Research
LSREP proposes a longitudinal protocol for evaluating conversational memory systems
An independent undergraduate researcher presents LSREP, a replay-based benchmark for how conversational memory accumulates and changes over 1,985 turns. ICE v2 did not outperform pure vector-RAG on the matched LongMemEval evaluation.
Every outlet that ran it 1 article
r/MachineLearningCommunityLead
LSREP: A Longitudinal State-Replay Protocol for Evaluating Conversational Memory [R]