FlyHash for LLM Agent Memory: An Empirical Validation and a New Application

A study empirically validated FlyHash, a locality-sensitive hashing scheme modeled on the fruit fly's olfactory circuit, as an alternative to dense vector embeddings searched with approximate nearest-neighbor methods for large language model agent memory, and proposed a new application for it. The research targets the external memory stores LLM agents use to retain information across conversations. The work matters because it tests a biologically inspired hashing approach against the standard embedding-plus-ANN retrieval stack that agent memory systems depend on.

Large language model LLM agents increasingly rely on external memory stores to retain information across conversations, typically via dense vector embeddings searched with approximate nearest-neighbor methods. The fruit fly's olfactory circuit computes a locality-sensitive hash of sensory input us