精选理由
别被 Arena 分数忽悠了——Kimi K3 虽好，但评分榜的坑和 Llama 4 时一样。
Kimi K3 是一个好模型，但人们再次过度解读 Arena 排行榜的 ELO 分数，就像之前对 Llama 4 那样。Arena 用户评分的 ELO 有局限性，前端是文本聊天，容易通过训练或系统提示让用户主观偏好表现更好，偏离真实能力评估。
原文 · Ethan Mollick
Kimi K3 is a very good model, but people are overi…
Kimi K3 is a very good model, but people are overindexing on an Arena score again (remember Llama 4?)
ELO scores as judged by Arena users are limited, and front-end is like text chat, relatively easy to train/system prompt to a state that people prefer when it is subjective.