RAM: Towards an Ever-Improving Memory System by Learning from Communications
This addresses the challenge of dynamic knowledge acquisition and lifelong learning for AI systems, though it appears incremental as it builds on existing RAG approaches.
The paper tackles the problem of improving retrieval-augmented generation (RAG) systems by introducing RAM, a framework with an ever-improving memory that learns from user feedback, resulting in significant improvements over traditional methods, especially in handling false premise and multi-hop questions.
We introduce an innovative RAG-based framework with an ever-improving memory. Inspired by humans'pedagogical process, RAM utilizes recursively reasoning-based retrieval and experience reflections to continually update the memory and learn from users' communicative feedback, namely communicative learning. Extensive experiments with both simulated and real users demonstrate significant improvements over traditional RAG and self-knowledge methods, particularly excelling in handling false premise and multi-hop questions. Furthermore, RAM exhibits promising adaptability to various feedback and retrieval methods, showcasing its potential for advancing AI capabilities in dynamic knowledge acquisition and lifelong learning.