Forgetting Without Restarting: Execution-State Unlearning for Stateful LLM Agents
Researchers propose provenance-guided selective replay letting LLM agents forget revoked information without restarts, matching full reset behavior.
The paper formalizes execution-state unlearning for stateful LLM agents, requiring that agents behave as if a revoked memory record was never observed across transcripts, compressed memory, tool plans, and KV caches. It proves exact unlearning requires at least T-τ+1 recomputed transitions and that Provenance-Guided Selective Replay attains this bound via a provenance graph, KV cache cropping, and sanitized replay. In audits across three agent suites, nine baselines, and three model families, memory deletion left leakage unchanged, instruction-based forgetting collapsed under elicitation (Leak@probes = 1.00), and selective replay matched full resets at up to 9x fewer recomputed tokens.
- Formalizes execution-state unlearning across prompt, memory, and KV cache
- Proves exact unlearning needs at least T-τ+1 recomputed transitions
- Selective replay indistinguishable from full reset at 9x fewer tokens
- Instruction-based forgetting leaks revoked data in 100% of probes
- Source redaction still leaked revoked preferences in 80% of episodes
- Organizations
- arXiv
From the original article
1 min read at arxiv.orgLong-running LLM agents are stateful: beyond the transcript they accrete compressed summaries, plaintext memory, pending tool plans, and, under every serving API, a KV cache. Yet today's "forget" operations delete a plaintext memory record and stop, leaving every artifact derived from the revoked information intact.
By Chao Yao · arxiv.org
A short excerpt; the article and its images belong to the publisher. Summary, key facts and assessment above are AI-written from the article and can be wrong.