1 Matching Annotations
  1. Last 7 days
    1. Astra is our most aligned model, with substantial improvements in understanding user intent and model behavior—you can delegate tasks with greater confidence in Astra's judgment.

      【非共识】这一关于模型对齐的声明是主观的,没有提供客观衡量标准。'对齐'是一个复杂且多方面的概念,仅凭OpenAI的自我评估不足以证明这一点。需要第三方验证和更具体的评估方法来支持这一说法。