Plain retrieval over raw chat history beats complex agent memory in six of seven settings
Raw-RAG keeps the raw conversation, finds relevant past messages by meaning, reranks them, and hands them to the answering model, with no rewriting of memory. Across LoCoMo, LongMemEval, MEME and MemoryAgentBench, it ranked first in six of seven benchmark-model settings, ahead of the strongest agentic memory system by 0.13-5.53 points. The authors say keeping raw evidence, not rewritten memories, is what matters.