AI 导读
如果你在发布模型之前就公布它的基准测试成绩(权重、产品、API 等"稍后"再放),那几乎可以肯定,在你正式推出之前,会有人在你这个类别里发布一个更好的模型。
正文
If you announce a model's benchmarks without releasing it (weights, product, api, etc "later"), feels like almost a sure thing that someone releases a better model in your category before you get it out.
来源:Nathan Lambert · x.com