2025-07-03-insights

Skywork-Reward-V2: Scaling Preference Data Curation via Human-AI Synergy

skywork之前开源了一个reward model，今天又迭代到了2期。足足训了2400万的pair数据