---
format: "aidr-story-markdown/v1"
id: "c44ac57feedcf64a76ed188a295ee6205722e64fa029ddb108769ccb5c9bfd7d"
canonical_url: "https://aidr.today/c44ac57f?lang=en"
title: "Cleaning an Agent's History Hurts When the World Still Holds the Error"
lang: "en"
requested_lang: "en"
available_langs: ["en","vi"]
translation_fallback: null
fallback_fields: []
published_at: "2026-11-13T05:00:00.000Z"
category: "Research"
topics: ["agent"]
source_urls: ["https://arxiv.org/abs/2611.14654"]
summary: "Tool-using language-model agents accumulate histories of their own mistakes. On AgentCE-Bench, adding a truthful note beside a preserved error raised recovery from 51% to 66% when the world state was held fixed."
---

# Cleaning an Agent's History Hurts When the World Still Holds the Error

> [Open the canonical story](<https://aidr.today/c44ac57f?lang=en>)

**Published:** 2026-11-13T05:00:00.000Z
**Category:** Research
**Topics:** agent

## Summary

Tool\-using language\-model agents accumulate histories of their own mistakes\. On AgentCE\-Bench, adding a truthful note beside a preserved error raised recovery from 51% to 66% when the world state was held fixed\.

## Sources

- [Story source](<https://arxiv.org/abs/2611.14654>)