Google DeepMind 在不共享提示或模型权重的情况下试行双盲 AI 评估
Google DeepMind 表示，其试点项目将外部评估提示和模型权重保密，尽管其首创性的主张尚未得到独立验证。
By RuntimeWire Staff · Published
Primary source: Google DeepMind
为何重要
封闭模型审计可能迫使评估者暴露私人测试，或迫使开发者共享有价值的模型权重。一个可信的双盲流程可以同时保护双方，但 Google DeepMind 未就其试点披露足够信息，无法评估其安全性、结果或评估者的独立性。
Google DeepMind, the AI lab co-founded by Demis Hassabis, said on August 27th that it is piloting a double-blind evaluation system intended to let outsiders test frontier models without exposing either side's sensitive material.
When you purchase through links in our articles, we may earn a small commission. This doesn’t affect our editorial independence.