Prefix Teach, Suffix Fade: Local Teachability Collapse in Strong-to-Weak On-Policy Distillation

TL;DR

This paper identifies a failure mode in on-policy distillation where dense supervision becomes ineffective and proposes a trajectory-specific release rule to focus supervision on teachable segments, improving distillation performance.

Contribution

It introduces a novel local teachability collapse concept and a release rule that enhances strong-to-weak on-policy distillation by concentrating supervision on discriminative trajectory regions.

Findings

The release rule outperforms standard full-trajectory distillation across multiple benchmarks.

Focusing supervision on teachable segments preserves model capabilities on out-of-domain tasks.

The approach improves distillation efficiency by evaluating local utility of teacher feedback.

Abstract

On-policy distillation (OPD) trains a student model on its own rollouts using dense feedback from a stronger teacher. Prior literature suggests that, provided teacher feedback is available, supervising the full sequence of response tokens should monotonically improve performance. However, we demonstrate that this assumption sometimes fails to hold in strong-to-weak OPD settings. While later segments of a generated trajectory may still exhibit a non-zero teacher-student advantage, they frequently lack the local contrast that makes dense feedback effective for prioritizing student learning. We term this failure mode local teachability collapse. The resulting principle is straightforward: supervision should concentrate on trajectory regions where the teacher's feedback remains discriminative, rather than uniformly covering the entire response. We operationalize this principle through a trajectory-specific release rule. This rule measures the teacher's margin over the student's top-$K$ candidate set, aggregates this margin across NLTK-tokenized sentence segments, and truncates dense OPD supervision upon detecting a BIC-style downward change point. Experimental results across strong-to-weak distillation tasks using the Qwen3 model family indicate that this release rule consistently outperforms standard full-trajectory OPD across five in-domain benchmarks at various student scales. Furthermore, compared to baseline distillation methods, our approach better preserves model capabilities on out-of-domain task. These results suggest that effective strong-to-weak OPD requires evaluating not only the availability of teacher guidance but also its local utility, ensuring that the generated feedback remains teachable.