QEMem Framework Addresses Relevance-Sufficiency Gap in LLM Agent Memory

QEMem shifts agent memory from simple relevance-based retrieval to the construction of query-specific evidence sets. The framework uses query-guided discovery and source-grounded reification to organize fragmented information into structured ViewCards, ensuring the agent has sufficient collective evidence for reasoning.