MidTool: Mid-training Data Synthesis for Agentic Tool Use
2026-08-20 • Artificial Intelligence
Artificial Intelligence
AI summaryⓘ
The authors explore improving large language models' ability to use tools by training them partway through their learning process, called mid-training. They created MidTool, a collection of data combining various sources and instructions to teach models how to understand and use tools effectively. By training models on MidTool and then fine-tuning them, the authors found that models performed better at tasks requiring tool use compared to those trained without this step. This suggests that mid-training focused on tool use can enhance models' practical abilities beyond regular training alone.
large language modelsmid-trainingtool usesupervised fine-tuningreinforcement learningAPIagentic capabilitiesworkflow compositiondata synthesisperformance benchmarks
Authors
Fengqing Jiang, Yite Wang, Boyi Liu, Zhaoyang Wang, Canwen Xu, Zhewei Yao, Radha Poovendran, Yuxiong He
Abstract
Mid-training is increasingly recognized as a critical stage for shaping the capabilities of large language models. Recent work has shown that targeted mid-training can strengthen reasoning-intensive abilities such as math and science, and can also improve agentic capabilities in software-engineering settings. In this work, we study the parallel but less explored agentic capability: general tool use. We present MidTool, an open corpus construction pipeline for agentic tool-use mid-training that combines large-scale web, PDF, and code data with synthesized supervision from real-world tool APIs, MCP skills, and document-grounded workflows. MidTool is designed to teach models how to recognize tool affordances, ground arguments from context, compose tool call workflow, and recover from incomplete information. We mid-train Qwen3-4B-Base and Qwen3-8B-Base on MidTool-Mix, and then apply follow-up post-training with both supervised fine-tuning and reinforcement learning. Compared with baselines, MidTool-Mix consistently improves downstream performance under both SFT and RL on BFCL, tau2-Bench, and MCP Universe. These results suggest that general tool use, like other important LLM capabilities, benefits from dedicated mid-training rather than being left entirely to post-training.