TASTE:AI 模型能否评判 AI 安全研究提案?
Anthropic 构建了 TASTE(The AI Safety Taste Evaluation)基准,用 92 组配对比较衡量模型评判 AI 安全研究提案的能力,以与资深人类研究者偏好的一致率为指标,估计人类研究者一致率为 77%。模型表现不及人类研究者,Fable 5 仅得 60%。该基准通过研究者讨论分歧并修订评分、再筛选自报“strong”置信度的标签来提升一致性,使估计人类一致率从讨论前的 53% 提升 15 个百分点至 68%。
ard-to-verify parts of safety research. One important hard-to-verify aspect of the research process is evaluating research proposals. Judging research proposals well is a high-leverage way to improve research quality, since some proposals are often much more important or tractable than others, and picking a poor initial direction can waste substantial time or resources. In order to evaluate models on the hardest-to-verify tasks, we have to use human judgment as ground truth. However, humans ofte
匹配字段正文
