← back
arXivSimran Kaur, Narutatsu Ri, Yinghui He, Liam Fowl, Sanjeev AroraMon, Jul 6, 2026, 8:01 AM PDT
score 17.0

Self-distillation harms AI models that think step-by-step

Original: Rethinking On-Policy Self-Distillation for Thinking Models

Source: arxiv.org

Writing ELI5 summary…

Self-distillation harms AI models that think step-by-step · TinyNews · TinyNews