[2605.23857] Strong Teacher Not Needed? On Distillation in LLM Pretraining

Abstract:Knowledge distillation generally assumes a strong-to-weak relationship where stronger teachers yield better students. In this work, we examine this assumption about distillation in large language model pretraining. By varying architecture sizes and training token budgets, we create strong-to-weak, same-level, and weak-to-strong teacher-student relationships, and study distillation's effectiveness under each. We find that the teacher need not be strong: with proper mixing of the language modeling and knowledge distillation losses, even small and undertrained teachers improve larger students. At the same time, a stronger teacher is not always better: pushing the teacher further, through more parameters or more training tokens, can saturate or even reverse the distillation gains. We further observe that distillation improves generalization (out-of-distribution and downstream performance) more readily than in-domain fitting. Together, these results challenge the common belief that distillation pretraining always requires a strong teacher.

# link_g9jjdwr8ex.pdf ## Metadata - PDFFormatVersion=1.7 - IsLinearized=false - IsAcroFormPresent=false - IsXFAPresent=false - IsCollectionPresent=false - IsSignaturesPresent=false - Author=Taiming Lu; Zhuang Liu - Creator=arXiv GenPDF (tex2pdf:a6404ea) - Custom.DOI=https://doi.org/10.48550/arXiv.2605.23857 - Custom.License=http://creativecommons.org/licenses/by/4.0/ - Custom.PTEX.Fullbanner=This is pdfTeX, Version 3.141592653-2.6-1.40.28 (TeX Live 2025) kpathsea version 6.4.1 - Custom.arXivID=https://arxiv.org/abs/2605.23857v1 - Producer=pikepdf 8.15.1 - Title=Strong Teacher Not Needed? On Di

saved by

• Christina Lee

related reading

• [2507.06187] The Delta Learning Hypothesis: Preference Tuning on Weak Data can Yield Strong Gainsarxiv.org
• [2605.12715] Scaling Laws for Mixture Pretraining Under Data Constraintsarxiv.org
• [2604.13010] Lightning OPD: Efficient Post-Training for Large Reasoning Models with Offline On-Policy Distillationarxiv.org
• On-Policy Distillation - Thinking Machines Labthinkingmachines.ai
• [2601.18734] Self-Distilled Reasoner: On-Policy Self-Distillation for Large Language Modelsarxiv.org
• [2604.00626] A Survey of On-Policy Distillation for Large Language Modelsarxiv.org
• [2510.13786] The Art of Scaling Reinforcement Learning Compute for LLMsarxiv.org
• [2605.06546] Efficient Pre-Training with Token Superpositionarxiv.org
• [2509.14786] Pre-training under infinite computearxiv.org
• [2305.18290] Direct Preference Optimization: Your Language Model is Secretly a Reward Modelarxiv.org
• [2604.22082] Removing Sandbagging in LLMs by Training with Weak Supervisionarxiv.org
• [2005.14165] Language Models are Few-Shot Learnersarxiv.org