发表机构
Dialpad Inc.（Dialpad 公司）
机构由 AI 辅助整理，请以论文原文为准。AI 中文总结
本研究通过跨数据、方法和规模的受控实验，发现LoRA监督微调在分布内最强，而强化学习在跨数据集迁移中略有优势，且LoRA优于全参数微调。
AI 中文摘要
关于训练数据、适应方法和模型规模如何共同影响语言模型智能体的工具调用性能，目前缺乏受控实验证据。我们评估了使用LoRA的监督微调（SFT）、通过组相对策略优化（GRPO）的强化学习（RL），以及SFT后接GRPO这三种方法，覆盖了从0.6B到32B参数的六个Qwen3模型，同时考察了分布内性能和跨数据集迁移。在0.6B-32B范围内，使用LoRA的SFT是分布内最强的方法，在18个实验设置中的15个中表现最佳。在跨数据集迁移上，各方法差距较小：GRPO在训练与测试数据集不同的54个设置中赢得29个，但其对SFT的平均优势不足一个百分点，而SFT->GRPO在两种比较中都很少最强。无论采用何种方法，数据集混合都能提供持续强劲的迁移效果，同时保持接近专门化分布内训练的性能。额外分析进一步证实，LoRA优于全参数微调，表明LoRA能更好地保留预训练的智能体行为。
英文摘要
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.
CommentsAccepted to the REALM Workshop at EMNLP 2026
