通过渐进式引导提升多轮智能体小模型的蒸馏效果
On-Policy Distillation with Curriculum Turn-level Guidance for Multi-turn Agents
- 在每轮中混合师生生成的对话，逐步降低教师干预概率
- 在ALFWorld等数据集上成功率提升25.5%，得分提高21.1%
- 特别适合资源受限场景下的小模型部署
多轮智能体通过规划、调用工具与环境交互，是解决复杂任务的有前景范式，但其能力通常依赖大模型，推理成本过高。在实践中，基于策略的蒸馏（OPD）可将大模型能力迁移到小模型，但存在典型缺陷：小模型错误在多轮中累积，使轨迹偏离教师熟悉的分布，导致教师监督在最需要时最不可靠。为此，本文提出引导式在策略蒸馏（Guided-OPD），在每轮采样中混合教师与学生生成的对话，并设计随训练进程衰减至零的教师干预课程。强引导确保早期轨迹贴近教师分布，随后逐步撤回以恢复推理时的纯在策略机制。在ALFWorld、ScienceWorld和WebShop上，将Qwen3-30B-A3B教师模型的Qwen3学生模型进行蒸馏，相比原始OPD，平均得分提升21.1%，成功率提升25.5%，且小模型收益更大。
原文摘要 · Abstract (English)
Multi-turn agents that plan, invoke tools, and interact with environments offer a promising paradigm for solving complex tasks, yet their capabilities typically rely on very large models whose inference cost is prohibitive in practice.On-Policy Distillation (OPD) is a natural recipe for transferring such capabilities to smaller students, but we find that it suffers a characteristic failure mode in this setting: small student errors compound across turns and push the trajectory out of the teacher's familiar state distribution, so the teacher's supervision becomes least reliable precisely where the student needs it most.We propose Guided On-Policy Distillation (Guided-OPD), a simple yet effective algorithm that mixes teacher- and student-generated turns within each rollout and schedules the teacher's intervention probability along a curriculum that decays to zero.Strong guidance keeps early trajectories close to the teacher distribution and is then gradually withdrawn to recover the purely on-policy regime used at inference.On ALFWorld, ScienceWorld, and WebShop, distilling Qwen3 students from a Qwen3-30B-A3B teacher, Guided-OPD improves Score by 21.1\% and Success Rate by 25.5\% over vanilla OPD on average, with larger gains on smaller students.
Thank you to arXiv for use of its open access interoperability. PaperDance 不是 arXiv 官方产品；中文卡片由大模型生成，请以原文为准。