I accidentally turned LLM memory into program analysis
Why I stopped trying to give LLM agents a better memory and instead built Lemmalog, a Datalog engine that maintains an agent's knowledge as analysis state, with provenance, retractions and incremental evaluation, plus what happened when I benchmarked it on LongMemEval and LoCoMo.
Summary
After shifting focus from enhancing LLM memory, the author developed Lemmalog, a Datalog engine designed to manage an agent’s knowledge effectively. The article details its features like provenance and incremental evaluation, along with performance insights from benchmarks on LongMemEval and LoCoMo.
Comments (69 of ) ⌛