AI agents gain step-change memory with just 0.12% extra parameters
A new memory module compresses an AI agent’s past interactions into a fixed-size matrix, cutting latency and token costs while reducing reliance on costly context expansion or external retrieval systems.