SFT or RL for Tool-Calling Agents? A Controlled Study Across Data, Method, and Scale

2026-09-15

AI Summary

研究通过控制实验，比较了SFT与LoRA、RL（GRPO）及两者结合的方法在不同规模Qwen3模型上工具调用性能的影响，发现SFT与LoRA在多数情况下表现最佳。

Abstract

Limited controlled evidence exists on how training data, adaptation method, and model scale jointly affect tool-calling performance in language-model agents. We evaluate supervised fine-tuning (SFT) with LoRA, reinforcement learning (RL) via Group Relative Policy Optimization (GRPO), and SFT followed by GRPO across six Qwen3 models from 0.6B to 32B parameters, covering both in-distribution performance and cross-dataset transfer. SFT with LoRA is the strongest in-distribution method throughout the 0.6B-32B range and best in 15 out of 18 experimental settings. On cross-dataset transfer, the methods are closer: GRPO wins 29 out of 54 settings where training and test datasets differ, but its margin over SFT averages under one point, and SFT->GRPO is rarely strongest in either comparison. Dataset mixing gives consistently strong transfer while staying close to specialized in-distribution training, regardless of method. Additional analysis further confirms that LoRA outperforms full-parameter fine-tuning, demonstrating that LoRA better preserves pretrained agentic behavior.

Problem

• training data
• adaptation method
• model scale
• tool-calling performance
• language-model agents

Innovation

• Supervised Fine-Tuning (SFT)
• LoRA
• Reinforcement Learning (RL)
• Cross-Dataset Transfer
• Group Relative Policy Optimization (GRPO)