Show HN: Fixing AI memory blind spot on connected facts with benchmark

Sobre

Semantic search alone is not enough to capture all connected facts, they will capture the semantically most identical memory only.<p>Tested on HotpotQA public dataset:<p>Vector + BM25 + entity graph: BothFound@5 71.5% Vector + BM25 only: BothFound@5 59.5%<p>Entity graph is the game changer to extract connected facts.<p>More Benchmark result:<p>LongMemEval-S: 84.8% recallAll@5 LoCoMo-10: 59% vs zep cloud 28%<p>What is your approach for connected facts retrieval ?