Event ·Multi-source

Anthropic研究：Claude价值观因模型和语言意外变化

Anthropic分析了309,815条用户对话，发现Claude的行为因模型版本和语言而异。Sonnet 4.6更温暖简洁，Opus 4.7更坦率谨慎，Opus 4.6直截了当。使用印地语和阿拉伯语的用户得到最温暖的Claude，英语和俄语用户得到最严谨的回应，荷兰语用户得到最易承认错误的Claude。Anthropic表示尚未理解这些变化的原因，也不确定这些差异是否期望的结果。

Current read

Anthropic发现Claude说着不同语言就像换了个人，连他们自己都搞不清原因，蛮有意思的研究。

11 sources71° AI heatLast updated 2026/7/13 19:39:48

Evidence

11 sources

Primary sourceThe Rundown AI

2026/7/13 19:39:48View source

Related source 1arXiv: Anthropic

2026/7/12 13:00:41View source

Related source 2小互

2026/7/14 06:25:54View source

Related source 3Anthropic

2026/7/13 17:24:56View source

Related source 4IT之家

2026/7/14 01:44:58View source

Related source 5Decoder

2026/7/14 11:00:12View source

Related source 6@koltregaskes

2026/7/14 19:04:39View source

Related source 7Simon Willison

2026/7/12 15:47:15View source

Related source 8Greg Brockman

2026/7/13 00:00:54View source

Related source 9techcrunch

2026/7/13 15:34:59View source

Related source 10AI 热点 · 原创

2026/7/13 16:17:43View source

Conflict check

Dedup data only shows that these sources discuss the same event, not that they agree. There is no structured supporting or refuting evidence yet, so no conflict is inferred automatically.

Open the story