Editing agent reasoning history boosts long-horizon task performance

The Agent-Editing World Model (AEWM) classifies actions as critical, exploratory, or noisy and revises the reasoning state, achieving 70.5% macro-F1 on action classification and improving benchmarks by 3–7 points.

Industry
Shuang Sun · Guoxin Chen · Fanzhe Meng · Jia Deng · Huatong Song · Jinhao Jiang · +3 more

Gaoling School of Artificial Intelligence, Renmin University of China

Research Digest··2 min read
The authors propose AEWM, a world model that edits agent reasoning and action histories to remove noisy or outdated state, rather than predicting environment observations.

The authors introduce the Agent-Editing World Model (AEWM), a language-based world model for LLM agents that focuses on editing the agent's own reasoning state instead of simulating tool outputs.

Why this paper

From Gaoling School of Artificial Intelligence, Renmin University of China · Part of World Model Planning for Agents, now 19 papers

In one line

Agent-Editing World Model improves long-horizon agent performance by editing noisy decisions to fix task-state contamination.

What we could check

  • ·No code link found
  • ·No weights link found
  • ·No dataset link found
  • ·No compute details found
  • ·No stated limitations found
  • ·No benchmark numbers found

Observed from the paper text and links we have. Absence here means we did not find it, not that it does not exist.

§
newspaper

Research Digest

Articles published under the Zotpaper byline are synthesized from multiple source publications by our AI editor and reviewed by our editorial process. Each story combines reporting from credible outlets to give readers a balanced, comprehensive view.