Learning from Saturated Data: Signals Beyond Correctness for LLM Training 文章

ArXiv CS.CL2026-06-02NEWSen作者: Hanno Hiss, Jasper Dekoninck, Martin Vechev
查看原文 →
关系图谱

Learning from Saturated Data: Signals Beyond Correctness for LLM Training · 相关技术

相关技术

SFTtoken-level entropypairwise LLM self-judgmentslarge language models