arXiv · cs.AI · Phuc Minh Nguyen, Ngoc-Hieu Nguyen, Parshin Shojaee, Chinh Duc La, Duy Minh Ho Nguyen, Rui Zhang, Khoa D. Doan, Ting Hua, Nitesh V. Chawl · 5 小时前模型评分40
研究揭示 RLVR 训练中的首因偏差：混合数据训练更擅长简单题
On the Primacy Bias in RLVR Training
导读
一项 arXiv 研究通过受控实验和大规模分析发现，可验证奖励强化学习（RLVR，Reinforcement Learning with Verifiable Reward）在混合数据上训练会进一步拉高模型在简单题上的表现，而难题仅带来边际提升。
分类置信度
中 · 待人工复核
- 只有单一媒体来源，需要人工复核。
来源级别：媒体与个人
事件类型：研究发布
证据摘录
- “In this paper, we study this question through the lens of primacy bias: the tendency of models to over-rely on already-learned representations, disproportionately improving performance on problems they already handle well while underexploring harder ones that require departing from those representations.”
来源：arXiv · cs.AI · arxiv.org