论文检索

Applications · Robotics

Kewei Chen, Yayu Long, mingsheng shang

Despite rapid progress in Vision-Language-Action (VLA) models for robotic control, instruction drift remains a persistent failure mode in long-horizon tasks. This paper reconceptualizes this phenomenon, positing that instruction drift is fundamentally a systematic sampling error: local greedy sampling is prone to collapsing into ''Negative Pivotal Windows''—irreversible local optima with high local probability that sever global success pathways. To address this, we propose \textbf{Context-Aware Power Sampling (CAPS)}, a training-free inference-time computation framework. CAPS leverages power distributions to sharpen global trajectory probabilities, effectively activating the model's implicit world model for lookahead planning. Furthermore, we introduce a metacognitive control mechanism based on Signal-to-Noise Ratio (SNR). This mechanism triggers adaptive MCMC search solely when drift risk is detected, enabling a dynamic transition from ''intuitive fast thinking'' to ''rational slow search.'' Experiments on RoboTwin, Simpler-WindowX, and Libero-long benchmarks demonstrate that CAPS significantly outperforms SOTA baselines, such as OpenVLA and TACO, without parameter updates. These results confirm that adaptive inference-time computation is a potent pathway to enhancing embodied long-horizon robustness.

Applications · Robotics

Keming Zhang, Sixian Zhang, Xinhang Song, Hongyu Wang, Yiyao Wang, Yingjie Wang, Shuqiang Jiang

Open-vocabulary mobile manipulation (OVMM) requires long-horizon navigation in unseen environments and object-centric manipulation. Most existing methods treat navigation and manipulation as separate stages, which can yield navigation endpoints that are poor for manipulation or manipulation-friendly poses that are globally inefficient. We address this mismatch with 3D Interaction Chains (3D-IC), a unified framework that couples multi-stage navigation and manipulation planning. 3D-IC maintains a shared 3D feature map for both skills, generates stage-aligned interaction waypoints, and links them into candidate multi-stage chains. A hierarchical policy then scores these chains by jointly considering feasibility (via VLM reasoning over waypoint-centric 3D features) and transition cost, selecting the best trade-off between success and path efficiency. The robot executes the next waypoint and replans as new observations arrive. Experiments in simulation and on a real Stretch 3 robot demonstrate consistent gains in both task success and trajectory efficiency.

Probabilistic Methods · Variational Inference

Ananyapam De, Linus Bleistein, Anton Thielmann, Benjamin Säfken

Optimal Transport (OT) traditionally relies on a fixed ground cost to produce a single deterministic transport plan—a practice that overlooks the inherent variability and noise in real-world data. While recent sampling based approaches of OT offer a principled way to quantify this uncertainty, these are computationally prohibitive and struggle to scale. In this paper, we introduce Sinkhorn-parameterized Variational Inference, a first scalable variational framework for performing posterior inference over transport plans. Our key insight is that the Sinkhorn map can be treated as a differentiable reparameterization of the set of entropic plans. This enables the use of flexible generative models like normalizing flows to approximate distributions over transport plans while enforcing marginal constraints. We experimentally demonstrate that our method matches the quality of intensive sampling techniques at a fraction of the computational cost, scaling effectively to large-scale problems.

Deep Learning · Large Language Models

Beibei Xiong, Hangyu Lv, Junqi Liu, Yisen Wang, Shaoshi Chen, Jianlin Wang, Zhengfeng Yang, Lihong Zhi

Automating formal proofs of combinatorial identities is challenging for LLM-based provers, as long-horizon proof planning is required and unconstrained search quickly explodes. Symbolic methods such as the Wilf--Zeilberger (WZ) method can achieve a mechanized proof of combinatorial identities by constructing special auxiliary functions and demonstrating that they satisfy specific recurrence relations. We propose WZ-LLM, a neuro-symbolic framework that turns WZ proof plans into executable proof sketches in Lean~4 and uses an LLM-based prover to discharge the resulting machine-checkable subgoals. We also train a dedicated WZ-Prover via a Lean-kernel-verified bootstrapping loop with expert-verified iteration, followed by DAPO-based refinement. Experiments show that WZ-LLM achieves a 34\% proof success rate on LCI-Test (100 classical combinatorial identities), outperforming strong baselines such as DeepSeek-V3 and Goedel-Prover-V2; moreover, on LCI-Test it proves 5 identities on which the symbolic-only baseline fails. WZ-LLM also improves performance on CombiBench and PutnamBench-Comb, suggesting the effectiveness of coupling symbolic proof sketches with learned formal reasoning. Experiments show that WZ-LLM achieves a 34\% proof success rate on LCI-Test (100 classic combinatorial identities), outperforming strong baselines such as DeepSeek-V3 and Goedel-Prover-V2, and delivering consistent gains on CombiBench and PutnamBench-Comb. These results indicate that our framework provides two complementary strengths: improved direct proving for identities beyond the scope of WZ, and substantially higher end-to-end success when WZ sketches guide a specialized prover.

Social Aspects · Safety

Mansur Ali Khan, Mehmet Efe Akengin, Osman Salahuddin, Ahmad A. Rushdi

While AI models advance at unprecedented rates, AI safety legislation in the United States remains largely stalled or unrealized. We observe that AI policy activity is increasing globally, yet binding enactments remain limited relative to the pace of technical capability releases. We argue for the need to bridge this gap between AI development and its regulation. Specifically, we support our position through a technical analysis of all U.S. AI-related bills introduced from 2017 to 2025, showing that only 4.23% of U.S. AI bills reach any terminal outcome. We identify that procedural bottlenecks, including committee pigeonholing, multi-sponsor coordination challenges, and expertise asymmetries, are primary correlates of legislative stalling. Our comprehensive analysis of institutional, economic, political, and informational constraints shows factors exacerbating these regulatory delays. To address this multi-faceted gap, we propose policy recommendations grounded in planned adaptation, preemptive enactment, and independent AI oversight. Finally, we highlight the need for coordinated action across policymakers, developers, and industry stakeholders so that AI safety governance keeps pace with technological innovation.

Jiajun Zhang, Yuheng Jing, Zeyu Cui, Hao Zheng, Wentao Chen, Kaixin LI, Jiaxi Yang, Tianbao Xie, Zeyao Ma, Tianyi Bai 等

The rapid evolution of Large Language Models (LLMs) has empowered even non-programmers to create visually appealing frontend mini-games with a single instruction. However, open-source models significantly lag behind proprietary counterparts in this domain. The core bottleneck is the lack of an evaluation mechanism that balances reliability with scalability, as existing methods either fail to verify dynamic interactivity or incur prohibitive computational costs. To bridge this gap, we introduce ALIVE (Aligning LLMs via Interactive Visual Execution), a high-throughput framework that leverages one-shot planning and DOM-based analysis to automatically evaluate generated games at scale. Extensive experiments demonstrate that ALIVE significantly outperforms static judge baselines in identifying functional flaws while remaining orders of magnitude more efficient than GUI agents. Functioning as a scalable `pre-flight' evaluation layer, it curates high-quality data for Supervised Fine-Tuning (SFT) and provides a consistent reward signal for Reinforcement Learning (RL). We leverage this pipeline to train ALIVE-Coder, a model achieving superior performance in interactive frontend generation. To the best of our knowledge, our work offers the first scalable path to evaluate and optimize interactive code, substantially advancing open-source capabilities.

Theory · Reinforcement Learning and Planning

Harin Lee, Kevin Jamieson

We study reinforcement learning with delayed state observation, where the agent observes the current state after some random number of time steps. We propose an algorithm that combines the augmentation method and the upper confidence bound approach. For tabular Markov decision processes (MDPs), we derive a regret bound of $\tilde{\mathcal{O}}(H \sqrt{D_{\max} SAK})$, where $S$ and $A$ are the cardinalities of the state and action spaces, $H$ is the time horizon, $K$ is the number of episodes, and $D_{\max}$ is the maximum length of the delay. We also provide a matching lower bound up to logarithmic factors, showing the optimality of our approach. Our analytical framework formulates this problem as a special case of a broader class of MDPs, where their transition dynamics decompose into a known component and an unknown but structured component. We establish general results for this abstract setting, which may be of independent interest.

Theory · Reinforcement Learning and Planning

Orin Levy, Aviv Rosenberg, Alon Peled-Cohen, Yishay Mansour

We introduce OPO-CMDP, the first policy optimization algorithm for stochastic Contextual Markov Decision Process (CMDPs) under general offline function approximation. Our approach achieves a high probability regret bound of $\widetilde{O}(H^4\sqrt{T|S||A|\log(|\mathcal{F}||\mathcal{P}|)}),$ where $S$ and $A$ denote the state and action spaces, $H$ the horizon length, $T$ the number of episodes, and $\mathcal{F}, \mathcal{P}$ the finite function classes used to approximate the losses and dynamics, respectively. This is the first regret bound with optimal dependence on $|S|$ and $|A|$, directly improving the current state-of-the-art (Qian, Hu, and Simchi-Levi, 2024). These results demonstrate that optimistic policy optimization provides a natural, computationally superior and theoretically near-optimal path for solving CMDPs.

Deep Learning · Large Language Models

Guangyue Peng, Zongchao Chen, Wen Luo, Yuntao Wen, Wei Li, Ruixiang Feng, Ran Le, Chen Yang, Zhenwei An, Yang Song 等

Reverse Chain-of-Thought Generation (RCG) synthesizes reasoning traces from query-answer pairs, but runs the risk of producing post-hoc rationalizations: when models can see the answer during generation, the answer serves as a cognitive anchor that shapes the entire explanation. We formalize this phenomenon through a three-level measurement hierarchy: lexical, entropic, and probabilistic anchoring, each captures surface artifacts, entropy dynamics, and latent answer dependence, respectively. We analyze semantic suppression, the intuitive mitigation strategy that instructs models to ignore the answer, to find out its counterproduction: while it reduces lexical overlap, it paradoxically increases entropic and probabilistic anchoring. Drawing on Ironic Process Theory from cognitive psychology, we attribute this failure to active monitoring of the forbidden answer, which inadvertently deepens dependence on it. To break this cycle, we propose Structural Skeleton-guided Reasoning (SSR), a two-phase approach that first generates an answer-invariant functional skeleton structure, then uses this skeleton to guide full trace generation. By redirecting the information flow to structural planning rather than answer monitoring, SSR consistently reduces anchoring across all three levels. We further introduce Distilled SSR (SSR-D), which fine-tunes models on teacher-generated SSR traces to ensure reliable structural adherence. Experiments across open-ended reasoning benchmarks demonstrate that SSR-D achieves up to 10\% improvement over suppression baselines while preserving out-of-distribution (OOD) generalization. Code and data will be open-sourced upon acceptance.

Theory · Game Theory

Xiaotie Deng, Yanru Guan, Ningyuan Li, Zihe Wang, Jie Zhang

We study mechanism design for selling data as a digital good when the value derived from training AI models follows a scaling law. The seller faces a linear cost when producing data, while the buyers benefit from additional data with diminishing returns as data volume increases. This departs from classical auction models by allowing allocations to be continuous quantities of data rather than binary outcomes. We first analyze an offline setting in which all buyer types are realized simultaneously, characterizing profit-optimal mechanisms and showing how virtual-value methods extend to continuous data allocations. We then consider an online setting with sequential arrivals, where production decisions must be made under demand uncertainty. We show that myopic allocation and fixed production plans can be arbitrarily suboptimal, whereas a simple two-stage algorithm that combines upfront production with adaptive expansion achieves a constant-factor approximation to the offline optimum. Finally, we study bilateral data trading under asymmetric information, where both the buyer’s value and the seller’s cost are private. Although the optimal truthful mechanism has a complex structure, we show that simple and implementable mechanisms recover a constant fraction of the first-best gain-from-trade. Overall, our results highlight how scaling laws introduce new algorithmic trade-offs in market design and provide performance guarantees for data markets under uncertainty.

Yuqi Jiang, Yumeng Liu, Zimu Li, Jinyuan Deng, Qian Jin, Yucheng Cui, YU LI, Xunzhao Yin, Qi Sun, Cheng Zhuo

As semiconductor technology nodes continue to shrink, computational lithography has become critical to yield and performance. However, real-world lithography is a continuous, multi-stage physical process driven by implicit interventions, which cannot be captured by the existing static or stage-wise models. To address these issues, we present \textbf{LithoDreamer}, the first physics-informed World Model (WM) framework for computational lithography, designed to represent the ``Layout-Mask-Resist Image-After Development Image (ADI)'' pipeline as a decision-driven multi-stage physical evolution system, enabling multi-step latent state rollouts within stages and intervention-aware decision-making across stages. First, we learn the feature variations between adjacent states in latent spaces to capture the physical dynamics of each stage. Second, the model plans continuous process interventions through physical mappings in the spaces, which in turn drive subsequent state transitions. Furthermore, we propose a contrastive variational optimization paradigm that jointly explores the evolutions of the interventions and states without discrete action supervision, enabling stable and continuous process rollouts in the WM. Extensive experiments show that LithoDreamer achieves state-of-the-art accuracy and generalization performance.

Qinsi Wang, Jing Shi, Kun Wan, Handong Zhao, Hancheng Ye, Zishan Shao, Jinghan Ke, Yudong Liu, Daniel Miranda, Purvak Lapsiya 等

Although Reinforcement Learning Fine-Tuning (RLFT) applied to Vision-Language Models (VLMs) substantially enhances multimodal reasoning capabilities, their prohibitive training cost limits broad adoption. Surprisingly, most existing methods simply port Large Language Model (LLM) RLFT techniques to VLMs, while ignoring a intrinsic property of multimodal models: their dynamic text–vision alignment. We ask a new question: Can this intrinsic alignment be turned into a training signal that makes VLM RLFT more efficient? We analyze how a VLM plans to attend, actually attends, and ideally should attend during reasoning, and derive two lightweight metrics from these patterns. Predictive View Accuracy (PVA) estimates sample difficulty, and Reasoning View Accuracy (RVA) reflects the quality of chain-of-thought (CoT) reasoning. These alignment signals enable automated data curriculum and dense reasoning supervision. We introduce FOCUS-RL, a plug-and-play framework that can be seamlessly integrated into any VLM and dramatically boosts RLFT training efficiency. FOCUS-RL achieves 2.5 x – 4 x faster convergence over vanilla GRPO and consistent accuracy gains (+4.4 on average) across six different benchmarks and multiple VLM families.

Reinforcement Learning · Batch/Offline

Lu Guo, Yixiang Shan, Zhengbang Zhu, Qifan Liang, Lichang Song, Ting Long, Weinan Zhang, Yi Chang

Offline reinforcement learning (RL) learns policies from fixed datasets, thereby avoiding costly or unsafe environment interactions. However, its reliance on finite static datasets inherently restricts the ability to generalize beyond the training distribution. Prior solutions based on synthetic data augmentation often fail to generalize to unseen scenarios in the (augmented) dataset. To address these challenges, we propose Retrieval High-quAlity Demonstrations (RAD) for decision-making, which innovatively introduces a retrieval mechanism into offline RL. Specifically, RAD retrieves high-return and reachable states from the offline dataset as target states, and leverages a generative model to generate sub-trajectories conditioned on these targets for planning. Since the targets are high-return states, once the agent reaches such a target, it can continue to obtain high returns by following the associated high-return actions, thereby improving policy generalization. Extensive experiments confirm that RAD achieves competitive or superior performance compared to baselines across diverse benchmarks, validating its effectiveness. Our code is available at https://anonymous.4open.science/r/RAD_ICML-6CC9.

Deep Learning · Large Language Models

Kaixiang Wang, Yidan Lin, Jiong Lou, Zihan Wang, Bunyod Suvonov, Zhaojiacheng Zhou, Yuxiang Zheng, Jiaxi Cao, Zhiheng Dong, Chentao Wu 等

The evolution of Large Language Model (LLM) agents towards System~2 reasoning, characterized by deliberative, high-precision problem-solving, necessitates maintaining rigorous logical integrity over extended horizons. However, prevalent memory preprocessing paradigms incur destructive de-contextualization. By compressing fluid sequential dependencies into pre-defined structures (e.g., embeddings or graphs), these methods sever the narrative integrity essential for deep reasoning. To address this, we propose E-mem, a framework shifting from Memory Preprocessing to Episodic Context Reconstruction inspired by biological engrams. E-mem employs a heterogeneous hierarchical architecture where multiple assistant agents maintain uncompressed memory contexts, while a central master agent orchestrates global planning. Unlike passive retrieval, our mechanism empowers assistants to locally reason within activated segments, extracting context-aware evidence before aggregation. Evaluations on the LoCoMo benchmark demonstrate that E-mem achieves over 54\% F1—surpassing the state-of-the-art GAM by 7.75\%—while reducing token cost by over 70\%. Our work is available on \url{https://anonymous.4open.science/r/E-mem-F6C3/}.

Social Aspects · Accountability, Transparency, and Interpretability

Emmy Liu, Varun Prashant Gangal, Chelsea Zou, Michael Yu, Xiaoqi Huang, Alex Chang, Zhuofu Tao, Karanpartap Singh, Sachin Kumar, Steven Feng

Despite numerous attempts at mitigation since the inception of language models, hallucinations remain a persistent problem even in today's frontier LLMs. Why is this? We review existing definitions of hallucination and fold them into a single, unified definition wherein prior definitions are subsumed. This position paper argues that hallucination can be unified by defining it as simply inaccurate (internal) world modeling, in a form where it is observable to the user. For example, stating a fact which contradicts a knowledge base OR producing a summary which contradicts the source. By varying the reference world model and conflict policy, our framework unifies prior definitions. We argue that this unified view is useful because it forces evaluations to clarify their assumed reference "world", distinguishes true hallucinations from planning or reward errors, and provides a common language for comparison across benchmarks and discussion of mitigation strategies. Building on this definition, we outline plans for a family of benchmarks using synthetic, fully specified reference world models to stress-test and improve world modeling components.

Reinforcement Learning · Multi-agent

Haibo Jin, Peng Kuang, Ye Yu, Xiaopeng Yuan, Haohan Wang

While existing multi-agent systems (MAS) can handle complex problems by enabling collaboration among multiple agents, they are often highly task-specific, relying on manually crafted agent roles and interaction prompts, which leads to increased architectural complexity and limited reusability across tasks. Moreover, most MAS communicate primarily through natural language, making them vulnerable to error accumulation and instability in long-context, multi-stage interactions within internal agent histories. In this work, we propose \textbf{Agent Primitives}, a set of reusable latent building blocks for LLM-based MAS. Inspired by neural network design, where complex models are built from reusable components, we observe that many existing MAS architectures can be decomposed into a small number of recurring internal computation patterns. Based on this observation, we instantiate three primitives: Review, Voting and Selection, and Planning and Execution. All primitives communicate internally via key–value (KV) cache, which improves both robustness and efficiency by mitigating information degradation across multi-stage interactions. To enable automatic system construction, an Organizer agent selects and composes primitives for each query, guided by a lightweight knowledge pool of previously successful configurations, forming a primitive-based MAS. Experiments show that primitives-based MAS improve average accuracy by 12.0–16.5\% over single-agent baselines, reduce token usage and inference latency by approximately 3$\times$–4$\times$ compared to text-based MAS, while incurring only 1.3$\times$–1.6$\times$ overhead relative to single-agent inference and providing more stable performance across model backbones.

Applications · Robotics

Feiyang Jia, Lin Liu, Ziying Song, Caiyan Jia, Hangjun Ye, Xiaoshuai Hao, Long Chen

End-to-end (E2E) autonomous driving has recently attracted increasing interest in unifying Vision–Language–Action (VLA) with World Models to enhance decision-making and forward-looking imagination. However, existing methods fail to effectively unify future scene evolution and action planning within a single architecture due to inadequate sharing of latent states, limiting the impact of visual imagination on action decisions. To address this limitation, we propose DriveWorld-VLA, a novel framework that unifies world modeling and planning within a latent space by tightly integrating VLA and world models at the representation level, which enables the VLA planner to benefit directly from holistic scene-evolution modeling and reducing reliance on dense annotated supervision. Additionally, DriveWorld-VLA incorporates the latent states of the world model as core decision-making states for the VLA planner, facilitating the planner to assess how candidate actions impact future scene evolution. By conducting world modeling entirely in the latent space, DriveWorld-VLA supports controllable, action-conditioned imagination at the feature level, avoiding expensive pixel-level rollouts. Extensive open-loop and closed-loop evaluations demonstrate the effectiveness of DriveWorld-VLA, which achieves state-of-the-art performance with 91.3 PDMS on NAVSIMv1, 86.8 EPDMS on NAVSIMv2, and 0.16 3-second average collision rate on nuScenes. Code and models will be publicly released.

Deep Learning · Large Language Models

Anthony Liang, Jonathan Berant, Adam Fisch, Abhimanyu Goyal, Kalpesh Krishna, Jacob Eisenstein

Reasoning models often spend a lot of time thinking before they generate a visible response. This creates a frustrating, but unfortunately common, experience: the user's time is wasted while the model reasons from a false premise that could have easily been corrected. In contrast, human speakers perform lightweight, incremental check-ins to ensure that conversational participants stay on common ground. With this motivation, we propose \textit{interleaved reasoning} (IR), in which the model alternates between thinking and surfacing intermediate responses, as an alternative to the standard ``think-then-answer'' approach. By providing useful information to the user earlier, IR reduces perceived latency, the time a user waits for an initial output, without compromising the quality of the final response. We focus on a specialization of interleaved reasoning, \method (\textul{Plan}-\textul{T}hought-\textul{A}nswer \textul{In}terleaving), where the first intermediate response is an explicit, step-by-step \textit{plan} for executing the task. This plan-first strategy allows for user intervention and early feedback for subsequent reasoning steps. \method\ yields an $\sim$6\% improvement in pass@1 across several challenging math reasoning and coding benchmarks, while reducing time-to-first-response by over 60\% relative to think-then-answer baselines.

Deep Learning · Algorithms

Kaiqi Lin, Jianping Luo

While LLM-driven Neural Architecture Search (NAS) leverages exceptional code generation and reasoning, it suffers from a critical "Semantic-Physical Misalignment": LLMs prioritize high-level semantic plausibility but are agnostic to intrinsic physical dynamics such as gradient flow, whereas Zero-Cost Proxies (ZCPs) capture these local sensitivities but lack global semantic planning. To bridge this gap, we propose SAGE-NAS, a closed-loop evolutionary framework that synergizes an LLM-Based Semantic Agent with a Graph-Based Evaluator. Specifically, SAGE-NAS coordinates an LLM-driven Semantic Agent to construct candidate architectures by dynamically scheduling complementary sub-policies that balance exploitation with exploration. Furthermore, the framework integrates a Dual-Modality Graph Evaluator that serves as a rapid performance predictor by fusing ZCP statistics with topological features, and a State-Aware Behavioral Atlas that guides sparsity-driven exploration to escape local optima. Experiments demonstrate that SAGE-NAS achieves state-of-the-art performance across multiple mainstream search spaces and downstream tasks, exhibiting a superior balance between search efficiency, model accuracy, and cross-task generalization capability.

Deep Learning · Generative Models and Autoencoders

Yangtian Zhang, Zhe Wang, Arthur Gretton, ZHITAO YING, David van Dijk, Michalis Titsias, Jiaxin Shi

Non-monotonic sequence generation methods, such as masked diffusion models, provide a flexible alternative to left-to-right autoregressive modeling by allowing tokens to be generated in non-fixed and prescribed orders. Despite their practical advantages, most existing non-monotonic models are order-agnostic and rely on a fixed-length masked token grid, limiting their ability to support variable-length generation and adaptive insertion order. In this work, we introduce a probabilistic framework for learning insertion order in variable-length insertion models. We formalize a bijective correspondence between insertion trajectories and permutations, which enables an exact reparameterization of the data likelihood as a sum over permutations. Building on this result, we propose the **Insertion Process (IP)**, a stochastic generative model that jointly learns *where* to insert, *what* to insert, and *when* to terminate, trained via permutation-based variational inference. Unlike prior masked or fixed-canvas approaches, IP natively supports variable-length generation and learns data-driven preferences over insertion orders. Experiments on planning benchmarks and molecular SMILES generation demonstrate that learning insertion order improves both modeling quality and generalization in domains without a canonical left-to-right structure.