Adaptive Focus Memory for Language Models

Christopher Cruz

Large language models (LLMs) are increasingly deployed in multi-turn dialogue settings, yet their behavior remains bottlenecked by naive history management strategies. Replaying th…