EvoMemState framework evolves memory states for LLM personal agents
Original title: EvoMemState: Evolving Memory State for Personal Agents
EvoMemState augments facts with relational and reasoning states to improve personalized assistance in agents.
The paper introduces EvoMemState, a state-centric framework for LLM agents that maintains memory validity and applies facts to requests. It includes ontology-guided induction, RL-based evolution, and state-aware reasoning. Experiments on dialogue and memory benchmarks show better retrieval and accuracy.
- Relational States capture fact validity and relations
- RL policy generates request-specific Reasoning State
- Improvements in retrieval completeness and response accuracy
- Organizations
- XMUDeepLIT
From the original article
1 min read at huggingface.co
Long-term memory enables large language model (LLM) agents to provide continuous personalized assistance by retaining user information across interactions. Existing systems support this process through fact extraction, organization, updating, and retrieval, but largely maintain memories as independent facts.
By Zhishang Xiang, Yunbo Tang, Zhongqiao Liu · huggingface.co
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.