Episodic-Semantic Memory Architecture for Long-Horizon Scientific Agents

In three linesDual-process memory architecture for scientific agents: decouples episodic window (10 messages) from semantic consolidation (3 tokens/message). Evaluation on 15,000 messages across 6 LLMs (OpenAI, Anthropic, Google): maintains 70-85% accuracy at 10,000 messages with 62% fewer tokens. Identifies trade-offs: Dual Process excels at numeric/temporal queries, RAG for historical retrieval.

Summary generated by Claude — human-verified