Distilling LLM Agent into Small Models with Retrieval and Code Tools

TL;DR

This paper introduces Agent Distillation, a method to transfer full task-solving abilities from large language models to smaller models using retrieval and code tools, enhancing small models' reasoning and robustness.

Contribution

It proposes a novel agent distillation framework with new prompting and self-consistent action generation techniques for small models.

Findings

Small models (0.5B-3B) achieve performance comparable to larger models.

Enhanced robustness and accuracy in reasoning tasks.

Effective transfer of reasoning and task-solving capabilities.

Abstract

Large language models (LLMs) excel at complex reasoning tasks but remain computationally expensive, limiting their practical deployment. To address this, recent works have focused on distilling reasoning capabilities into smaller language models (sLMs) using chain-of-thought (CoT) traces from teacher LLMs. However, this approach struggles in scenarios requiring rare factual knowledge or precise computation, where sLMs often hallucinate due to limited capability. In this work, we propose Agent Distillation, a framework for transferring not only reasoning capability but full task-solving behavior from LLM-based agents into sLMs with retrieval and code tools. We improve agent distillation along two complementary axes: (1) we introduce a prompting method called first-thought prefix to enhance the quality of teacher-generated trajectories; and (2) we propose a self-consistent action generation for improving test-time robustness of small agents. We evaluate our method on eight reasoning tasks across factual and mathematical domains, covering both in-domain and out-of-domain generalization. Our results show that sLMs as small as 0.5B, 1.5B, 3B parameters can achieve performance competitive with next-tier larger 1.5B, 3B, 7B models fine-tuned using CoT distillation, demonstrating the potential of agent distillation for building practical, tool-using small agents. Our code is available at https://github.com/Nardien/agent-distillation.