ICLR 2026 论文列表

Current zero-shot unsupervised segmentation methods based on normalized cuts (NCut) face three key limitations. First, they rely on recursive bipartitions with repeated eigen-decompositions, making them prohibitively expensive at scale. Second, each split requires spectral relaxation followed by rounding, introducing layers of approximation where the final partition may diverge from the true NCut objective. Third, existing heuristics lack convergence guarantees, and recursive bipartitioning offers no principled assurance of producing a stable $K$-way segmentation. We propose \textbf{Falcon}, a proximal-gradient solver that directly optimizes the discrete $K$-way NCut objective without spectral relaxation. We prove linear convergence in the number of tokens. Falcon computes closed-form gradient scores weighted by cluster volumes and performs row-wise one-hot proximal updates stabilized by inertia. A monotone backtracking scheme adaptively tunes the proximal parameter, ensuring non-decreasing NCut values. This design preserves discrete feasibility, removes repeated eigen-decomposition, and guarantees convergence under the \text{Kurdyka--\L{}ojasiewicz} framework. Across six benchmarks, Falcon outperforms the strongest official baseline (DiffCut) by wide margins, e.g., +13.2 mIoU on VOC, +27.7 on COCO-Object, and +3.1 on Cityscapes, while remaining competitive on Pascal Context. It also runs up to an order of magnitude faster than recursive NCut. By pairing pretrained foundation models with a principled NCut solver, Falcon sets a new state of the art across six benchmarks and achieves the best performance on 17 of 18 benchmark–encoder pairs, underscoring both its robustness and its generality in bridging the gap between unsupervised and supervised segmentation.

Transferring the optimal learning rate from small to large neural networks can enable efficient training at scales where hyperparameter tuning is otherwise prohibitively expensive. To this end, the Maximal Update Parameterization (µP) proposes a learning rate scaling designed to keep the update dynamics of internal representations stable across different model widths. However, the scaling rules of µP rely on strong assumptions, particularly about the geometric alignment of a layer’s inputs with both its weights and gradient updates. In this large-scale empirical investigation, we show that these assumptions hold only briefly at the start of training in the practical setups where learning rate transfer is most valuable, such as LLM training. For the remainder of training it is weight decay rather than µP that correctly stabilizes the update dynamics of internal representations across widths, facilitating learning rate transfer. This suggests µP's scaling primarily acts as a form of implicit learning rate warmup, allowing us to largely replace it with modified warmup schedules. Together these findings fundamentally challenge prevailing beliefs about learning rate transfer and can explain empirical observations such as why µP requires the independent weight decay variant for good transfer.

The pursuit of human-like conversational agents has long been guided by the Turing test. For modern speech-to-speech (S2S) systems, a critical yet unanswered question is whether they can converse like humans. To tackle this, we conduct the first Turing test for S2S systems, collecting 2,968 human judgments on dialogues between 9 state-of-the-art S2S systems and 28 human participants. Our results deliver a clear finding: no existing evaluated S2S system passes the test, revealing a significant gap in human-likeness. To diagnose this failure, we develop a fine-grained taxonomy of 18 human-likeness dimensions and crowd-annotate our collected dialogues accordingly. Our analysis shows that the bottleneck is not semantic understanding but stems from paralinguistic features, emotional expressivity, and conversational persona. Furthermore, we find that off-the-shelf AI models perform unreliably as Turing test judges. In response, we propose an interpretable model that leverages the fine-grained human-likeness ratings and delivers accurate and transparent human-vs-machine discrimination, offering a powerful tool for automatic human-likeness evaluation. Our work establishes the first human-likeness evaluation for S2S systems and moves beyond binary outcomes to enable detailed diagnostic insights, paving the way for human-like improvements in conversational AI systems.

In this paper, we propose Label-Aware Noise Elimination (LANE), a new approach that improves the robustness of deep learning models when trained under increased label noise in fine-grained text classification. LANE leverages the semantic relations between classes and monitors the training dynamics of the model on each training example to dynamically lower the importance of training examples that are perceived to have noisy labels. We test the effectiveness of LANE in fine-grained text classification and benchmark our approach on a wide variety of datasets with various number of classes and various amounts of label noise. LANE considerably outperforms strong baselines on all datasets, obtaining significant improvements ranging from an average improvement of 2.4% in F1 on manually annotated datasets to a considerable average improvement of 4.5% F1 on datasets with higher levels of label noise. We carry out comprehensive analyses of LANE and identify the key components that lead to its success.

Recent advances in large language models have laid the foundation for multimodal LLMs (MLLMs), which unify text, speech, and vision within a single framework. As these models are rapidly evolving toward general-purpose instruction following across diverse and complex tasks, a key frontier is evaluating their crosslingual and multimodal capabilities over both short- and long-form inputs. However, existing benchmarks fall short in evaluating these dimensions jointly: they are often limited to English, mostly focus on a single modality at a time, rely on short-form inputs, or lack human annotations--hindering comprehensive assessment of model performance across languages, modalities, and task complexity. To address these gaps, we introduce MCIF (Multimodal Crosslingual Instruction Following), the first crosslingual human-annotated benchmark based on scientific talks on NLP and beyond. MCIF evaluates instruction following in crosslingual, multimodal settings over different input lengths and spans four macro-tasks: recognition, translation, question answering, and summarization. It covers three core modalities (speech, vision, and text) and four diverse languages (English, German, Italian, and Chinese), fully aligned across all dimensions. This parallel design enables a systematic evaluation of MLLMs' abilities to interpret instructions across languages and effectively integrate multimodal contextual information. Our benchmarking and analysis of 23 models highlight universal challenges across modalities and tasks, indicating substantial room for improvement in future MLLMs development. MCIF is released under CC-BY 4.0 license to promote open research.

Transformer-based architectures traditionally employ softmax to compute attention weights, which produces dense distributions over all tokens in a sequence. While effective in many settings, this density has been shown to be detrimental for tasks that demand precise focus on fixed-size patterns: as sequence length increases, non-informative tokens accumulate attention probability mass, leading to dispersion and representational collapse. We show in this paper that dynamically sparse attention mechanisms using $\alpha$-entmax can avoid these issues, due to their ability to assign exact zeros to irrelevant tokens. Furthermore, we introduce Adaptive-Scalable Entmax (ASEntmax), which endows $\alpha$-entmax with a learnable temperature parameter, allowing the attention distribution to interpolate between sparse (pattern-focused) and dense (softmax-like) regimes. Our empirical evaluation on synthetic tasks and language modeling demonstrates that ASEntmax substantially outperforms softmax, scalable softmax, and fixed-temperature $\alpha$-entmax baselines, achieving up to 1000$\times$ length extrapolation on synthetic benchmarks and superior long-context generalization on language modeling while preserving short-context performance, including better perplexity trends and higher retrieval accuracies at 8$\times$ training length.

While text-to-3D generation has attracted growing interest, existing methods often struggle to produce 3D assets that align well with human preferences. Current preference alignment techniques for 3D content typically rely on hardly-collected preference-paired multi-view 2D images to train 2D reward models, when then guide 3D generation — leading to geometric artifacts due to their inherent 2D bias. To address these limitations, we construct 3D-MeshPref, the first large-scale unpaired 3D preference dataset, featuring diverse 3D meshes annotated by a large language model and refined by human evaluators. We then develop RewardCS, the first reward model trained directly on unpaired 3D-MeshPref data using a novel Cauchy-Schwarz divergence objective, enabling effective learning of human-aligned 3D geometric preferences without requiring paired comparisons. Building on this, we propose DreamCS, a unified framework that integrates RewardCS into text-to-3D pipelines — enhancing both implicit and explicit 3D generation with human preference feedback. Extensive experiments show DreamCS outperforms prior methods, producing 3D assets that are both geometrically faithful and human-preferred. Code and models will be released publicly.

Knowledge Editing has emerged as a promising solution for efficiently updating embedded knowledge in large language models (LLMs). While existing approaches demonstrate effectiveness in integrating new knowledge and preserving the original capabilities of LLMs, they fail to maintain fine-grained irrelevant knowledge facts that share the same subject as edited knowledge but differ in relation and object. This challenge arises because subject representations inherently encode multiple attributes, causing the target and fine-grained irrelevant knowledge to become entangled in the representation space, and thus vulnerable to unintended alterations during editing. To address this, we propose DiKE, a novel approach that Disentangles Knowledge representations for LLM Editing (DiKE). DiKE consists of two key components: a Knowledge Representation Disentanglement (KRD) module that decomposes the subject representation into target-knowledgerelated and -unrelated components, and a Disentanglement-based Knowledge Edit (DKE) module that updates only the target-related component while explicitly preserving the unrelated one. We further derive a closed-form, rank-one parameter update based on matrix theory to enable efficient and minimally invasive edits. To rigorously evaluate fine-grained irrelevant knowledge preservation, we construct FINE-KED, a new benchmark comprising fine-grained irrelevant knowledge at different levels of relational similarity to the edited knowledge. Extensive experiments across multiple LLMs demonstrate that DiKE substantially improves fine-grained irrelevant knowledge preservation while maintaining competitive general editing performance.

Reconstruction techniques, such as 3D Gaussian Splatting (3DGS), are increasingly used for generating scenarios in autonomous driving system (ADS) research. Existing 3DGS-based works for autonomous driving scenario generation have, through various optimizations, achieved high visual similarity in reconstructed scenes. However, this route is built on a strong assumption: that higher scene similarity directly translates into better preservation of ADS behaviour. Unfortunately, this assumption has not been effectively validated, and ADS behaviour is more closely related to the objects within the field of view rather than the global image. Thus, we focus on the perception module—the entry point of ADS. Preliminary experiments reveal that although current methods can produce reconstructions with high overall similarity, they often fail to ensure that the perception module outputs remain consistent with those obtained from the original images. Such a limitation can significantly harm the applicability of reconstruction in the ADS domain. To address this gap, we propose two complementary solutions: a perception-aligned loss, which directly leverages the output differences between reconstructed and ground truth images during the training process; and an object zone quality loss, which specifically reinforces the training on object locations identified by the perception model on ground-truth images. Experiments demonstrate that both of our methods improve the ability of reconstructed scenes to preserve consistency between the perception module outputs and those from the ground-truth inputs.

Inference-time scaling enhances a model’s reasoning by extending its chain-of-thought (CoT). However, existing approaches typically rely on a single policy trained with outcome-reward reinforcement learning (RL), which often suffers from long-horizon plan failures, i.e., the implicit plan drifts away from any valid strategy. This problem is particularly severe for smaller language models (LMs) with long CoTs due to their limited capacity. To address this, we propose Multi-Level Reasoning (MLR), which reformulates long-CoT generation as a two-level stochastic process. Specifically, MLR employs two policies: a high-level planner that generates step descriptors (abstract subgoals) and a low-level executor that produces detailed content conditioned on these descriptors. The planner then generates the next subgoal based on the summarized current step, forming an alternating plan–execute loop. To maintain scalability, we adopt a minimal design, where the base model serves as the low-level policy and a lightweight LoRA module implements the high-level policy. For training, we observe that outcome-reward RL is inefficient and weakly informative for long trajectories (e.g., those exceeding 4K tokens). To overcome this, we introduce online Step-DPO, a process-level preference optimization scheme that leverages Twisted Sequential Monte Carlo (TSMC) to provide scalable stepwise supervision. This yields more effective training, improved stability, and higher accuracy. Extensive experiments on challenging math, science, and logical reasoning benchmarks show that, with only 10% SFT data and 5% of preference data, MLR outperforms both the DeepSeek-R1 distillation and the outcome-reward RL baselines across multiple base models and tasks. More importantly, MLR exhibits slower performance degradation on long-horizon reasoning, demonstrating stronger robustness under extended CoT generation.

Vision-language-action models (VLAs) have garnered significant attention for their potential in advancing robotic manipulation. However, previous approaches predominantly rely on the general comprehension capabilities of vision-language models (VLMs) to generate action signals, often overlooking the rich temporal and causal structure embedded in visual observations. In this paper, we present UniVLA, a unified and native multimodal VLA model that autoregressively models vision, language, and action signals as discrete token sequences. This tokenized formulation naturally supports flexible multimodal task learning, particularly from large-scale video data, and further demonstrates that generative vision supervision can significantly enhance visual understanding. By incorporating world modeling during post-training, UniVLA captures causal dynamics from videos, facilitating effective transfer to downstream policy learning—especially for long-horizon tasks. Our approach sets new state-of-the-art results across several widely used simulation benchmarks, including CALVIN, LIBERO, and Simplenv-Bridge, substantially outperforming prior methods. For example, UniVLA achieves 95.5% average success rate on LIBERO benchmark, surpassing π₀-FAST's 85.5%. We further demonstrate its broad applicability through experiments on real-world ALOHA manipulation tasks and autonomous driving scenarios.

We present WinT3R, a feed-forward reconstruction model capable of online prediction of precise camera poses and high-quality point maps. Previous methods suffer from a trade-off between reconstruction quality and real-time performance. To address this, we first introduce a sliding window mechanism that ensures sufficient information exchange among frames within the window, thereby improving the quality of geometric predictions without introducing a large amount of extra computation. In addition, we leverage a compact representation of cameras and maintain a global camera token pool, which enhances the reliability of camera pose estimation without sacrificing efficiency. These designs enable WinT3R to achieve state-of-the-art performance in terms of online reconstruction quality, camera pose estimation, and reconstruction speed, as validated by extensive experiments on diverse datasets.

It has long been conjectured and empirically observed that neural networks tend to preserve clustering structure. This paper formalizes this conjecture. Specifically, we establish precise conditions for cluster structure preservation and derive bounds to quantify its extent. Through this analysis we are able to show that certain neural networks are learning parameters that preserve the clustering structure of the original data in their embeddings, without the need to impose mechanisms to promote this behavior. Extensive numerical analysis and experiments validate our results. Our findings offer deeper insight into neural network behavior, explaining why certain data types (such as images, audio, and text) benefit more from deep learning. Beyond theory, our findings guide better initialization, feature encoding, and regularization strategies.

Large language models (LLMs) rarely admit uncertainty, often producing fluent but misleading answers, rather than abstaining (i.e., refusing to answer). This weakness is even evident in temporal question answering (QA), where models frequently ignore time-sensitive evidence and conflate facts across different time-periods. In this paper, we present the first empirical study of training LLMs with abstention ability while reasoning about temporal QA. Existing approaches such as calibration might be unreliable in capturing uncertainty in complex reasoning. We instead frame abstention as a teachable skill and introduce a pipeline that couples Chain-of-Thought (CoT) supervision with Reinforcement Learning (RL) guided by abstention-aware rewards. Our goal is to systematically analyze how different information types and training techniques affect temporal reasoning with abstention behavior in LLMs. Through extensive experiments studying various methods, we find that RL yields strong empirical gains on reasoning: a model initialized by Qwen2.5-1.5B-Instruct surpasses GPT-4o by 3.46% and 5.80% in Exact Match on TimeQA-Easy and Hard, respectively. Moreover, it improves the True Positive rate on unanswerable questions by 20% over a pure supervised fine-tuned (SFT) variant. Beyond performance, our analysis shows that SFT induces overconfidence and harms reliability, while RL improves prediction accuracy but exhibits similar risks. Finally, by comparing implicit reasoning cues (e.g., original context, temporal sub-context, knowledge graphs) with explicit CoT supervision, we find that implicit information provides limited benefit for reasoning with abstention. Our study provides new insights into how abstention and reasoning can be jointly optimized, providing a foundation for building more reliable LLMs.

Large language models (LLMs) face growing challenges in efficient generative inference due to the increasing memory demands of Key-Value (KV) caches, especially for long sequences. Existing eviction methods typically retain KV pairs with high attention weights but overlook the impact of attention redistribution caused by token removal, as well as the spatial-temporal dynamics in KV selection. In this paper, we propose ReST-KV, a robust KV eviction method that combines layer-wise output **Re**construction and **S**patial-**T**emporal smoothing to provide a more comprehensive perspective for the KV cache eviction task. Specifically, ReST-KV formulates KV cache eviction as an optimization problem that minimizes output discrepancies through efficient layer-wise reconstruction. By directly modeling how each token’s removal affects the model output, our method naturally captures attention redistribution effects, going beyond simplistic reliance on raw attention weights. To further enhance robustness, we design exponential moving average smoothing to handle temporal variations and an adaptive window-based mechanism to capture spatial patterns. Our method, ReST-KV, significantly advances performance on long-context benchmarks. It surpasses state-of-the-art baselines by 2.58\% on LongBench and 15.2\% on RULER. Additionally, ReST-KV consistently outperforms existing methods on Needle-in-a-Haystack and InfiniteBench, all while achieving a remarkable 10.61$\times$ reduction in decoding latency at 128k context length. The code is included in the supplementary material and is designed for easy reproduction.

The alignment of Large Language Models (LLMs) and brain activity provides a powerful framework to advance our understanding of cognitive neuroscience and artificial intelligence. In this work, we zoom into one of the fundamental units of LLMs—the transformer block—to provide the first systematic computational neuroanatomy of its internal operations and human brain acitivity during language processing. Analyzing 21 state-of-the-art LLMs across five model families, we extract and evaluate 13 distinct intermediate states per transformer block—from initial layer normalization through attention mechanisms to feed-forward networks (FFNs). Our analysis reveals three key findings: (1) The commonly used hidden states in LLMs are surprisingly suboptimal, with over 90\% of brain voxels in sensory and language regions better explained by previously unexplored intermediate computations; (2) Different computational stages within a single transformer block map to anatomically distinct brain systems, revealing an intra-block hierarchy where early attention states align with sensory cortices while later FFN states correspond to association areas—mirroring the cortical processing hierarchy; (3) Rotary Positional Embeddings (RoPE) specifically enhance alignment along the brain's auditory processing streams. Per-head queries with RoPE best explain 74\% of auditory cortex activity compared to 8\% without RoPE, providing the first neurobiological validation of this architectural component in LLMs. Building on these insights, we propose MindTransformer, a feature selection framework that learns brain-aligned representations from all intermediate states. MindTransformer achieves significant brain alignment performance, with correlation improvements in primary auditory cortex exceeding gains from 456× model scaling. Our computational neuroanatomy approach opens new directions for understanding both biological intelligence through the lens of transformer computations and artificial intelligence through principles of brain organization.

As large language models (LLMs) continue to revolutionize AI research, there is a growing interest in building large-scale brain foundation models to advance neuroscience. While most existing brain foundation models are pre-trained on time-series signals or connectome features, we propose a novel graph-based pre-training paradigm for constructing a brain graph foundation model. In this paper, we introduce the Brain Graph Foundation Model, termed BrainGFM, a unified framework that leverages graph contrastive learning and graph masked autoencoders for large-scale fMRI-based pre-training. BrainGFM is pre-trained on a diverse mixture of brain atlases with varying parcellations, significantly expanding the pre-training corpus and enhancing the model’s ability to generalize across heterogeneous fMRI-derived brain representations. To support efficient and versatile downstream transfer, we integrate both graph prompts and language prompts into the model design, enabling BrainGFM to flexibly adapt to a wide range of atlases, neurological and psychiatric disorders, and task settings. Furthermore, we employ meta-learning to optimize the graph prompts, facilitating strong generalization to previously unseen disorders under both few-shot and zero-shot learning conditions via language-guided prompting. BrainGFM is established on 27 neuroimaging datasets spanning 25 common neurological and psychiatric disorders, encompassing 2 types of brain atlases (functional and anatomical) across 8 widely used parcellations, and covering over 25,000 subjects, 60,000 fMRI scans, and a total of 400,000 graph samples aggregated across all atlases and parcellations. The code is available at https://github.com/weixinxu666/BrainGFM.

While Large Language Models have been used to produce interpretable stock forecasts, they mainly focus on analyzing textual reports but not historical price data, also known as Technical Analysis. This task is challenging as it switches between domains: the stock price inputs and outputs lie in the time-series domain, while the reasoning step should be in natural language. In this work, we introduce Verbal Technical Analysis (VTA), a novel framework that combine verbal and latent reasoning to produce stock time-series forecasts that are both accurate and interpretable. To reason over time-series, we convert stock price data into textual annotations and optimize the reasoning trace using an inverse Mean Squared Error (MSE) reward objective. To produce time-series outputs from textual reasoning, we condition the outputs of a time-series backbone model on the reasoning-based attributes. Experiments on stock datasets across U.S., Chinese, and European markets show that VTA achieves state-of-the-art forecasting accuracy, while the reasoning traces also perform well on evaluation metrics judged by industry experts.

Embodied cognition argues that intelligence arises from continuous sensorimotor interaction with the world. It raises an intriguing question: do modern vision-language models (VLMs), trained largely in a disembodied manner, exhibit signs of embodied cognition? To investigate this, we introduce **ENACT**, a benchmark that probes this question through world modeling from egocentric interaction. Grounded in a partially observable Markov decision process (POMDP) framework, **ENACT** comprises two complementary sequence reordering tasks: forward world modeling (predicting an ordered sequence of future states from actions) and inverse world modeling (inferring an ordered sequence of actions from state changes). Correctly solving these tasks indicates that the model has a solid understanding of how the environment will evolve given one's actions. Our scalable dataset contains 8,972 QA pairs derived from diverse, long-horizon household activities in the BEHAVIOR simulator. Experiments reveal a significant performance gap between state-of-the-art VLMs and humans, which widens dramatically as interaction horizons lengthen. We find that models consistently solve the inverse problem better than the forward one and exhibit strong embodied biases, showing a preference for right-handed actions and performance degradation with camera perspectives that deviate from those of human vision. Code and supplementary materials are available in our [anonymous repository](https://github.com/iclrsubmission2026/iclr-2026-submission).