On-Policy Distillation with Curriculum Turn-level Guidance for Multi-turn Agents
Enables efficient deployment of capable multi-turn agents by distilling large models into smaller ones, addressing a key bottleneck in practical agent systems.
On-policy distillation for multi-turn agents suffers from compounding student errors that push trajectories out of the teacher's expertise. Guided-OPD mixes teacher and student turns with a decaying curriculum, improving Score by 21.1% and Success Rate by 25.5% over vanilla OPD on average across ALFWorld, ScienceWorld, and WebShop.
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that decays to zero.Strong guidance keeps early trajectories close to the teacher distribution and is then gradually withdrawn to recover the purely on-policy regime used at inference.On ALFWorld, ScienceWorld, and WebShop, distilling Qwen3 students from a Qwen3-30B-A3B teacher,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.