcome of an action by selecting the most informative effect frame based on semantic relevance and visual quality. It then extracts complementary cues from visual grounding and symbolic scene graphs, aligning them in a shared latent space to form robust effect-aware representations. To detect mistakes, we further design a prompt-based detector that incorporates task-specific prompts and aligns each action segment with its intended execution semantics. Our approach achieves state-of-the-art performance on the EgoPER and CaptainCook4D benchmarks under the challenging one-class classification (OCC) setting. These results demonstrate that modeling both execution and outcome yields more reliable mistake detection, and highlight the potential of effect-aware representations to benefit a broader range of downstream applications.

As large language models (LLMs) are increasingly deployed worldwide, ensuring their fair and comprehensive cultural understanding is important. However, LLMs exhibit cultural bias and limited awareness of underrepresented cultures, while the mechanisms underlying their cultural understanding remain underexplored. To fill this gap, we conduct a neuron-level analysis to identify neurons that drive cultural behavior, introducing a gradient-based scoring method with additional filtering for precise refinement. We identify both culture-general neurons contributing to cultural understanding regardless of cultures, and culture-specific neurons tied to an individual culture. These neurons account for less than 1% of all neurons and are concentrated in shallow to middle MLP layers. We validate their role by showing that suppressing them substantially degrades performance on cultural benchmarks (by up to 30%), while performance on general natural language understanding (NLU) benchmarks remains largely unaffected. Moreover, we show that culture-specific neurons support knowledge of not only the target culture, but also related cultures. Finally, we demonstrate that training on NLU benchmarks can diminish models' cultural understanding when we update modules containing many culture-general neurons. These findings provide insights into the internal mechanisms of LLMs and offer practical guidance for model training and engineering.

Large language models (LLMs) have significantly advanced protein representation learning. However, their capacity to interpret and design antibodies through natural language remains limited. To address this challenge, we present AFD-Instruction, the first large-scale instruction dataset with functional annotations tailored to antibodies. This dataset encompasses two key components: antibody understanding, which infers functional attributes directly from sequences, and antibody design, which enables de novo sequence generation under functional constraints. These components provide explicit sequence-function alignment and support antibody design guided by natural language instructions. Extensive instruction-tuning experiments on general-purpose LLMs demonstrate that AFD-Instruction consistently improves performance across diverse antibody-related tasks. By linking antibody sequences with textual descriptions of function, AFD-Instruction establishes a new foundation for advancing antibody modeling and accelerating therapeutic discovery.

Although Video Large Language Models (VLLMs) have shown remarkable capabilities in video understanding, they are required to process high volumes of visual tokens, causing significant computational inefficiency. Existing VLLMs acceleration frameworks usually compress spatial and temporal redundancy independently, which overlooks the spatiotemporal relationships, thereby leading to suboptimal spatiotemporal compression. The highly correlated visual features are likely to change in spatial position, scale, orientation, and other attributes over time due to the dynamic nature of video. Building on this insight, we introduce FlashVID, a training-free inference acceleration framework for VLLMs. Specifically, FlashVID utilizes Attention and Diversity-based Token Selection (ADTS) to select the most representative tokens for basic video representation, then applies Tree-based Spatiotemporal Token Merging (TSTM) for fine-grained spatiotemporal redundancy elimination. Extensive experiments conducted on three representative VLLMs across five video understanding benchmarks demonstrate the effectiveness and generalization of our method. Notably, by retaining only $\textbf{10}$% of visual tokens, FlashVID preserves $\textbf{99.1}$% of the performance of LLaVA-OneVision. Consequently, FlashVID can serve as a training-free and plug-and-play module for extending long video frames, which enables a $\textbf{10$\times$}$ increase in video frame input to Qwen2.5-VL, resulting in a relative improvement of $\textbf{8.6}$% within the same computational budget. Code is available at https://github.com/Fanziyang-v/FlashVID.

Large Language Model (LLM) agents are reshaping the game industry, by enabling more intelligent and human-preferable characters. Yet, current game benchmarks fall short of practical needs: they lack evaluations of diverse LLM capabilities across various game genres, studies of agentic modules crucial for complex gameplay, and fine-tuning datasets to adapt pre-trained LLMs into gaming agents. To fill these gaps, we present Orak, a benchmark for training and evaluating LLM agents across 12 popular video games spanning all major genres. Using a plug-and-play interface built on Model Context Protocol (MCP), Orak supports systematic and reproducible studies of agentic modules in varied game scenarios. We further release a fine-tuning dataset of expert LLM gameplay trajectories spanning multiple genres, turning general LLMs into effective game agents. Orak offers a comprehensive evaluation framework, including game leaderboards, LLM battle arenas, and in-depth analyses of input modality, agentic strategies, and fine-tuning effects, establishing a foundation towards versatile gaming agents. Code is available at https://anonymous.4open.science/r/Orak-5013/.

Fingerspelling plays a vital role in sign languages, particularly for conveying names, technical terms, and words not found in the standard lexicon. However, evaluation of two-handed fingerspelling detection and recognition is rarely addressed in existing sign language datasets—particularly for BANZSL (British, Australian, and New Zealand Sign Language), which share a common two-handed manual alphabet. To bridge this gap, we curate a large-scale dataset, dubbed BANZ-FS, focused on BANZSL fingerspelling in both controlled and real-world environments. Our dataset is compiled from three distinct sources: (1) live sign language interpretation in news broadcasts, (2) controlled laboratory recordings, and (3) diary vlogs from online platforms and social media. This composition enables BANZ-FS to capture variations in signing tempos and fluency across diverse signers and contents. Each instance in BANZ-FS is carefully annotated with multi-level alignment: video ↔ subtitles, video ↔ fingerspelled letters, and video ↔ target lexicons. In total, BANZ-FS includes over 35,000 video-aligned fingerspelling instances. Importantly, BANZ-FS highlights the unique linguistic and visual challenges posed by two-handed fingerspelling, including handshape coarticulation, self-occlusion, intra-letter variation, and rapid inter-letter transitions. We benchmark state-of-the-art models on the key tasks, including fingerspelling detection, isolated fingerspelling recognition, and fingerspelling recognition in context. Experimental results show that BANZ-FS presents substantial challenges while offering rich opportunities for BANZSL understanding and broader sign language technology. The dataset and benchmarks are available at BANZ-FS.

Matrix operations, such as linear solves, eigendecompositions, and log determinants, are foundational building blocks for any number of downstream applications. Therefore, any broadly capable learning system should be able to effectively approximate these operations in its internal representation. Accordingly, there is great motivation to study transformers for linear algebra --- for if transformers cannot even semi-competently perform matrix operations, then we cannot expect them to form a basis for a generally intelligent system. We demonstrate that current techniques developing transformers for linear algebra have striking failure modes, prohibitive scaling, and particularly poor out-of-distribution generalization to other matrix distributions, and matrices of different sizes. Investigating further, we find that current transformer approaches operate as statistical interpolators, rather than discovering algorithms that will generalize to matrices from other distributions. Based on our understanding of these limitations, we develop a sequence of interventions that substantially improve scaling and performance, including matrix embeddings through a learnable projection, linear attention, looping, and a data pre-training distribution of structured matrices. We term the resulting method the \emph{RangeFormer}, which we show has significantly improved scaling and performance on challenging OOD matrices from the \emph{matrix market}. Moreover, with RangeFormer we show for the first time that transformers can be successfully applied to downstream tasks that involve iterative matrix operations, including Gaussian process learning, and improving the sampling distribution of randomized methods.

Multimodal large language models have demonstrated impressive capabilities in visual-language understanding, particularly in offline video tasks. More recently, the emergence of online video modeling has introduced early forms of active interaction. However, existing models, typically limited to tens of minutes, are not yet capable of all-day proactive understanding over ultra-long video streams. They struggle to maintain long-term context online, as they suffer from token accumulation and lack scalable memory mechanisms. These limitations hinder critical tasks such as reminding users that medication was taken hours earlier—an ability that exemplifies the shift from reactive to memory-oriented assistants with long-term reasoning. To bridge this gap, we present Memento, the first proactive vision-language framework for ultra-long streaming video. To avoid token growth and support scalable long-duration understanding, we introduce Dynamic Memory and Query-related Memory Selection, enabling sparse memory retention and efficient retrieval. To address the training challenges of memory-based modeling, we propose Step-Aware Memory Attention, which aligns memory access with temporal steps for stable supervision. To support both training and evaluation of active, long-term behavior, we construct Memento-54K and MementoBench, a dataset-benchmark suite covering diverse tasks on text, object, and action across video streams up to 7 hours. Experiments demonstrate that Memento achieves superior performance, paving the way toward reliable all-day proactive video assistants.

Large language model (LLM) alignment is typically achieved through learning from human preference comparisons, making the quality of preference data critical to its success. Existing studies often pre-process raw training datasets to identify valuable preference pairs using external reward models or off-the-shelf LLMs, achieving improved overall performance but rarely examining whether individual, selected data point is genuinely beneficial. We assess data quality through individual influence on validation data using our newly proposed truncated influence function (TIF), which mitigates the over-scoring present in traditional measures and reveals that preference data quality is inherently a property of the model. In other words, a data pair that benefits one model may harm another. This leaves the need to improve the preference data selection approaches to be adapting to specific models. To this end, we introduce a set of candidate scoring functions (SFs) that are computationally simpler than TIF and positively correlated with it. These functions are also model dependent and can serve as potential indicators of individual data quality for preference data selection. Furthermore, we observe that these SFs inherently exhibit errors when compared to TIF. To this end, we combine them to offset their diverse error sources, resulting in a simple yet effective data selection rule that enables the models to achieve a more precise selection of valuable preference data. We conduct experiments across diverse alignment benchmarks and various LLM families, with results demonstrating that better alignment performance can be achieved using less data, showing the generality and robustness of our findings and new methods.

Imitation learning for manipulation has a well-known data scarcity problem. Unlike natural language and 2D computer vision, there is no Internet-scale corpus of data for dexterous manipulation. One appealing option is egocentric human video, a passively scalable data source. However, existing large-scale datasets such as Ego4D do not have native hand pose annotations and do not focus on object manipulation. To this end, we use Apple Vision Pro to collect EgoDex: the largest and most diverse dataset of dexterous human manipulation to date. EgoDex has 829 hours of egocentric video with paired 3D hand and finger tracking data collected at the time of recording, where multiple calibrated cameras and on-device SLAM can be used to precisely track the pose of every joint of each hand. The dataset covers a wide range of diverse manipulation behaviors with everyday household objects in 194 different tabletop tasks ranging from tying shoelaces to folding laundry. Furthermore, we train and systematically evaluate imitation learning policies for hand trajectory prediction on the dataset, introducing metrics and benchmarks for measuring progress in this increasingly important area. By releasing this large-scale dataset, we hope to push the frontier of robotics, computer vision, and foundation models. EgoDex is publicly available for download.

The development of video diffusion models unveils a significant challenge: the substantial computational demands. To mitigate this challenge, we note that the reverse process of diffusion exhibits an inherent entropy-reducing nature. Given the inter-frame redundancy in video modality, maintaining full frame rates in high-entropy stages is unnecessary. Based on this insight, we propose TPDiff, a unified framework to enhance training and inference efficiency. By dividing diffusion into several stages, our framework progressively increases frame rate along the diffusion process with only the last stage operating on full frame rate, thereby optimizing computational efficiency. To train the multi-stage diffusion model, we introduce a dedicated training framework: stage-wise diffusion. By solving the partitioned probability flow ordinary differential equations (ODE) of diffusion under aligned data and noise, our training strategy is applicable to various diffusion forms and further enhances training efficiency. Comprehensive experimental evaluations validate the generality of our method, demonstrating 50% reduction in training cost and 1.5x improvement in inference efficiency.

GPT has shown its remarkable success in natural language processing. However, the language sequence is not sufficient to describe spatial-temporal details in the visual world. Alternatively, the video sequence is good at capturing such details. Motivated by this fact, we propose a concise Video-GPT in this paper by treating video as new language for visual world modeling. By analogy to next token prediction in GPT, we introduce a novel next clip diffusion paradigm for pretraining Video-GPT. Different from the previous works, this distinct paradigm allows Video-GPT to tackle both short-term generation and long-term prediction, by autoregressively denoising the noisy clip according to the clean clips in the history. Extensive experiments show our Video-GPT achieves the state-of-the-art performance on video prediction, which is the key factor towards world modeling (Physics-IQ Benchmark: Video-GPT 34.97 vs. Kling 23.64 vs. Wan 20.89). Moreover, it can be well adapted on 6 mainstream video tasks in both video generation and understanding, showing its great generalization capacity in downstream.

While protein language models (PLMs) are one of the most promising avenues of research for future de novo protein design, the way in which they transform sequences to hidden representations, as well as the information encoded in such representations is yet to be fully understood. Several works have attempted to propose interpretability tools for PLMs, but they have focused on understanding how individual sequences are transformed by such models. Therefore, the way in which PLMs transform the whole space of sequences along with their relations is still unknown. In this work we attempt to understand this transformed space of sequences by identifying protein structure and representation with square-root velocity (SRV) representations and graph filtrations. Both approaches naturally lead to a metric space in which pairs of proteins or protein representations can be compared with each other. We analyze different types of proteins from the SCOP dataset and show that the Karcher mean and effective dimension of the SRV shape space follows a non-linear pattern as a function of the layers in ESM2 models of different sizes. Furthermore, we use graph filtrations as a tool to study the context lengths at which models encode the structural features of proteins. We find that PLMs preferentially encode immediate as well as local relations between amino acids, but start to degrade for larger context lengths. The most structurally faithful encoding tends to occur close to, but before the last layer of the models, indicating that training a folding model ontop of these layers might lead to improved folding performance.

Understanding the negation in a sentence is an important part of compositional understanding and logic in natural language. Many practical AI applications, such as autonomous driving, include precise instruction with negations. For example, following instruction to an AI assistant ”locate a parking spot without a vehicle” requires the assistant to not confuse between presence and absence of vehicles. Al- though joint embedding-based Vision Language Models (VLMs) like CLIP have revolutionized multi-modal tasks, they struggle to interpret negation. To address this limitation, recently many works proposed to solve the problem through a data- centric approach by introducing additional datasets with hard-negative samples for both image and text data. Contrary to these approaches, we present a zero-shot approach to tackle the negation understanding problem. We probe the properties of CLIP text embeddings and show that they follow compositional arithmetic op- erations, which allow the addition or removal of semantic information directly in the embedding space. We then present a rule-based approach to extract negated text from given caption and then use it to explicitly remove corresponding se- mantic information from original embedding, improving negation understanding in VLMs. Our approach does not require expensive training process to induce negation understanding into the model, and achieves the state-of-the-art perfor- mance on popular benchmark for negation understanding. We improve baseline CLIP model performance on NegBench from 25.5% to 67.0% for MCQ and from 50.9% to 56.1% for retrieval tasks. Even NegCLIP model which is fine-tuned on negtion datasets, our approach boosts its MCQ accuracy from 54.03% to 66.22% and retrieval accuracy from 59.25% to 60.1% showing strong performance.

Recent advances have shown that scaling test-time computation enables large language models (LLMs) to solve increasingly complex problems across diverse domains. One effective paradigm for test-time scaling (TTS) involves LLM generators producing multiple solution candidates, with LLM verifiers assessing the correctness of these candidates without reference answers. In this paper, we study generative verifiers, which perform verification by generating chain-of-thought (CoT) reasoning followed by a binary verdict. We systematically analyze verification dynamics across three dimensions -- problem difficulty, generator capability, and verifier generation capability -- through empirical studies on 12 benchmarks across mathematical reasoning, knowledge, and natural language reasoning tasks using 14 open-source models (2B to 72B parameter range) and GPT-4o. Our experiments reveal three key findings about verification effectiveness: (1) Easy problems allow verifiers to more reliably certify correct responses; (2) Weak generators produce errors that are easier to detect than strong generators; (3) Verification ability is generally correlated with the verifier's own problem-solving capability, but this relationship varies with problem difficulty. These findings reveal opportunities for optimizing basic verification strategies in TTS applications. First, given the same verifier, some weak generators can nearly match stronger ones in post-verification TTS performance (e.g., the Gemma2-9B to Gemma2-27B performance gap shrinks by 75.7%). Second, we identify cases where strong verifiers offer limited advantages over weak ones, as both fail to provide meaningful verification gains, suggesting that verifier scaling alone cannot overcome fundamental verification challenges.

Advancing machine intelligence requires developing the ability to perceive across multiple modalities, much as humans sense the world. We introduce OmniVinci, an initiative to build a strong, open-source, omni-modal LLM. We carefully study the design choices across model architecture and data curation. For model architecture, we present three key innovations: (i) OmniAlignNet for strengthening alignment between vision and audio embeddings in a shared omni-modal latent space; (ii) Temporal Embedding Grouping for capturing relative temporal alignment between vision and audio signals; and (iii) Constrained Rotary Time Embedding for encoding absolute temporal information in omni-modal embeddings. We introduce a curation and synthesis pipeline that generates 24M single-modal and omni-modal conversations. We find that modalities reinforce one another in both perception and reasoning. Our model, OmniVinci, improves over Qwen2.5-Omni with +19.05 on DailyOmni (cross-modal understanding), +1.7 on MMAR (audio), and +3.9 on Video-MME (vision), while using just 0.2T training tokens — a 6× reduction compared to Qwen2.5-Omni’s 1.2T. We finally demonstrate omni-modal advantages in downstream applications spanning robotics, medical AI, and smart factory.

Sparse autoencoders (SAEs) have recently emerged as a powerful tool for interpreting the features learned by large language models (LLMs). By reconstructing features with sparsely activated networks, SAEs aim to recover complex superposed polysemantic features into interpretable monosemantic ones. Despite their wide applications, it remains unclear under what conditions SAEs can fully recover the ground truth monosemantic features from the superposed polysemantic ones. In this paper, we provide the first theoretical analysis with a closed-form solution for SAEs, revealing that they generally fail to fully recover the ground truth monosemantic features unless the ground truth features are extremely sparse. To improve the feature recovery of SAEs in general cases, we propose a reweighting strategy targeting at enhancing the reconstruction of the ground truth monosemantic features instead of the observed polysemantic ones. We further establish a theoretical weight selection principle for our proposed weighted SAE (WSAE). Experiments across multiple settings validate our theoretical findings and demonstrate that our WSAE significantly improves feature monosemanticity and interpretability.

Task vector is a compelling mechanism for accelerating inference in in-context learning (ICL) by distilling task-specific information into a single, reusable representation. Despite their empirical success, the underlying principles governing their emergence and functionality remain unclear. This work proposes the *Task Vectors as Representative Demonstrations* conjecture, positing that task vectors encode single in-context demonstrations distilled from the original ones. We provide both theoretical and empirical support for this conjecture. First, we show that task vectors naturally emerge in linear transformers trained on triplet-formatted prompts through loss landscape analysis. Next, we predict the failure of task vectors in representing high-rank mappings and confirm this on practical LLMs. Our findings are further validated through saliency analyses and parameter visualization, suggesting an enhancement of task vectors by injecting multiple ones into few-shot prompts. Together, our results advance the understanding of task vectors and shed light on the mechanisms underlying ICL in transformer-based models.

Mixture-of-Experts (MoE) has been a prevalent method for scaling up large language models at a reduced computational cost. Despite its effectiveness, the routing mechanism of MoE still lacks a clear understanding from the perspective of cross-layer mechanistic interpretability. We propose a light-weight methodology at which we can break down the routing decision for MoE to contribution of model components, in a recursive fashion. We use our methodology to dissect the routing mechanism by decomposing the input of routers into model components. We study how different model components contribute to the routing in different widely used open models. Our findings on four different LLMs reveal common patterns such as: a) MoE layer outputs contribute more than attention layer outputs to the routing decisions of latter layers, b) \emph{MoE entanglement} at which MoE firing up in layers consistently correlate with firing up of MoE in latter layers, and c) some components can persistently influence the routing in many following layers. Our study also includes findings on how different models have different patterns when it comes to long range and short range inhibiting/promoting effects that components can have over MoE in latter layers. Our results indicate importance of quantifying the impact of components across different layers on MoE to understand the mechanism of routing.