EviMem: Evidence-Gap-Driven Iterative Retrieval for Long-Term Conversational Memory
About
Long-term conversational memory requires retrieving evidence scattered across multiple sessions, yet single-pass retrieval fails on temporal and multi-hop questions. Existing iterative methods refine queries via generated content or document-level signals, but none explicitly diagnoses the evidence gap, namely what is missing from the accumulated retrieval set, leaving query refinement untargeted. We present EviMem, combining IRIS (Iterative Retrieval via Insufficiency Signals), a closed-loop framework that detects evidence gaps through sufficiency evaluation, diagnoses what is missing, and drives targeted query refinement, with LaceMem (Layered Architecture for Conversational Evidence Memory), a coarse-to-fine memory hierarchy supporting fine-grained gap diagnosis. On LoCoMo, EviMem improves Judge Accuracy over MIRIX on temporal (73.3% to 81.6%) and multi-hop (65.9% to 85.2%) questions at 4.5x lower latency. Code: https://github.com/AIGeeksGroup/EviMem.
Related benchmarks
| Task | Dataset | Result | Rank | |
|---|---|---|---|---|
| Long-context Conversational Question Answering | LoCoMo Overall | G-EVAL2.81 | 3 | |
| Long-context Conversational Question Answering | LoCoMo Single-Hop | G-EVAL Score2.98 | 3 | |
| Long-context Conversational Question Answering | LoCoMo Multi-Hop | G-EVAL2.89 | 3 | |
| Long-context Conversational Question Answering | LoCoMo Temporal | G-EVAL3.08 | 3 | |
| Long-context Conversational Question Answering | LoCoMo Adversarial | G-EVAL Score1.94 | 3 | |
| Long-context Conversational Question Answering | LoCoMo Open-Domain | G-EVAL3.17 | 3 |