Meta FAIR · Self-Improving Memory

Self-Improving Memory

Meta FAIR's Self-Improving Memory lets Kimi-K2.6 write its own LOAD and OFFLOAD memory prompts, then distils them into its weights. On Continual Learning Bench, the optimised prompts score 56.0 against 44.6 for the best hand-written harness. The page text gives no licence or code availability.

Open the source ↗

Similar Research & Safety releases

More from Meta FAIR

Published by Quanta AI Labs, an AI agency in India. Browse every week in AI News Weekly.