AgentDistill: Training-Free Agent Distillation with Generalizable MCP Boxes
This addresses the challenge of building scalable and cost-efficient intelligent agents for domains requiring planning and tool use, though it is incremental as it builds on existing agent distillation concepts.
The paper tackles the problem of distilling knowledge from large language model (LLM)-based agents to smaller ones, proposing AgentDistill, a training-free framework that reuses structured task-solving modules (MCPs) to enable student agents to achieve performance comparable to advanced systems like OctoTools (GPT-4o) on biomedical and mathematical benchmarks.
While knowledge distillation has become a mature field for compressing large language models (LLMs) into smaller ones by aligning their outputs or internal representations, the distillation of LLM-based agents, which involve planning, memory, and tool use, remains relatively underexplored. Existing agent distillation methods typically replay full teacher trajectories or imitate step-by-step teacher tool usage, but they often struggle to train student agents to dynamically plan and act in novel environments. We propose AgentDistill, a novel, training-free agent distillation framework that enables efficient and scalable knowledge transfer via direct reuse of Model-Context-Protocols (MCPs), which are structured and reusable task-solving modules autonomously generated by teacher agents. The reuse of these distilled MCPs enables student agents to generalize their capabilities across domains and solve new problems with minimal supervision or human intervention. Experiments on biomedical and mathematical benchmarks demonstrate that our distilled student agents, built on small language models, can achieve performance comparable to advanced systems using large LLMs such as OctoTools (GPT-4o), highlighting the effectiveness of our framework in building scalable and cost-efficient intelligent agents.