From Isolated Conversations to Hierarchical Schemas: Dynamic Tree Memory Representation for LLMs
This addresses memory management challenges for LLMs handling complex reasoning and extended interactions, though it appears incremental as an improvement over existing memory augmentation methods.
The paper tackles the problem of effective long-term memory management in large language models by introducing MemTree, a dynamic tree-structured memory representation algorithm that organizes information hierarchically. Evaluations on multi-turn dialogue and document QA benchmarks show it significantly enhances performance in scenarios requiring structured memory.
Recent advancements in large language models have significantly improved their context windows, yet challenges in effective long-term memory management remain. We introduce MemTree, an algorithm that leverages a dynamic, tree-structured memory representation to optimize the organization, retrieval, and integration of information, akin to human cognitive schemas. MemTree organizes memory hierarchically, with each node encapsulating aggregated textual content, corresponding semantic embeddings, and varying abstraction levels across the tree's depths. Our algorithm dynamically adapts this memory structure by computing and comparing semantic embeddings of new and existing information to enrich the model's context-awareness. This approach allows MemTree to handle complex reasoning and extended interactions more effectively than traditional memory augmentation methods, which often rely on flat lookup tables. Evaluations on benchmarks for multi-turn dialogue understanding and document question answering show that MemTree significantly enhances performance in scenarios that demand structured memory management.