智能体
arxiv.org ·
Agent-Editing World Model：重新思考 LLM 智能体的世界建模
该论文提出 Agent-Editing World Model（AEWM），认为在真实反馈可用时，重建高熵、依赖执行的工具响应价值有限，转而建模推理与动作如何塑造未来任务进展。AEWM 包含 Action Judge（区分 Critical、Exploratory、Noisy 决策）与 State Revision（编辑同一观测历史中的噪声推理-动作续写），EditAct 将二者与真实执行结合，直接改变后续决策所依赖的状态。作者在 Search、Terminal、Software Engineering 上通过 mid-training 与监督微调训练 AEWM，报告 Action Judge 基准 macro-F1 达 70.5%，超最强前沿基线 10.6 分；在六个基准、三种 agent backbone 上 EditAct 平均提升 3.2–6.7 分；基于已验证 EditAct 轨迹的拒绝采样微调 AEWM-RFT 较 Self-RFT 提升 2.2–2.6 分。
AI推荐理由 arXiv 新论文，提出 Agent-Editing World Model（AEWM），不再模拟工具响应，而是建模推理与动作如何影响任务进展，并用 Action Judge 区分 Critical/Exploratory/Noisy 决策、State Revision 编辑噪声推理-动作轨迹，EditAct 与真实执行结合。摘要给出具体量化结果：Action Judge 基准 macro-F1 70.5%，超最强前沿基线 10.6 分；六个基准、三种 agent backbone 上 EditAct 平均提升 3.2–6.7 分；AEWM-RFT 较 Self-RFT 提升 2.2–2.6 分。针对 LLM agent 的 task-state contamination 问题，方向具体且有可验证指标，对 agent 研究有参考价值。局限：结果均为论文自报，尚无第三方复现，且未提供代码或数据可用性信息。
82
正文
Computer Science > Computation and Language
arXiv:2609.28416 (cs)
[Submitted on 23 Sep 2026]
Title:Agent-Editing World Model: Rethinking World Modeling for LLM Agents
Authors:Shuang Sun, Guoxin Chen, Fanzhe Meng, Jia Deng, Huatong Song, Jinhao Jiang, Wayne Xin Zhao, Hongteng Xu, Ji-Rong Wen
View a PDF of the paper titled Agent-Editing World Model: Rethinking World Modeling for LLM Agents, by Shuang Sun and 8 other authors
View PDF
HTML (experimental)
Abstract:Recent advances in large language models (LLMs) have enabled agents to tackle long-horizon tasks across diverse environments. To further improve agent performance, existing language world models typically predict environment observations, yet reconstructing high-entropy, execution-dependent tool responses offers limited value when real feedback is available. Meanwhile, agents suffer from \emph{task-state contamination}, where unsupported assumptions and outdated plans persist in history and distort subsequent decisions. We propose the \textbf{Agent-Editing World Model (AEWM)}, which models how reasoning and actions shape future task progress rather than simulating tool responses. AEWM combines \textbf{Action Judge} to distinguish \textsc{Critical}, \textsc{Exploratory}, and \textsc{Noisy} decisions with \textbf{State Revision} to edit noisy reasoning--action continuations from the same observed history. \textbf{EditAct} integrates these capabilities with real execution, directly changing the state underlying subsequent decisions rather than merely providing critiques. We train AEWM across Search, Terminal, and Software Engineering through mid-training and supervised fine-tuning. AEWM achieves 70.5\% macro-F1 on our Action Judge benchmark, exceeding the strongest frontier baseline by 10.6 points. Across six benchmarks and three agent backbones, EditAct improves average scores by 3.2--6.7 points over the strongest baseline. Furthermore, rejection sampling fine-tuning on verified EditAct trajectories, termed \textbf{AEWM-RFT}, improves over Self-RFT by 2.2--2.6 points across three domains without online AEWM guidance.
Subjects:
Computation and Language (cs.CL); Artificial Intelligence (cs.AI); Machine Learning (cs.LG)
Cite as:
arXiv:2609.28416 [cs.CL]
(or
arXiv:2609.28416v1 [cs.CL] for this version)
https://doi.org/10.48550/arXiv.2609.28416
Focus to learn more
arXiv-issued DOI via DataCite (pending registration)
Submission history
From: Shuang Sun [view email]
[v1]
Wed, 23 Sep 2026 17:18:26 UTC (2,129 KB)
Full-text links:
Access Paper:
View a PDF of the paper titled Agent-Editing World Model: Rethinking World Modeling for LLM Agents, by Shuang Sun and 8 other authors
View PDF
HTML (experimental)
TeX Source
view license
Current browse context:
cs.CL
new
|
recent
| 2026-09
Change to browse by:
cs
cs.AI
cs.LG
References & Citations
NASA ADS
Google Scholar
Semantic Scholar
export BibTeX citation
Loading...
BibTeX formatted citation
×
loading...
Data provided by:
Bookmark
Bibliographic Tools
Bibliographic and Citation Tools
Bibliographic Explorer Toggle
Bibliographic Explorer (What is the Explorer?)
Connected Papers Toggle
Connected Papers (What is Connected Papers?)
Litmaps Toggle
Litmaps (What is Litmaps?)
scite.ai Toggle
scite Smart Citations (What are Smart Citations?)
Code, Data, Media
Code, Data and Media Associated with this Article
alphaXiv Toggle
alphaXiv (What is alphaXiv?)
Links to Code Toggle
CatalyzeX Code Finder for Papers (What is CatalyzeX?)
DagsHub Toggle
DagsHub (What is DagsHub?)
GotitPub Toggle
Gotit.pub (What is GotitPub?)
Huggingface Toggle
Hugging Face (What is Huggingface?)
ScienceCast Toggle
ScienceCast (What is ScienceCast?)
Demos
Demos
Replicate Toggle
Replicate (What is Replicate?)
Spaces Toggle
Hugging Face Spaces (What is Spaces?)
Spaces Toggle
TXYZ.AI (What is TXYZ.AI?)
Related Papers
Recommenders and Search Tools
Link to Influence Flower
Influence Flower (What are Influence Flowers?)
Core recommender toggle
CORE Recommender (What is CORE?)
Author
Venue
Institution
Topic
About arXivLabs
arXivLabs: experimental projects with community collaborators
arXivLabs is a framework that allows collaborators to develop and share new arXiv features directly on our website.
Both individuals and organizations that work with arXivLabs have embraced and accepted our values of openness, community, excellence, and user data privacy. arXiv is committed to these values and only works with partners that adhere to them.
Have an idea for a project that will add value for arXiv's community? Learn more about arXivLabs.
Which authors of this paper are endorsers? |
Disable MathJax (What is MathJax?)