Honest Lying: Understanding Memory Confabulation in Reflexive Agents

Hugging Face Daily Papers Papers

Summary

This paper identifies memory confabulation in Reflexion-style agents, where agents store incorrect task interpretations and persist in errors across environment resets. The authors introduce the Reflection Repetition Rate (RRR) metric to detect this and propose a mitigation that replaces open-ended self-diagnosis with programmatic failure signal extraction.

Reflexion-style agents rely on self-generated reflections as memory, implicitly assuming that agents can accurately diagnose their own failures. We show that this assumption can fail systematically: across ALFWorld and HumanEval, agents store confident but incorrect interpretations of the task and continue acting on them across trials, even though the environment resets to the correct task each time. We call this failure mode memory confabulation and introduce the Reflection Repetition Rate (RRR), a log-based metric that detects repeated reliance on incorrect reflective content. Using RRR, we identify 16 frozen environments in ALFWorld, where 0 of 121 reflections mention the correct target object, and 4 analogous cases in HumanEval. Our mitigation replaces open-ended self-diagnosis with programmatic extraction of trajectory-level failure signals, increasing correct object mention from 0% to 86%, reducing RRR from 0.64 to 0.10, and solving 3 of 16 frozen ALFWorld environments, suggesting that reflective memory can reinforce false beliefs rather than correct them.
Original Article
View Cached Full Text

Cached at: 06/09/26, 08:41 AM

Paper page - Honest Lying: Understanding Memory Confabulation in Reflexive Agents

Source: https://huggingface.co/papers/2605.29463

Abstract

Agents relying on self-generated reflections can store confident but incorrect task interpretations, leading to persistent errors despite environment resets, which is identified through a new metric called Reflection Repetition Rate.

Reflexion-style agentsrely onself-generated reflectionsas memory, implicitly assuming that agents can accurately diagnose their own failures. We show that this assumption can fail systematically: across ALFWorld and HumanEval, agents store confident but incorrect interpretations of the task and continue acting on them across trials, even though the environment resets to the correct task each time. We call this failure modememory confabulationand introduce theReflection Repetition Rate(RRR), a log-based metric that detects repeated reliance on incorrect reflective content. Using RRR, we identify 16 frozen environments in ALFWorld, where 0 of 121 reflections mention the correct target object, and 4 analogous cases in HumanEval. Our mitigation replaces open-ended self-diagnosis with programmatic extraction oftrajectory-level failure signals, increasing correct object mention from 0% to 86%, reducing RRR from 0.64 to 0.10, and solving 3 of 16 frozen ALFWorld environments, suggesting that reflective memory can reinforce false beliefs rather than correct them.

View arXiv pageView PDFAdd to collection

Get this paper in your agent:

hf papers read 2605\.29463

Don’t have the latest CLI?curl \-LsSf https://hf\.co/cli/install\.sh \| bash

Models citing this paper0

No model linking this paper

Cite arxiv.org/abs/2605.29463 in a model README.md to link it from this page.

Datasets citing this paper0

No dataset linking this paper

Cite arxiv.org/abs/2605.29463 in a dataset README.md to link it from this page.

Spaces citing this paper0

No Space linking this paper

Cite arxiv.org/abs/2605.29463 in a Space README.md to link it from this page.

Collections including this paper0

No Collection including this paper

Add this paper to acollectionto link it from this page.

Similar Articles

@omarsar0: // The Memory Curse in LLM Agents // (bookmark it) Long histories apparently degrades agents as they become increasingl…

X AI KOLs Following

This research paper identifies the 'memory curse' in LLM agents, demonstrating that expanded context windows systematically degrade cooperative behavior in multi-agent social dilemmas by eroding forward-looking intent. The authors show that targeted fine-tuning, synthetic memory sanitization, and reducing explicit Chain-of-Thought reasoning can effectively mitigate this behavioral decay.