Back to AI News
arxiv

Falsification-Gated Self-Evolution Lets LLM Agents Test Memory Updates Before Committing

The paper proposes Falsification-Gated Self-Evolution (FGSE), a verifier-grounded framework for deciding when lessons derived from an agent episode should become persistent memory. FGSE represents each candidate lesson as a structured hypothesis with a precondition, behavior change, expected effect, and verifier, then tests it in a temporary state using target-transfer, falsification, and archived-regression probes. It commits, refines, or rejects the hypothesis based on measured gain and risk. Reported results include 39.8 WebArena success rate, 78.8 AppWorld average completion, 0.724/0.487 Pass^1 on tau-Retail/tau-Airline, 5.0% falsification failures, 1.3% regression damage, and 3.6% harmful committed updates.

AI Brief

It offers a concrete, benchmarked method for making long-term memory updates in LLM agents safer and more reliable rather than writing episode lessons directly into persistent state.

arXiv:2611.14326v1 Announce Type: new Abstract: LLM agents increasingly operate in long-horizon, verifier-rich environments, but current self-improvement mechanisms often write episode-derived lessons directly into memory, making local success an unreliable signal for reusable and safe knowledge. This paper addresses the core question of when an experience-derived behavioral update should be admitted into persistent agent state. We propose Falsification-Gated Self-Evolution (FGSE), a verifier-grounded framework that represents each candidate lesson as a structured hypothesis with an explicit precondition, behavior change, expected effect, and verifier. FGSE applies the hypothesis only in a temporary state, tests it on target-transfer, falsification, and archived-regression probes, and then commits, refines, or rejects it according to measured gain and risk. Across web, app, tool-use, and long-term memory benchmarks, FGSE achieves strong task performance, including 39.8 WebArena success rate, 78.8 AppWorld average completion, and 0.724/0.487 Pass$^1$ on $\tau$-Retail/$\tau$-Airline, while reducing falsification failures to 5.0%, regression damage to 1.3%, and harmful committed updates to 3.6%. These results suggest that testing memory updates as hypotheses supports more reliable long-term adaptation.
Be the first to rate this article.

Comments & Questions (0)

Captcha
Please be respectful — let's keep the conversation friendly.

No comments yet

Be the first to comment!