Aoqian Zhang, Chunzheng Wang, Haoran Chen, Annan Tang, Cheng Xiang, Fan Shi, Shuzhi Sam Ge
Humanoids have become increasingly agile and capable in real-world environments. To operate safely in dynamic scenarios, they must rapidly evade fast-moving obstacles while maintaining stability and physically feasible whole-body motion. Learning such behaviors remains challenging due to the high-dimensional dynamics of humanoid systems and the need to jointly balance reactive obstacle avoidance, motion naturalness, and control robustness. In this work, we present Humanoid Dodgeball, a reinforcement learning (RL) framework for agile dynamic obstacle avoidance in humanoids. Our method enables highly dynamic whole-body evasive maneuvers against incoming projectiles through threat-aware reward shaping, while multiple direction-specific adversarial motion priors regularize the learned behaviors toward stable and diverse human-like motions. Extensive simulation and hardware experiments demonstrate that the proposed framework substantially improves dynamic obstacle avoidance performance across diverse obstacle speeds and launch distances, while preserving robust and physically plausible humanoid motions.
Xiangfei Sheng, Xiaofeng Pan, Leida Li, Qian Zhang, Biao Wang, Johan Wagemans, Weisi Lin
Aesthetic image cropping (AIC) aims to improve image composition and aesthetic quality by removing extraneous or distracting content. Over the past decade, AIC methods have evolved from specialized vision models to multimodal large language models(MLLMs). Reliable evaluation that reflects human preferences is essential for benchmarking AIC methods and guiding model optimization. However, the current AIC evaluation relies largely on geometry-based metrics such as intersection over union (IoU), as well as image aesthetic assessment(IAA) models. Although inconsistencies with human preferences have long been observed, the robustness and reliability of these evaluation criteria remain underexplored. To bridge this gap, we first construct CropBench, a large-scale benchmark comprising 106,995 human preference pairs for 1,019 images, with each candidate set containing the original image and candidate crops produced by 14 AIC methods. A systematic study on CropBench reveals that the existing criteria, ranging from IoU to off-the-shelf IAA assessment models, only exhibit limited alignment with human preferences. Building upon CropBench, we further propose CropReward, an MLLM-based reward model that evaluates a candidate crop by comparing with the original uncropped image and outputs a score indicating explicit aesthetic improvement or degradation relative to the original. In addition, we employ CropReward to construct CropReward-300k, the largest AIC dataset to date, and further develop CropAlign, a preference-aligned AIC model trained solely with CropReward supervision, without human-annotated crop boxes. Experimental results demonstrate that CropReward aligns with human preferences significantly better than AIC existing criteria. CropAlign also surpasses the state-of-the-art AIC methods, demonstrating the advantage of CropReward in guiding AIC model optimization.
Existing code reveals past implementation decisions but rarely explains the reasoning behind them. When coding agents implement new features or modify existing functionality, past decisions may conflict with new task instructions. This raises a fundamental question for agentic software evolution: when multiple implementations satisfy the same functional requirements, how do coding agents decide whether to follow existing code or new task instructions? We investigate this question through bidirectional, controlled experiments in purpose-built repositories, followed by validation in real-world open-source repositories. Our approach isolates three competing sources of guidance: (1) agents' default preferences, (2) repository precedents, and (3) explicit task instructions. We conduct 5,625 trials across four coding agents and six implementation pairs, with validation in four open-source repositories spanning three programming languages. We find that agents share the same default implementation choices, yet overwhelmingly follow repository precedent when it conflicts with those defaults (359/360 trials). Crucially, not all precedents carry equal influence: three supporting local files yield 96% adoption, compared with only 4% for sibling-resource files, and changing file placement alone can shift adoption from 0% to 100% at a fixed example count. When new task intent conflicts with existing code, written priority rules achieve substantially higher success than code examples alone (59/60 versus 38/60 joint successes for two agents on two implementation pairs). The effectiveness of explicit rules also extends to open-source repositories, where a harness-loaded rule raises joint success from 0/120 to 117/120. These findings motivate design-decision traceability that preserves not only implementation choices but also the rationale and priorities that should guide future changes.
Andie Huang, Ying Pang, Qian Zhang, Yihang Cheng, Lin Ma
Many LLMs now emit an explicit reasoning trace before they write code, but benchmark results for code generation and bug fixing usually report correctness alone. It is therefore unclear how much the extra reasoning contributes relative to the tokens it consumes. We examine this question with 866 Python tasks drawn from HumanEval Pro, MBPP Pro, and BugsInPy under five model settings. We look at the accuracy gain and token overhead of reasoning, the behaviors that appear inside reasoning traces, and whether a shortened trace still leads to a correct solution. Turning reasoning on in Qwen3-32B over five runs improves mean pass@1 on HumanEval Pro from 71.1% to 76.5% and mean tests pass on BugsInPy from 3.5% to 7.9%. On MBPP Pro the improvement from 68.1% to 69.5% is not statistically significant, even though the model uses a median of over 2,000 thinking tokens per task. Across models, a longer median trace does not reliably mean higher accuracy. A qualitative analysis shows that traces mostly choose and check an implementation in code generation, while in bug fixing they mostly locate the fault and plan the repair. We then truncate stored traces offline and ask Qwen3-32B with reasoning off to produce a solution from what remains. Retaining the first half of a trace yields no statistically significant drop in correctness. Stopping when the trace becomes repetitive seldom triggers, and stopping once solution code first appears benefits some models but harms others. Neither rule beats plain fixed truncation. These results suggest that more reasoning is not always better and that studies should report reasoning cost alongside correctness.
Junxing Hu, Ai Han, Haolan Zhan, Pu Wei, Zhiqian Zhang, Yuhang Guo, Jiawei Lu, Zhen Chen 等 11 位作者
Long-horizon agentic systems become long-context systems by accumulation rather than through a single long document, since every step reassembles a prompt from role instructions, tool and sub-agent descriptions, recalled memories, dialogue history, and previous tool outputs, separately for every agent. This paper studies how such accumulated context can be trained, evaluated, and managed in hierarchical multi-agent assistants. We present HiMA-Ecom, a hierarchical multi-agent benchmark for e-commerce assistants with 22.8K instances, of which 17.7K carry memory entries in context, pairing agent-specific supervised fine-tuning data with system-level supervision for end-to-end multi-agent reinforcement learning. Building upon it, HiMA-R1 keeps the context budget of a trajectory under control through Variance-Reduction Group Relative Policy Optimization (VR-GRPO), which samples along an initial trajectory so that sampling cost grows additively with trajectory length and updates the agents with the largest reward variance, together with an efficiency reward on trajectory length and a memory evolution mechanism that repurposes GRPO rewards as cost-free signals for what stays in context. Agents built upon 3B and 7B open-source models match DeepSeek-R1 and surpass DeepSeek-V3 by 6.0 points on average, while reward-driven memory reduces the update steps needed for convergence by roughly 55%.
Miner extractable value (MEV) poses a persistent challenge for blockchain ecosystems. It arises from the fact that a block builder can exploit its unilateral control over transaction inclusion and ordering to extract profit, potentially at users' expense. Recent work has explored consensus-aided mechanism design to mitigate MEV. This approach combines generic consensus-layer guarantees with application-layer mechanisms that use these guarantees to address application-specific MEV problems. Prior work exemplifies this approach for automated market makers (AMMs). When the consensus layer offers fair sequencing, which orders transactions according to their arrival times, their proposed mechanism achieves incentive compatibility (IC) and weak local efficiency (wLE). Informally, IC rules out profitable strategic deviations by users and builders. Meanwhile, wLE ensures that after execution, any order willing to trade at the initial price cannot benefit from further trade with the pool at the ending marginal price. In this work, we resolve the main open question left by prior works: whether IC and wLE can be achieved under plain consensus (e.g., current Ethereum) without fair sequencing. We prove that no deterministic, anonymous batch AMM can achieve both properties. The impossibility holds even when the consensus layer guarantees censorship resilience and block space is unlimited, establishing the first separation between censorship resilience and fair sequencing in the context of decentralized mechanism design. Our impossibility proof combines three strategic capabilities: creating Sybil identities, over-reporting budgets, and misreporting values. We further show the tightness of our impossibility result: removing any one from the strategy space restores feasibility under censorship resilience. Specifically, we construct a mechanism achieving IC and wLE for each restricted setting.
The adaptive clinching auction of Dobzinski, Lavi, and Nisan (FOCS'08) is a cornerstone of mechanism design with budget-constrained bidders, and it has since inspired a broad range of extensions to substantially richer market settings. Adaptive clinching enjoys compelling guarantees: it is Pareto optimal, and truthful bidding is a dominant strategy whenever budgets cannot be overreported. In many real-world environments, however, identities are cheap to create, making resilience to Sybil attacks an additional requirement. Whether the guarantees of adaptive clinching survive when a bidder can split its budget across fake identities has remained open. We introduce a generalized framework for adaptive clinching and prove that it satisfies Sybil-robust incentive compatibility: a bidder cannot benefit from splitting into any finite number of identities with arbitrary value reports, provided their aggregate reported budget does not exceed the bidder's enforceable true budget. Sybil resilience thus joins the list of ideal properties of adaptive clinching. Our work is motivated by practical problems in decentralized finance, where automated market makers (AMMs) are among the most widely used applications. We design a new AMM mechanism that extends classical adaptive clinching in two ways: it supports a two-sided market and protects against two-sided strategies, and it allows the supply function to depend on the auction state. Its one-sided special case is captured by our generalized adaptive clinching framework, and our proof techniques show that the full two-sided mechanism simultaneously satisfies Sybil-robust incentive compatibility and a relaxed notion of Pareto optimality. We argue that these guarantees are essentially the best possible via impossibility results. More broadly, our work reveals new connections between classical algorithmic game theory and modern decentralized blockchains.
Large language models (LLMs) can describe what a piece of code should compute, but they do so in free-form text. Such text cannot be checked against program executions, and it does not reveal whether a statement is correct. We present SemLoc, which turns LLM reasoning into executable, verifiable semantic constraints. With structured semantic grounding, the LLM writes each property as a constraint with a fixed category, attached to an anchor in the code graph, such as a variable definition, branch, loop, or return. SemLoc inserts each constraint as a runtime check and runs the tests, which yields a semantic violation spectrum over passing and failing tests. Counterfactual verification then patches the code at an anchor and reruns the tests; a patch that a constraint suggests counts as verified only if all tests then pass. We study SemLoc on fault localization, where the tests serve as the oracle for verification, with four LLMs on QuixBugs and on SemBugs, a new dataset of 809 real bugs from 750 GitHub projects. The constraints are executable: 99.4% become runtime checks, and on programs where coverage ties the faulty line with at least half of the covered lines, their violations alone rank it first in 40.9% of the cases, compared with 2.2% for spectrum-based fault localization and 7.0% for Daikon invariants. They make the LLM's answers verifiable: 68.4% of SemLoc's top-ranked answers carry a patch that the tests confirm, and 93.6% of these are correct, which the LLM's own confidence cannot match. SemLoc's first-rank accuracy does not differ significantly, for any of the four LLMs, from that of an LLM given the same inputs (81.4% vs. 82.0%), and it far exceeds that of an LLM that sees only the function (50.0%).
Chengqian Zhang, Yucheng Jin, Duo Zhang, Tiejun Li, Han Wang
Crystal generative models mainly learn what stable crystals look like, with little explicit supervision for what makes them stable. We reveal a substantial representation gap between state-of-the-art crystal generative models and pretrained universal machine learning interatomic potentials (MLIPs) via energy probing, and show this gap can be closed by a simple training-time alignment. We propose Crystal REPresentation Alignment (CrystalREPA), a plug-and-play framework that aligns the atom-wise hidden states of generative encoders with frozen MLIP representations through an element-aware contrastive objective, transferring stability-aware atomistic priors with marginal training overhead and no additional inference cost. Across three generative frameworks, ten MLIP teachers, and two benchmark datasets, CrystalREPA consistently improves the thermodynamic stability, structural validity, and structural fidelity of generated crystals. Equally important, we find that an MLIP's transfer effectiveness is poorly predicted by its accuracy on standard leaderboards (e.g., Matbench Discovery) but strongly predicted by the distinguishability of its atom-wise representation space, yielding a practical, accuracy-independent criterion for selecting MLIP teachers for generative transfer. Code is available at https://github.com/Chengqian-Zhang/CrystalREPA.
Junxing Hu, Zicheng Zhang, Zhiqian Zhang, Ai Han, Yifeng Zhang, Pengzhang Liu, Ling Li
Reasoning models trained with Group Relative Policy Optimization (GRPO) now produce chains of thought thousands of tokens long, and recent studies reveal that such training is prone to policy collapse. Entropy regularization is the standard remedy, but it works by pushing probability mass away from the few tokens that carry it and toward the vast low-probability tail of the vocabulary, and it applies this force once for every token of the reasoning trace. This puts the bonus in direct conflict with the entropy decay that convergence requires, and makes its coefficient difficult to set in the long-trace regime. To resolve this tension, we propose Info-GRPO, an information-theoretic framework that cultivates correlation between the policy and a latent prior. Rather than raising entropy everywhere, Info-GRPO builds on the mutual information between a latent variable and the trajectory it induces: augmenting prompts with latent seeds during training lets the model explore diverse policies correlated with the prior, while the seed-conditioned entropy is guided toward convergence. Diversity is therefore carried across whole reasoning traces rather than by per-token randomness, and exploration no longer fights entropy reduction. Experiments show that Info-GRPO outperforms vanilla GRPO and entropy-regularized GRPO across diverse reasoning benchmarks. On AIME24, it improves $Avg@8$ over GRPO by 3.75%, 1.66%, 4.16%, and 4.17% on four backbones, and it reaches the same majority-vote accuracy with $8\times$ fewer samples. Code and models will be publicly released.
Miner extractable value (MEV) in automated market makers allows block builders to profit from transaction ordering and injected trades, imposing costs on users and contributing to builder centralization. We introduce Otter (Optimal Truthful Trading with Excess Redistribution), a two-asset batch AMM that achieves provable MEV resilience when the consensus layer provides censorship resilience and block space is uncongested. Otter makes truthful behavior a dominant strategy for both users and builders. Consequently, a builder cannot profit from strategic deviations, including reordering bids or injecting sybil bids, even when the builder is itself a trader with intrinsic value. Moreover, we show that Otter maximizes social welfare, in a strong sense, within a natural class of mechanisms satisfying the desired game-theoretic properties. To achieve these guarantees, we introduce a new paradigm called surplus redistribution, which provably prevents residual surplus from being captured as MEV by redirecting it to the broader community. Specifically, the pool's output tokens need not be distributed entirely among users in the current batch. Instead, residual surplus may be transferred to a smart contract governed by the decentralized community. The accumulated surplus can be used to benefit community members --- for example, by subsidizing transaction fees, rewarding liquidity providers, or returning assets to the pool. Our approach relies on the underlying consensus layer to provide censorship resilience. We motivate its necessity through an impossibility result showing that the desired game-theoretic guarantees become unattainable when the builder is additionally allowed to censor transactions. Thus, our results also provide a mathematically formal demonstration of how consensus-level security guarantees can fundamentally expand what is achievable at the application layer.
Haibo Tong, Zeyang Yue, Feifei Zhao Institute of automation, Chinese academy of science, Chinese Academy of Sciences, Erliang Lin, Lu Jia, Ruolin Chen Institute of automation, Chinese academy of science, Chinese Academy of Sciences, Yinqian Sun Institute of automation, Chinese academy of science, Chinese Academy of Sciences, Qian Zhang Institute of automation, Chinese academy of science, Chinese Academy of Sciences 等 9 位作者
Whether Large Language Models (LLMs) truly possess human-like Theory of Mind (ToM) capabilities has garnered increasing attention. However, existing benchmarks remain largely restricted to narrow paradigms like false belief tasks, failing to capture the full spectrum of human cognitive mechanisms. We introduce CogToM, a comprehensive, theoretically grounded benchmark comprising over 8000 bilingual instances across 46 paradigms, validated by 49 human annotators. A systematic evaluation of 22 representative models, including frontier models like GPT-5.1 and Qwen3-Max, reveals significant performance heterogeneities and highlights persistent bottlenecks in specific dimensions. Further analysis based on human cognitive patterns suggests potential divergences between LLM and human cognitive structures. CogToM offers a robust instrument and perspective for investigating the evolving cognitive boundaries of LLMs. We release our code and data at https://github.com/Beijing-AISI/CogToM.
尚无 AI 解读，展示原文摘要
cs.CL
Proceedings of the 64th Annual Meeting of the Association for Computational Linguistics (Volume 1: Long Papers) · 2026 · 正式发表
Xin Chen Nanjing University, Feng Jiang Shenzhen University of Advanced Technology, Yiqian Zhang, Hardy Chen, Shuo Yan University of Texas at Dallas, Wenya Xie University of Minnesota - Twin Cities, Min Yang, Shujian Huang Nanjing University
Reasoning-oriented Large Language Models (LLMs) have achieved remarkable progress with Chain-of-Thought (CoT) prompting, yet they remain fundamentally limited by a blind self-thinking paradigm: performing extensive internal reasoning even when critical information is missing or ambiguous. We propose Proactive Interactive Reasoning (PIR), a new reasoning paradigm that transforms LLMs from passive solvers into proactive inquirers that interleave reasoning with clarification. Unlike existing search- or tool-based frameworks that primarily address knowledge uncertainty by querying external environments, PIR targets premise- and intent-level uncertainty through direct interaction with the user. PIR is implemented via two core components: (1) an uncertainty-aware supervised fine-tuning procedure that equips models with interactive reasoning capability, and (2) a user-simulator-based policy optimization framework driven by a composite reward that aligns model behavior with user intent. Extensive experiments on mathematical reasoning, code generation, and document editing demonstrate that PIR consistently outperforms strong baselines, achieving up to 32.70% higher accuracy, 22.90% higher pass rate, and 41.36 BLEU improvement, while reducing nearly half of the reasoning computation and unnecessary interaction turns. Further reliability evaluations on factual knowledge, question answering, and missing-premise scenarios confirm the strong generalization and robustness of PIR.
尚无 AI 解读，展示原文摘要
cs.CL
Proceedings of the 64th Annual Meeting of the Association for Computational Linguistics (Volume 1: Long Papers) · 2026 · 正式发表
Jianing Zhang, Runan Li, Honglin Pang, Ding Xia The University of Tokyo, Zhou Zhu, Qian Zhang, Chuntao Li Jilin University, Xi Yang Jilin University
Deciphering ancient Chinese Oracle Bone Script (OBS) is a challenging task that offers insights into the beliefs, systems, and culture of the ancient era. Existing approaches treat decipherment as a closed-set image recognition problem, which fails to bridge the “interpretation gap”: while individual characters are often unique and rare, they are composed of a limited set of recurring, pictographic components that carry transferable semantic meanings. To leverage this structural logic, we propose an agent-driven Vision-Language Model (VLM) framework that integrates a VLM for precise visual grounding with an LLM-based agent to automate a reasoning chain of component identification, graph-based knowledge retrieval, and relationship inference for linguistically accurate interpretation. To support this, we also introduce OB-Radix, an expert-annotated dataset providing structural and semantic data absent from prior corpora, comprising 1,022 character images (934 unique characters) and 1,853 fine-grained component images across 478 distinct components with verified explanations. By evaluating our system across three benchmarks of different tasks, we demonstrate that our framework yields more detailed and precise decipherments compared to baseline methods.
尚无 AI 解读，展示原文摘要
cs.CL · cs.CV
Proceedings of the 64th Annual Meeting of the Association for Computational Linguistics (Volume 1: Long Papers) · 2026 · 正式发表
Zining Liu, Yunhai Hu, Tianhua Xia, BO Bao, Eric Sather Cerebras Systems, Inc, Vithursan Thangarasa Cerebras Systems, Inc, Sai Qian Zhang New York University
Speculative decoding (SD) has proven to be an effective technique for accelerating autoregressive generation in large language models (LLMs), however its application to vision-language models (VLMs) remains relatively unexplored. We propose DREAM-S, a novel SD framework designed specifically for fast and efficient decoding in VLMs. DREAM-S leverages a neural architecture search (NAS) framework with target-aware supernet training to automatically identify both the optimal interaction strategy between the draft and target models, and the most suitable draft model architecture for the underlying hardware implementation platform. DREAM-S additionally incorporates adaptive intermediate feature distillation, guided by attention entropy, to enable efficient draft training. Experiments on a range of well-established VLMs show that DREAM-S achieves up to a 3.85\times speedup compared to standard decoding approaches and significantly outperforms existing SD baselines.
尚无 AI 解读，展示原文摘要
cs.CL · cs.CV
Proceedings of the 64th Annual Meeting of the Association for Computational Linguistics (Volume 1: Long Papers) · 2026 · 正式发表
Qian Zhang, Xinye Li, Xiaokai Wu, Junhao Xu, Zhanyue Qin Harbin Institute of Technology (Shenzhen) and Tencent, Qingbin Liu, Junxian Cai Tencent, Xi Chen 等 13 位作者
Model Editing, also known as knowledge editing, is receiving increasing attention in the field of Large Language Models (LLMs). However, existing model editing approaches predominantly focus on knowledge-level or static visual domains, overlooking dynamic semantics. This paper exploratively applies six representative model editing methods (FT, IKE, MEND, SERAC, MEMIT and AlphaEdit) to Video Large Language Models (Vid-LLMs) and introduces the first benchmark specifically designed for Vid-LLMs editing—VMEB (Vid-LLMs Model Editing Benchmark)—systematically extending model editing research from static modalities to dynamic video scenarios. We position this work as a forward-looking benchmark and a foundational diagnostic study: in the video paradigm, our evaluation dimensions encompass traditional metrics including Reliability, Locality, and Generality, while also introducing a video-specific metric: Robustness. Based on experimental results, we analyze the strengths and limitations of existing model editing approaches, and identify new challenges and research directions for the future development of the model editing field within the context of multimodal and video paradigms. Our benchmark is available at https://github.com/Sakabamrisa/VMEB.
尚无 AI 解读，展示原文摘要
cs.CL · cs.CV
Proceedings of the 64th Annual Meeting of the Association for Computational Linguistics (Volume 1: Long Papers) · 2026 · 正式发表
Yifan Ji, Zhipeng Xu, Zhenghao Liu Northeastern University, Zulong Chen, Qian Zhang Alibaba Group, Zhibo Yang, Junyang Lin Alibaba Group, Yu Gu 等 10 位作者
Key Information Extraction (KIE) from real-world documents remains challenging due to substantial variations in layout structures, visual quality, and task-specific information requirements. Recent Large Multimodal Models (LMMs) have shown promising potential for performing end-to-end KIE directly from document images. To enable a comprehensive and systematic evaluation across realistic and diverse application scenarios, we introduce UNIKIE-BENCH, a unified benchmark designed to rigorously evaluate the KIE capabilities of LMMs. UNIKIE-BENCH consists of two complementary tracks: a constrained-category KIE track with scenario-predefined schemas that reflect practical application needs, and an open-category KIE track that extracts any key information that is explicitly present in the document. Experiments on 15 state-of-the-art LMMs reveal substantial performance degradation under diverse schema definitions, long-tail key fields, and complex layouts, along with pronounced performance disparities across different document types and scenarios. These findings underscore persistent challenges in grounding accuracy and layout-aware reasoning for LMM-based KIE. All codes and datasets are available at https://github.com/NEUIR/UNIKIE-BENCH.
尚无 AI 解读，展示原文摘要
cs.CL · cs.CV
Proceedings of the 64th Annual Meeting of the Association for Computational Linguistics (Volume 1: Long Papers) · 2026 · 正式发表