评分 6 · 来源:cs.AI updates on arXiv.org · 发布于 2026-04-14
评分依据:PRISM uses gradient concentration to diagnose SFT vs RL data allocation needs. Addresses underexplored question of how to split data between SFT and RL stages.
Consolidation or Adaptation? PRISM: Disentangling SFT and RL Data via Gradient Concentration
发布
采集
学术前沿 6.0 分
— PRISM uses gradient concentration to diagnose SFT vs RL data allocation needs. Addresses underexplored question of how to split data between SFT and RL stages.