MemU Achieves 92.09% Accuracy on the LoCoMo Benchmark

About Long Conversation Memory

The LoCoMo (Long Conversation Memory) dataset is the industry‑standard benchmark for evaluating long‑term memory and reasoning in conversational AI systems. It is built from very long multi‑session dialogues with rich temporal, personal, and event‑driven context, enabling comprehensive evaluation across multiple reasoning categories such as single‑hop retrieval, multi‑hop inference, temporal understanding, and open‑domain question answering. Originally developed through a hybrid human–machine annotation process and adopted widely by the research community, LoCoMo provides a standardized framework to assess an AI’s ability to recall, reason, and persist information across extended interactions.