arXiv subjects

Meng Fang

Publications and source records attributed to Meng Fang.

At least 19 recordsLinked to original sources

arXiv2026

Continual Learning Exam: Benchmarking Long-Horizon RSI and Experiential Learning in LLM Agents

Continual learning is the ability of LLM agents to improve over time by acquiring, refining, and applying knowledge from experience while retaining useful prior knowledge. It is essential for agents learning over long time horizons, yet current evaluations largely assess individual task performance rather than improvement through accumulated experience. We introduce Continual Learning Exam (CLE), a large-scale benchmark developed by 50+ domain experts to measure experience-driven improvement in realistic professional workflows. CLE comprises 120 human-authored tasks, each organised as a five-level ladder, yielding 600 levels spanning 89 professional roles across 12 occupational domains. Agents must pass each level to unlock the next, with later levels building on experience accumulated during earlier assignments. An evaluation agent assesses each submission against human-authored rubrics. We compare a single-submission setting with one that provides diagnostic feedback after failure and allows one revision. Across 17 harness-model configurations using frontier LLMs, feedback and revision yield a median completion gain of 5.3 percentage points. Yet the best-performing system clears only 47.7% of levels and completes all five levels in just 23.3% of tasks, highlighting the difficulty of sustained learning from experience for existing frontier LLMs. As such, the dataset can also be used to post-train LLMs to strengthen their continual learning capabilities.

cs.LG↗

arXiv2026

Automotive-ENV: Benchmarking Multimodal Models in Automotive Cockpit Environments

Multimodal models have demonstrated strong capabilities in web, desktop, and mobile GUI environments, but their performance in automotive cockpit systems remains largely unexplored. In-vehicle GUIs introduce unique challenges beyond traditional GUI interaction, including vehicle-state reasoning, implicit intent understanding, and safety-critical decision making. To address this gap, we introduce Automotive-ENV, an interactive benchmark for evaluating multimodal models in automotive cockpit environments. Built on Android Automotive OS, Automotive-ENV combines cockpit-screen interaction with vehicle-state signals and deterministic state-based evaluation. The benchmark contains 420 carefully constructed tasks covering routine cockpit operation and safety-critical scenarios. Unlike prior GUI benchmarks that rely on action matching or LLM-based judging, Automotive-ENV evaluates model behavior through reproducible validators over UI states, vehicle signals, and safety-related conditions. We benchmark representative general-purpose multimodal models and GUI-specialized vision-language models on Automotive-ENV. Experimental results show that current models remain far from reliable in automotive environments, with substantial performance degradation in implicit and safety-critical tasks. Further analysis reveals consistent failure modes in safety-aware reasoning, vehicle-control understanding, and implicit hazard recognition. We release Automotive-ENV, together with tasks, validators, and evaluation tools, to support future research on automotive multimodal systems.

cs.RO↗

arXiv2026

Cracking Mobile Manipulation: From Failure Diagnosis to Agentic Robot System

Recent vision-language-action (VLA) models have shown strong performance on local robot manipulation, yet their success does not readily translate to long-horizon mobile manipulation. In our experiments, a pretrained subtask VLA achieves 80% success rate on atomic manipulation tasks but only 19% when these skills are composed into mobile manipulation. We argue that this gap stems from failures across the execution process rather than from local manipulation capability alone. To uncover these bottlenecks, we systematically diagnose failures in long-horizon mobile manipulation and identify three dominant regimes: task planning, navigation and base positioning, and local manipulation. Based on this diagnosis, we develop failure-guided agentic intervention that augments a fixed subtask VLA with demonstration-based planning guidance, adaptive visual memory, accessibility-aware base adjustment, and local end-effector adjustment. On RoboCasa365, these interventions progressively improve task success from 19% to 47%, with the finalized system reaching 49% on held-out episode seeds, compared with 18% for the baseline. The same design improves performance from 14% to 37% on BEHAVIOR-1K and from 5/20 to 13/20 on real-robot tasks. These results suggest that reliable mobile manipulation can be advanced by diagnosing and intervening on the current bottleneck, rather than uniformly scaling the underlying policy. Our code will be released at https://github.com/PKU-VaLuE-Lab/Cracking-Mobile-Manipulation.git.

cs.RO↗

arXiv2026

MARS: Systematic Multi-Level Acceleration for World Action Models

World Action Models (WAMs) have emerged as a promising alternative for embodied intelligence by jointly predicting future world states and robot actions. However, their heavy visual encoding and iterative generative inference incur substantial latency, limiting their applicability to high-frequency closed-loop control. Existing training-free acceleration methods typically optimize a single stage, yielding limited speedups or degraded control under aggressive reuse. In this work, we systematically profile the inference overhead of WAMs and observe abundant redundancy across Transformer computation, diffusion denoising, and consecutive policy queries. However, the safe removal of such redundancy depends strongly on its impact on action prediction and control sensitivity. Based on these observations, we introduce \textbf{MARS}, a training-free and plug-and-play framework for multi-level WAM acceleration. MARS selectively removes low-impact attention computation, adaptively accelerates diffusion inference according to motion granularity, and speculatively reuses predicted future visual representations across consecutive queries with online verification. Together, these complementary mechanisms provide remarkable acceleration while limiting approximation error. Extensive experiments across LIBERO, LIBERO-PRO, and RoboCasa show that MARS achieves up to $4.00\times$, $3.94\times$, and $2.10\times$ end-to-end speedups, respectively, while maintaining task success rate close to the original policy. Moreover, MARS generalizes across different model backbones and across simulation and real-robot experiments, substantially improving the efficiency--performance trade-off of WAM inference. Our code is available at https://github.com/PKU-VaLuE-Lab/MARS.

cs.CV↗

arXiv2026

OpenSocialJax: Evaluating Social Adaptation in Sequential Social Dilemmas with Open-Ended Environmental Dynamics

Sequential social dilemma benchmarks capture diverse social interactions, but rarely test whether agents can infer hidden environmental rules through repeated interaction while adapting to other agents. We introduce \textbf{OpenSocialJax}, which extends Clean Up and Commons Harvest across an open-ended distribution of latent environment configurations, systematically varying spatial layouts and hidden rules governing tool crafting, payoffs, and resource dynamics. Each episode samples one configuration from tens of millions of possibilities and holds it fixed across five trials. We evaluate eight large language models, including three closed source models and five open source models. Our evaluation tests whether models can discover hidden rules and adapt to other agents in homogeneous and mixed teams. We find that rule discovery often improves performance but does not reliably sustain cooperation, as the cleaning burden becomes concentrated among fewer agents in OpenCleanup and every team eventually exhausts the shared resource in OpenHarvest. In our mixed arena, one cooperative agent was able to sustain the public good in OpenCleanup at a cost to its own return, but could not prevent resource depletion in OpenHarvest. Together, these results highlight the difficulty of adapting jointly to unfamiliar environmental dynamics and the changing behavior of other agents.

cs.MA↗

arXiv2026

Memento 3: Model-Based Recursive Self-Improvement through Reflective Rulebooks

Learning to act in unfamiliar environments requires agents to infer how the world works and revise that understanding as new evidence arrives. Yet limited observations can support multiple world models that explain past interactions but predict different outcomes in unseen states. We introduce Memento 3, building on the Memento series to enable frozen LLM agents to continually learn explicit world models through external memory. The agent maintains a natural-language rulebook as persistent semantic memory, recording revisable hypotheses about environment dynamics while leaving unknown aspects underspecified. It compiles this rulebook into executable code for prediction and planning. Through a continual loop of observation, reflection, rule revision, compilation, and verification, the agent uses prediction errors to refine both the rulebook and its code. Updated code is accepted only when the LLM judges it faithful to the rulebook and cell-exact replay reproduces the observed transitions. We investigate this process as a model-based route to recursive self-improvement (RSI): the agent autonomously explores the environment, revises its world model, and uses verified updates to guide subsequent interaction and learning, while the underlying LLM remains fixed. A population extension maintains multiple world models in parallel, sharing interaction evidence and using their predictions to guide exploration. On ARC-AGI-3, the single-model agent clears every level of all 25 public games, achieves a mean Relative Human Action Efficiency (RHAE) of 100.0, and uses 44% of the human action count. In an Atari Pong case study, a learned feedback controller wins 21:0 in each of three evaluated episodes with different openings, without further LLM calls.

cs.AI↗

arXiv2026

GPTBIAS: A Comprehensive Framework for Evaluating Bias in Large Language Models

Warning: This paper contains content that may be offensive or upsetting. There has been a significant increase in the usage of large language models (LLMs) in various applications, both in their original form and through fine-tuned adaptations. As a result, LLMs have gained popularity and are being widely adopted by a large user community. However, one of the concerns with LLMs is the potential generation of socially biased content. The existing evaluation methods have many constraints, and their results exhibit a limited degree of interpretability. In this work, we propose a bias evaluation framework named GPTBIAS that leverages the high performance of LLMs (e.g., GPT-4 \cite{openai2023gpt4}) to assess bias in models. We also introduce prompts called Bias Attack Instructions, which are specifically designed for evaluating model bias. To enhance the credibility and interpretability of bias evaluation, our framework not only provides a bias score but also offers detailed information, including bias types, affected demographics, keywords, reasons behind the biases, and suggestions for improvement. We conduct extensive experiments to demonstrate the effectiveness and usability of our bias evaluation framework.

cs.CL↗

arXiv2026

Bilevel Coordinated Reflection: A Game-Theoretic Approach to Multi-Agent LLM Systems

Multi-agent LLM systems commonly use an orchestrator to decompose a task for a team of workers and then improve through textual reflection. Despite strong empirical results, these systems lack a unified account of coordination, memory improvement, and the role of external verification. We model orchestrator-worker interaction as a bilevel coordination game: under bounded coupling, the workers' local-update game is an approximate potential game whose equilibrium slack is controlled by decomposition quality. We then analyse reflection as stochastic movement over semantic memory states. For free-form reflection, we derive a finite-time upper bound, prove worst-case tightness, and give a positive lower bound under a falsifiable persistent-harm condition. We further prove an information-theoretic impossibility result: no gate that observes only the generated transcript can improve uniformly over text-indistinguishable environments, whereas an environment-grounded gate can. Motivated by this separation, we introduce Stochastic Reflective Memory Ascent (SRMA), which accepts a candidate memory only after a grounded evaluation risk strictly decreases. Under calibration and non-degenerate corrective mass, SRMA converges exactly, geometrically or polynomially; matching constructions show that both rate regimes are order-tight. We also provide confidence gating for stochastic evaluation and re-anchoring guarantees for piecewise-stationary environments. Experiments instantiate these objects with environment-grounded metrics and test the predicted coordination and drift laws. On 500 SWE-bench instances, the complete Kimi-based system resolves 72.2% versus a 70.8% public mini-SWE-agent reference. Code: https://github.com/YihangChen9/Bilevel-Coordinated-Reflection

cs.AI↗

arXiv2026

SocialReasonBench: A Video-QA Benchmark for Social Reasoning with Counterfactual Narrative Videos

Recent advances in Large Multimodal Models (LMMs) have greatly improved video understanding, yet their ability to reason about human-centered social situations remains limited. Existing benchmarks typically rely on videos with a single observed trajectory, making it difficult to determine whether models truly understand social dynamics or merely exploit recurring narrative patterns. We introduce SocialReasonBench, a video multiple-choice QA benchmark for evaluating socially grounded reasoning in scenarios derived from interactive narratives. Built from gameplay videos of Detroit: Become Human, the benchmark leverages branching storylines where player decisions lead to alternative social outcomes that can be checked against the game's own script, flowchart, and recorded branches. We develop a multi-agent curation pipeline that localizes socially meaningful clips, grounds answer labels in game-state signals, and generates theory-guided questions with diagnostic distractors. SocialReasonBench covers seven reasoning dimensions, including intent recognition, emotional empathy, moral dilemma, counterfactual reasoning, and causal antecedent. Experiments on contemporary LMMs show that models perform reasonably well on basic social understanding but struggle with counterfactual and causal reasoning. Further ablation and diagnostic error analyses reveal that models often depend on incomplete modality cues and fall into reasoning traps such as visual shortcuts, highlighting a gap between observable event recognition and deeper reasoning over latent social states.

cs.CL↗

arXiv2026

SAGE: From Direct Answering to Evidence-Grounded Inference for Chinese Ancient Document Understanding

Chinese ancient document understanding demands complex visual, linguistic, and historical reasoning. Current Large Vision-Language Models (LVLMs) typically rely on an opaque, single-pass generation paradigm, often producing overconfident and weakly grounded responses. To address this, we propose SAGE, an evidence-grounded multi-agent framework that reformulates Chinese ancient document understanding as evidence-grounded inference rather than direct answer generation. SAGE coordinates specialized agents for task-aware planning, tool-mediated evidence acquisition, claim-level verification, and bounded replanning under a constrained shared-state runtime. This design supports bounded evidence seeking, answer revision, and abstention when grounding is insufficient. Experiments on the AncientDoc benchmark show that SAGE consistently outperforms matched direct-answering baselines across three LVLM backbones. Remarkably, SAGE with Qwen3.5-9B surpasses much larger monolithic LVLMs on most evaluated metrics, highlighting the importance of structured, evidence-grounded inference beyond model scaling.

cs.CL↗

arXiv2026

Scaling GUI Agents with Visual State Transitions

We introduce State Transition Pretraining (STP) as a new scaling axis for GUI agents. During the STP stage, we continually pretrain a unified multimodal model on visual state transitions by jointly optimizing inverse dynamics (predicting actions from state changes) and forward dynamics (predicting next states from current states and actions). This optimization equips the model with better action-grounded visual representations and an internal world model of GUI dynamics. When subsequently fine-tuned on trajectories with task instructions, our STP-trained models consistently outperform baselines trained solely via direct trajectory fine-tuning across agent benchmarks in both desktop and mobile GUI scenarios (AgentNetBench, AndroidControl, and GUIOdyssey). Further empirical studies show that joint dynamics optimization yields stable improvements over single-objective training, and downstream performance scales steadily with the volume of transition data.

cs.AI↗

arXiv2026

NeMo: Needle in a Montage for Video-Language Understanding

Recent advances in video large language models (VideoLLMs) call for new evaluation protocols and benchmarks for video-language understanding. Inspired by the needle in a haystack test widely used by LLMs, we introduce a novel task of Needle in a Montage (NeMo), which is designed to assess the temporal understanding capabilities of advanced VideoLLMs. Specifically, the proposed task focuses on two fundamental abilities critical for temporal understanding, i.e., retrieval-style long-context recall and temporal grounding. To generate video question answering data for our task, we develop a scalable automated data generation pipeline that facilitates high-quality data synthesis. Built upon the proposed pipeline, we present NeMoBench, a video-language benchmark centered on our task. Specifically, our full set of NeMoBench features 31,378 automatically generated question-answer (QA) pairs from 13,486 videos with various durations ranging from seconds to hours. Experiments demonstrate that our pipeline can reliably and automatically generate high-quality evaluation data, enabling NeMoBench to be continuously updated with the latest videos. We evaluate 20 state-of-the-art models on our benchmark, providing extensive results and key insights into their capabilities and limitations. Our project page is available at: https://lavi-lab.github.io/NeMoBench.

cs.CV↗

arXiv2026

MEAL: A Benchmark for Continual Multi-Agent Reinforcement Learning

Benchmarks play a central role in reinforcement learning (RL) research, yet their computational constraints often shape what is studied. Despite the motivation of lifelong learning, most continual RL papers consider only 3-10 sequential tasks, as CPU-bound environments make longer sequences impractical. Meanwhile, continual learning in cooperative multi-agent settings remains largely unexplored. To address these gaps, we introduce MEAL (Multi-agent Environments for Adaptive Learning), the first benchmark for continual multi-agent RL. By leveraging JAX and GPU acceleration, MEAL enables training on sequences of 100 tasks in a few hours on a single GPU. We find that long task sequences reveal failure modes that do not appear at smaller scales.

cs.AI↗

arXiv2026

Self-evolving LLM agents with in-distribution Optimization

Large Language Models (LLMs) have recently emerged as powerful controllers for interactive agents in complex environments, yet training them to perform reliable long-horizon decision making remains a fundamental challenge. A key difficulty lies in credit assignment: agents often receive delayed rewards only at the end of episodes. In this paper, we propose Q-Evolve, a self-evolving framework for LLM agents that unifies automatic process-reward labeling and policy learning within a principled in-distribution reinforcement learning paradigm. In each evolving iteration, our method learns an in-distribution critic from a hybrid off-policy dataset that combines expert demonstrations with agent-generated trajectories, stabilizing Bellman backups in sparse-reward settings via a weighted Implicit Q-Learning objective. The learned value function is then used to derive step-wise process rewards through advantage estimation, enabling dense and reliable supervision without environment backtracking or human annotation. Leveraging these signals, we perform behavior-proximal policy optimization that evolves the agent over the data used for process reward labeling, allowing iterative self-improvement without exacerbating distribution shift. We evaluate our method on AlfWorld, WebShop, and ScienceWorld, showing Q-Evolve outperforms strong baselines in sample efficiency, robustness, and overall task performance. Our results demonstrate that stable agent self-evolution is achievable through the co-evolution of process-level supervision and policy, both grounded within a shared in-distribution learning loop.

cs.LG↗

arXiv2026

MACCA: Offline Multi-agent Reinforcement Learning with Causal Credit Assignment

Offline Multi-agent Reinforcement Learning (MARL) is valuable in scenarios where online interaction is impractical or risky. While independent learning in MARL offers flexibility and scalability, accurately assigning credit to individual agents in offline settings poses challenges because interactions with an environment are prohibited. In this paper, we propose a new framework, namely Multi-Agent Causal Credit Assignment (MACCA), to address credit assignment in the offline MARL setting. Our approach, MACCA, characterizing the generative process as a Dynamic Bayesian Network, captures relationships between environmental variables, states, actions, and rewards. Estimating this model on offline data, MACCA can learn each agent's contribution by analyzing the causal relationship of their individual rewards, ensuring accurate and interpretable credit assignment. Additionally, the modularity of our approach allows it to integrate with various offline MARL methods seamlessly. Theoretically, we proved that under the setting of the offline dataset, the underlying causal structure and the function for generating the individual rewards of agents are identifiable, which laid the foundation for the correctness of our modeling. In our experiments, we demonstrate that MACCA not only outperforms state-of-the-art methods but also enhances performance when integrated with other backbones.

cs.LG↗

arXiv2026

PaperVoyager : Building Interactive Web with Visual Language Models

Recent advances in visual language models have enabled autonomous agents for complex reasoning, tool use, and document understanding. However, existing document agents mainly transform papers into static artifacts such as summaries, webpages, or slides, which are insufficient for technical papers involving dynamic mechanisms and state transitions. In this work, we propose a Paper-to-Interactive-System Agent that converts research papers into executable interactive web systems. Given a PDF paper, the agent performs end-to-end processing without human intervention, including paper understanding, system modeling, and interactive webpage synthesis, enabling users to manipulate inputs and observe dynamic behaviors. To evaluate this task, we introduce a benchmark of 19 research papers paired with expert-built interactive systems as ground truth. We further propose PaperVoyager, a structured generation framework that explicitly models mechanisms and interaction logic during synthesis. Experiments show that PaperVoyager significantly improves the quality of generated interactive systems, offering a new paradigm for interactive scientific paper understanding.

cs.CL↗

arXiv2026

I-WebGenBench : Evaluating Interactivity in LLM-Generated Scientific Web Applications

cs.CL↗

arXiv2026

Web2BigTable: A Bi-Level Multi-Agent LLM System for Internet-Scale Information Search and Extraction

Agentic web search increasingly faces two distinct demands: deep reasoning over a single target, and structured aggregation across many entities and heterogeneous sources. Current systems struggle on both fronts. Breadth-oriented tasks demand schema-aligned outputs with wide coverage and cross-entity consistency, while depth-oriented tasks require coherent reasoning over long, branching search trajectories. We introduce \textbf{Web2BigTable}, a multi-agent framework for web-to-table search that supports both regimes. Web2BigTable adopts a bi-level architecture in which an upper-level orchestrator decomposes the task into sub-problems and lower-level worker agents solve them in parallel. Through a closed-loop run--verify--reflect process, the framework jointly improves decomposition and execution over time via persistent, human-readable external memory, with self-evolving updates to each single-agent. During execution, workers coordinate through a shared workspace that makes partial findings visible, allowing them to reduce redundant exploration, reconcile conflicting evidence, and adapt to emerging coverage gaps. Web2BigTable sets a new state of the art on WideSearch, reaching an Avg@4 Success Rate of \textbf{38.50} ($7.5\times$ the second best at 5.10), Row F1 of \textbf{63.53} (+25.03 over the second best), and Item F1 of \textbf{80.12} (+14.42 over the second best). It also generalises to depth-oriented search on XBench-DeepSearch, achieving 73.0 accuracy. Code is available at https://github.com/web2bigtable/web2bigtable.

cs.AI↗