Back to AI News
arxiv

ENGRAM: From Agent Experience to Weights through Verified Consolidation

ENGRAM is a framework for consolidating a language-model agent's stored experience into model weights while separating training admission from memory release. A demonstration-aware gate admits skills based on student mastery or sufficient successful teacher demonstrations, and the student learns from selected action traces with context from an Engram Trace Graph. A verification gate then compares the updated model without episodic retrieval against the base model with retrieval on reserved probes, and only releases records whose scores meet relative and absolute criteria. Experiments across interactive and reasoning benchmarks report gains over frozen-weight controls, a reduced retrievable footprint close to retain-all performance, and lower input-token usage and model calls.

AI Brief

It offers a concrete, verifiable method for deciding when agent experience can be consolidated into weights and safely removed from retrieval, which is directly useful to teams building memory-equipped agents.

arXiv:2611.09472v1 Announce Type: new Abstract: Language-model agents use experience through external retrieval and parameter learning. Consolidating stored experience into model weights requires deciding both which experience to train on and when its external records can leave retrieval. Admission based solely on student mastery can exclude useful teacher demonstrations, while removal immediately after training can discard guidance that the model still needs. We introduce ENGRAM, a framework that separates training admission from memory release. A demonstration-aware gate admits skills based on student mastery or sufficient successful teacher demonstrations. The student learns from selected action traces with context from an Engram Trace Graph (ETG). A verification gate then compares the updated model without episodic retrieval against the base model with retrieval on reserved probes. Release requires valid training participation and scores meeting both relative and absolute criteria; records for skills that fail verification remain retrievable. Experiments across interactive and reasoning benchmarks show consistent gains over frozen-weight ETG controls, with ablations supporting demonstration-aware admission and selective adapter activation. A controlled held-out comparison shows that verified release substantially reduces the retrievable footprint while remaining close to retain-all performance and outperforming immediate and budget-matched random release. ENGRAM also improves task performance while reducing input-token usage and model calls. Skill-level verification further reduces retrievable experience through selective archival, retaining external guidance for skills that do not meet the release criteria. Code is available in anonymous.4open.science/r/engram-anonymous.
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!