cs.LGNov 9, 2026

Speculative Self-Distillation enables Efficient Knowledge Internalization

Authors: Shayan Talaei, Agam Bhatia, Arshia Soltani Moakhar, Jonas Hübotter, Amin Saberi, Azalia Mirhoseini

Organizations: Stanford University · ETH Zürich

Abstract

Many real-world applications require language models to internalize knowledge that was absent from pretraining, to incorporate continually updated facts, or to learn new domain skills. Self-distillation has emerged as a natural approach for this task. A model conditioned on the source material acts as the teacher, and an unconditioned student is trained to reproduce the teacher's behavior closed-book. We propose Speculative Self-Distillation (SSD), an efficient per-token mixed-policy method for self-distillation. The student generates each rollout by default, while the teacher intervenes only at positions where its next-token distribution substantially diverges from the student's. Unlike off-policy distillation with teacher-generated prefixes, this design keeps training rollouts close to the student's test-time behavior, while avoiding the low-signal states that fully on-policy rollouts drift into. In effect, SSD lets the teacher steer the rollout exactly where the privileged context changes the prediction. On three closed-book QA benchmarks measuring fact recall, knowledge updates, and skill acquisition, SSD nearly matches on-policy distillation while using 38-69% fewer supervised tokens and consistently outperforms off-policy distillation on the held-out test set. On a forgetting stress test, SSD retains the most general capability while also achieving the best in-domain accuracy and using less supervision than the on-policy baseline. Overall, SSD establishes a new accuracy-efficiency-forgetting Pareto frontier for self-distillation.

Figures & tables

Appendix figures & tables11 assets

Supplementary material from the paper’s appendix.

Appendix

Explore similar work

CardsList
  1. Bridged Self-Distillation for On-Policy Learning

    Oct 19, 2026Runsheng Chen, Yihong Liu, Raoyuan Zhao +3Reasoning DistillationOn-Policy Distillation

  2. DemoPSD: Disagreement-Modulated Policy Self-Distillation

    Jul 2, 2026Yunhe Li, Hao Shi, Wenhao Liu +5On-Policy Self-DistillationRL for Language Model Reasoning

  3. On-Policy Self-Distillation without Any Supervision

    Aug 6, 2026Yijiang Li, Bingyang Wang, Yijun Liang +3On-Policy Self-DistillationLanguage Model Self-Improvement