Description

MEMI is a framework that casts robot memory modeling as an inverse problem, inferring what to remember by analyzing demonstration data in hindsight. Using models like GPT-6 Astra and Claude Fable 5.1, it optimizes memory annotations and policy prompts to help robots self-improve on long-horizon manipulation tasks, reaching human-level success rates.

Descriptions, tags, and model credits may be AI-generated or inferred from public sources and can be incomplete or wrong. Learn more.