Reproduce “Is Complex Agentic Memory Always Needed? Naive RAG Is Stronger Than We Thought” on OpenResearch
Turn your favorite coding agents into research agents that review literature, develop hypotheses, and run experiments.
Universal app for Apple silicon and Intel. Requires macOS 11 or later.
About this paper
Zhengye Bao, Shengtao Zhang, Haokai Zhang, Shiqi Lou, Muning Wen, Minnan Luo
RAW-RAG tests whether conversation histories require memories by retrieving overlapping raw-text windows, reranking candidates, and giving the top 15 to a reader. Across four benchmarks, it ranks first among memory systems in six of seven benchmark–model settings. On MemoryAgentBench with DeepSeek-V4-Flash, it scores 53.41% versus Memora’s 47.88%, but loses to MemMachine on MEME, limiting the finding to evaluated offline conversation tasks.
Summary by alphaXiv.
Tip: change arxiv.org to autoarxiv.org in any arXiv abstract link to open the paper here.