Reproduce “TA-Mem: Tool-Augmented Autonomous Memory Retrieval for LLM in Long-Term Conversational QA” on OpenResearch
Turn your favorite coding agents into research agents that review literature, develop hypotheses, and run experiments.
Universal app for Apple silicon and Intel. Requires macOS 11 or later.
About this paper
Mengwei Yuan, Jianan Liu, Jing Yang, Xianyou Li, Weiran Yan, Yichao Wu, Penghao Liang
TA-Mem is a tool-augmented framework that empowers Large Language Models to autonomously retrieve information from memory in long-term conversational QA. It achieves superior performance, particularly on temporal and multi-hop questions, while maintaining token efficiency by adapting retrieval strategies to query types.
Key insights
- Empowering LLMs with autonomous tool selection for memory retrieval enables dynamic adaptation to specific query types, moving beyond static, one-size-fits-all approaches.
- Structuring memory into rich, multi-indexed notes (e.g., facts, events, persons, tags) provides versatile access points, allowing for more targeted and precise information retrieval.
- Integrating an iterative agentic loop with a memory cache optimizes token usage by focusing on relevant context and preventing redundant fetching, even with multiple LLM interactions.
Summary by alphaXiv.
Tip: change arxiv.org to autoarxiv.org in any arXiv abstract link to open the paper here.