AI RESEARCH
LightReasoner: Can Small Language Models Teach Large Language Models Reasoning?
arXiv CS.CL
•
ArXi:2510.07962v2 Announce Type: replace Large language models (LLMs) have nstrated remarkable progress in reasoning, often through supervised fine-tuning (SFT). However, SFT is resource-intensive, relying on large curated datasets, rejection-sampled nstrations, and uniform optimization across all tokens, even though only a fraction carry meaningful learning value.