Back to AI News
arxiv

Speculative Self-Distillation Enables Efficient Knowledge Internalization

The paper proposes Speculative Self-Distillation (SSD), a per-token mixed-policy method for teaching language models knowledge absent from pretraining or updated facts. The student generates rollouts by default, while the teacher intervenes only where its next-token distribution diverges substantially. On three closed-book QA benchmarks, SSD nearly matches on-policy distillation while using 38-69% fewer supervised tokens and outperforms off-policy distillation on held-out tests. A forgetting stress test shows SSD retains the most general capability while achieving the best in-domain accuracy with less supervision than the on-policy baseline.

AI Brief

It introduces a practical self-distillation method that reduces supervision cost while improving the accuracy-efficiency-forgetting tradeoff for knowledge internalization.

arXiv:2611.12016v1 Announce Type: cross 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.
Be the first to rate this article.

Comments & Questions (0)

Captcha
Please be respectful — let's keep the conversation friendly.

No comments yet

Be the first to comment!