Skip to content

Teacher Should Think Ahead: Adaptive Continuations for Reliable On-Policy Distillation

Sep 2026 · 3 citations · 36 references
Computer Science

TL;DR

This work identifies Teacher Uncertainty Contraction (TUC), a systematic phenomenon whereby the teacher's predictive uncertainty decreases as it continues from a student-generated prefix, and proposes Adaptive-Continuations On-Policy Distillation (AC-OPD), which augments informative states along student rollouts with teacher continuations and adaptively selects their effective supervision horizons.

Abstract

On-policy distillation (OPD) is a promising approach for transferring knowledge between language models, where a student receives dense token-level supervision along its own generated trajectories. However, teacher supervision can be unreliable when conditioned on incomplete or low-quality student prefixes. We identify Teacher Uncertainty Contraction (TUC), a systematic phenomenon whereby the teacher's predictive uncertainty decreases as it continues from a student-generated prefix. We theoretically characterize this trade-off through a variance-bias decomposition of teacher-branch gradients, showing that uncertainty contraction reduces variance while teacher-student path divergence increases bias, thereby favoring a finite continuation. Guided by this insight, we propose Adaptive-Continuations On-Policy Distillation (AC-OPD), which augments informative states along student rollouts with teacher continuations and adaptively selects their effective supervision horizons. Experiments on mathematical reasoning and code generation across model scales demonstrate that AC-OPD consistently improves over standard OPD. Controlled-continuations and matched-budget analyses further validate the adaptive-continuations design, highlighting adaptive teacher continuations as an effective principle for reliable on-policy distillation.The code will be made publicly available upon publication.

View source

Similar papers

Reward-Aligned Reweighting for On-Policy Distillation

Reward-Aligned Reweighting for On-Policy Distillation (R$^{2}$-OPD), which uses outcome agreement and the magnitude of teacher--student disagreement to continuously reallocate teacher supervision, achieves the highest average accuracy among the compared training methods in both cross-size and same-size distillation.

Hao Xu, Junwei Su, Lan-Song Diao et al. · 0 citations
#artificial intelligence Preprint Oct 2026

Stochastic Teacher Intervention for Agentic On-Policy Distillation

STI-OPD, a stochastic teacher intervention framework for multi-turn agentic OPD, and an Importance-Weighted Reverse KL objective that corrects the token sampling mismatch between teacher-generated responses and the student policy to preserve the original OPD objective are introduced.

Jun-Nan Liu, Lin-Hao Luo, Zhijun Chen et al. · 0 citations

On the Off-Policy Teacher in On-Policy Distillation

Control experiments show that SCOUT improves the teacher's ability to continue from student-generated prefixes, supporting the intended mechanism of student-conditioned teacher adaptation, and consistently improves the effectiveness of on-policy distillation.

Lang-Lin Huang, Hao Liu, Mononito Goswami et al. · 1 citation

Learning from a Thoughtful Teacher: Adaptive On-Policy Self-Distillation for Mathematical Reasoning

Adaptive On-Policy Self-Distillation is proposed, which adapts what information the teacher receives and how strongly its feedback influences learning, and encodes each solution as a reasoning DAG, orders problems by the student's evolving capability, and reveals only the affordable subgraph and its next frontier as PI...

Jia-Cheng Du, Wei-Wei Xie, Tian-Yi Du et al. · 0 citations
#natural language process... Preprint Sep 2026

Dense Is Not Enough: Hierarchical Supervision Allocation for Long-Horizon On-Policy Distillation

This work forms long-horizon OPD as hierarchical supervision allocation as well as a coarse-to-fine framework that organizes supervision through Locate, Validate, and Refine and suggests that effective long-horizon distillation requires teaching at the right depth, the right decision, and the right token.

Yu-Hao Sun, Bin-Rui Wu, Zhuo-Er Xu et al. · 0 citations
#machine learning Preprint Oct 2026

DiffGate: Difficulty-Gated Teacher Guidance for On-Policy Distillation

DiffGate is introduced, an outcome-gated objective that combines GRPO with selective, bounded teacher guidance, and improves pass@8 across all four model--domain settings, demonstrating improved solution coverage under the evaluation protocol.

Karn Tiwari, V. Chordia, P. PrathoshA · 0 citations

Related blog posts

MIT News · Artificial Intelligence Sep 29, 2026

Who we become when we talk to machines

Professor Sherry Turkle’s new book, “Artificial Intimacy,” offers a withering critique of chatbots and the antisocial dynamics she believes they encourage.

We use cookies to run the site and, with your consent, for analytics and to show ads. See our Cookie Policy.