Tag
The paper proposes Gated-Memory Routing, a framework for multi-agent LLM systems that uses learned gates to manage memory, improving accuracy and reducing inference costs on reasoning and code-generation benchmarks.