返回探索

Jev 模型智能性获认可,但公开评估并非可靠指标

作者表示人们认为 Jev 模型智能是好事,但公开评估会奖励不良行为者,因此已预先承诺降低其重要性。

experts has warned me not to rant about this, but it's before work hours 💅: I love that people are finding our model intelligent (if jev was dumb it would be useless) but public evals are not a good sign of that (they reward bad actors) I've precommitted to de-emphasizing

Jev 模型智能性获认可,但公开评估并非可靠指标 1
· 522 次赞在 X 打开