Share your thoughts, 1 month free Claude Pro on usSee more
WorkDL logo mark

When Should the Teacher Move? Temporal Coupling and Stability in Self On-Policy Distillation

About

Self on-policy distillation trains a student policy against a teacher derived from its own parameter history, yet the teacher's update schedule -- which governs the \emph{temporal coupling} between teacher and student -- has not been systematically studied as a stability variable. Through a controlled schedule sweep on Qwen3-8B, we establish that \emph{isolation periods}, defined as complete teacher freezing between updates, are the key structural property enabling stable learning, not teacher age. To characterize these underlying training dynamics, we introduce a diagnostic framework of temporal KL structure, refresh shock, and length-tail risk. This framework further uncovers \emph{state-oblivious collapse}: optimal short-horizon fixed schedules catastrophically fail under long-horizon training because a clock-driven refresh can copy a transiently drifting student into the teacher in a single, irreversible step. This failure mode is invisible under short-horizon evaluation and mechanistically distinct from EMA's chronic contamination. To address this, we propose \emph{Consolidation-Gated Teacher Refresh} (CGTR), which preserves isolation periods while gating each refresh on joint evidence of reward improvement and length-tail safety, ensuring every teacher movement responds to genuine student consolidation rather than a clock signal. With a single shared parameter set and no per-dataset retuning, CGTR achieves \textbf{zero collapse} and the best final score on all four tasks (Chemistry, Biology, Physics, ToolUse), self-regulating its refresh frequency to each task's learning dynamics.

Haowei Guo, Baolong Bi, Ruicheng Zhang, Bingqian Sun, Wentao Zhang• 2026

Related benchmarks

TaskDatasetResultRank
Cross-task generalizationChemistry
Test Final Score0.806
3
Cross-task generalizationBiology
Test Fin Score48.5
3
Cross-task generalizationPhysics
Test Final Score70.3
3
Cross-task generalizationToolUse
Test Final Score53.6
3
Multiple-Choice ReasoningSciKnowEval Chemistry L3 (test)
Final Test Mean Score@1680.6
3
Showing 5 of 5 rows

Other info

Follow for update