QEMem: A framework to enhance memory utilization in LLM agents
In brief
- QEMem reframes memory use by constructing query-specific evidence sets instead of relying on individual retrieval.
- It combines query-guided discovery, source-grounded reification, and sufficiency-guided refinement.
- The approach boosts evidence coverage and accuracy by 7.6 to 10.7 points on LoCoMo.
LLM agents rely on long-term memory to reuse past experiences, but retrieving relevant memories does not always provide sufficient evidence for accurate answers. QEMem introduces a method that builds query-specific evidence sets, improving response accuracy. The framework achieves 91.43% accuracy on LoCoMo with Claude Opus 4.5, outperforming existing baselines.
This summary gives you the essentials. The full article, its figures and quotes are on cs.AI updates on arXiv.org.
Read the original articleAI-generated summary of the source, checked by rules. Report an error