The paper presents the first dataset that captures how humans learn and use mechanisms in unfamiliar environments, using ARC‑AGI‑3 as a testbed. It includes 144 trajectories from 64 participants across 25 public games, each paired with time‑synchronized actions and think‑aloud audio, and 250 trajectories from large language model agents with and without harnesses. The dataset is annotated for mechanism triggers, action purposes, errors, and recovery, enabling a comparison of human and LLM cognition, highlighting differences in hypothesis testing, exploration, and use of qualitative representations.
By Shuzhi Liu, Archana Warrier, Dat Nguyen, Chengqi Zheng, Kevin Ellis, Yewen Pu
arXiv:2609.16679v1 Announce Type: new
Abstract: Foundation models, alongside advances in learned game-world models, are reshaping AI across the game lifecycle. Beyond playing games, recent systems mo...
By Meng Luo, Yanlin Li, Hao Li, Hongzhan Lin, Pengfei Zhou, Tianjie Ju, Ran Zhang, Yeying Jin, Mong-Li Lee, Wynne Hsu
arXiv:2607. 26598v1 Announce Type: cross Abstract: Large language model (LLM) agents may recover from a failure within an episode or after a retry, yet the same execution failure can recur in later tasks because post-episode feedback rarely revises the persistent harness that guides future interactions.
By Yuetian Du, Yucheng Wang, He Xu, Jiexu Xu, Shanwen Tan, Bing Zhao, Boyu Yang, Zhijie Xu, Ming Kong, Hu Wei, Jie Liu, Qiang Zhu
Learn2Play Bench is a new benchmark that tests how well large language model agents learn from experience in unfamiliar, text‑based games with novel or counterintuitive rules. The benchmark provides reproducible feedback, automatic scoring, and varied game instances to evaluate learning across repeated attempts and transfer to new situations. Findings show that retaining full action records aids learning, human players outperform agents, and the choice of harness significantly impacts performance and inference cost.
By Yibo Li, Jinhang Qiu, Zhi Zheng, Qianyun Guo, Jiaying Wu, Shuo Ji, Bryan Hooi
The paper introduces ReTRAIL, a framework that allows long‑horizon agents to rewind the environment to earlier states while preserving forward progress in knowledge. By training a shared policy with State‑Anchored GRPO and a branch‑advancement reward, ReTRAIL enables agents to recover from unproductive states and reuse insights from abandoned exploration. Experiments on MineSweeper, Sokoban, ALFWorld, and ScienceWorld show that this combined rewinding and experience‑reuse strategy outperforms simple rewind primitives and improves long‑horizon task solving.
By Ziang Ye, Wentao Shi, Lei Chen, Zhengzhou Cai, Yu Wang, Zhiyuan Yao, Zihan Wang, Xierui Song, Fuli Feng, Yixin Cao
arXiv:2610.14807v1 Announce Type: new
Abstract: Evaluating autonomous agents requires examining whether they can translate goals into coherent action across extended interactions. In persistent envir...
By Jianfeng Lin, Runze Wang, Haorong Hong, Yuxuan Xie, Zeno Zeng, Zhongyuan Liu
arXiv:2609.21423v1 Announce Type: new
Abstract: Online agent deployments produce abundant execution traces, while task-specific verification and expert annotation are costly to scale. We study how to...
By Siyuan Liu (Fudan University, Meituan Longcat Team), Fan Yu (Fudan University, Meituan Longcat Team), Dongyu Ru (Meituan Longcat Team), Yizhu Liu (Meituan Longcat Team), Yifan Yang (Meituan Longcat Team), Xuezhi Cao (Meituan Longcat Team), Xunliang Cai (Meituan Longcat Team), Yixin Cao (Fudan University)
arXiv:2610.10426v1 Announce Type: new
Abstract: Terminal-agent capability depends jointly on model weights and the runtime harness that formats prompts, binds tools, and handles error recovery. Exist...
By Jixuan Chen, Jiaxin Zhang, Qinyuan Ye, Yada Pruksachatkun, Haoxiang Zhang, Jingming Zhuo, Yifan Zhang, Yutong Dai, Juntao Tan, Xiangyu Peng, Silvio Savarese, Zeyuan Chen, Lianhui Qin, Chien-Sheng Wu
The paper introduces an online skill‑evolution framework that transforms interaction traces and evaluator feedback into a persistent, versioned library of reusable procedures for computer‑use agents. By executing each iteration against a frozen library snapshot, the system updates skills without altering the underlying model parameters. Experiments across four OSWorld domains show that the evolving library consistently outperforms an empty‑library baseline, with gains ranging from 5.7 to 18.6 percentage points, while also revealing domain‑specific temporal stability and challenges in skill retrieval and revision.
By Longtao Hu, Xiao Liang, Linchao Zhu
SWE-Game is a benchmark comprising 247 tasks based on 41 Godot games across 13 gameplay categories, testing coding agents on tasks such as brief-to-game, design-document implementation, skeleton completion, fault repair, and Godot-to-Unity porting. Evaluation uses engine-state checks, replay of certified reference inputs, and agent-authored demonstrations to judge mechanic correctness, playability, and post-repair behavior, supplemented by vision‑language rubrics for presentation. Across six models, Opus5 leads but overall scores stay below 60/100, highlighting common issues like omitted requirements and gameplay logic errors.
By Xiaoyu Chen, Lai Wei, Jin Wang, Xiangyu Zou, Ruochen Fan, Enze Luo, Mingzhe Yao, Jiahui Zhu, Yuhua Wen, Linghe Kong, Weiran Huang
MC‑EvoBench is a controlled Minecraft benchmark designed to audit experience‑driven self‑improvement in agents. It features 19 evolution chains (12 primary, 7 calibration/probe) and three graph‑aligned probe families, each task represented as a task‑state graph with stage‑level verifiers and strict source‑target boundaries. The benchmark evaluates probes under no‑prior, related, unrelated, and shuffled source conditions, reporting metrics such as TSR, WGP, and VMP while preserving provenance records.
By Jingwei Song, Zhisheng Chen, Bingfan Zeng, Jing Tang, Yubo Hou, Jinhao Jing, Zhengwei Xie, Linfeng Zhang
arXiv:2610.16125v1 Announce Type: new
Abstract: An agent can retain successful trajectories without learning which part of an experience should repair a new discrepancy. We formulate passive imitatio...
By Kai Hu, Hua Huang