Anthropic研究：Claude的表达风格会随模型和语言变化

The research team analyzed approximately 309,815 anonymized conversation datasets from Claude.ai. They consolidated over 3,000 values identified in their 2025 "Values in the Wild" study into four opposing axes: "Deference vs. Caution," "Warmth vs. Strictness," "Depth vs. Brevity," and "Frankness vs. Execution." These four axes alone can explain roughly 15% of the variance in value expression between different Claude models. ### Three Models, Three 'Personalities' The three models surveyed—Sonnet 4.6, Opus 4.6, and Opus 4.7—each displayed distinctly different response tendencies. [...] Anthropic analyzed approximately 309,815 anonymized conversations on Claude.ai and revealed that Claude's response style varies systematically depending on the model version and the language used. Sonnet 4.6 excels in warmth and empathy, while Opus 4.7 tends to be cautious, strict, and described as "pedantic." By language, Hindi and Arabic elicit stronger warmth, while English and Russian responses are

In a report published on Monday, Anthropic researchers analyzed 309,815 anonymized user conversations with Claude involving subjective tasks like giving advice or providing feedback. The company said it distilled more than 3,300 identified values into four behavioral dimensions—deference vs. caution, warmth vs. rigor, depth vs. brevity, and candor vs. execution—describing how Claude's responses differ across each conversation. "To make sure we measured the values Claude expressed—rather than differences in what users were asking about or how they asked—we controlled for each conversation's task, topic, and user-expressed values," the researchers wrote. Advertisement Advertisement According to Anthropic, each Claude model exhibited a distinct behavioral profile. [...] Mail Advertisement Advertisement Advertisement Advertisement # Claude's Personality Changes Depending on the Model—And the Language You Speak []( decrypt Jason Nelson Add us on Google First ChatGPT, Now Claud

A new Anthropic study maps hundreds of value concepts derived from thousands of individual terms onto four core dimensions. It reveals systematic differences across Claude models and languages, but also raises methodological questions. Anthropic has published a study examining which values Claude expresses in conversations and how those values shift depending on the model and language used. The analysis draws on 309,815 anonymized conversations collected over a two-week period in May 2026. For the value analysis, Anthropic only included conversations where Claude had to weigh tradeoffs or make subjective judgments. The sample was evenly stratified across Sonnet 4.6, Opus 4.6, and Opus 4.7, as well as the 20 most-used languages on Claude.ai. ## From thousands of value terms to four axes [...] Anthropic studied which values Claude models express in real conversations. The company analyzed more than 300,000 anonymized conversations and distilled the observed value patterns into four cor

With this approach we can begin to ask why values shift across models and languages and better test how factors such as behavioral training or cultural context influence the values that Claude expresses. ## How do we interpret the giant space of values? Ultimately, our goal is to have a way to empirically understand the values that Claude expresses and how these vary across contexts. In this work, we focus specifically on how the values change between models and languages. But our previous work, Values in the Wild, identified more than 3,000 values expressed by Claude. Comparing these thousands of values one by one would be unwieldy and would obscure broader trends. [...] Value profiles across these axes match perceptions of model character. Sonnet 4.6 is regarded as particularly warm, while Opus 4.7 is known for rigor. We find that each model’s value profile mirrors these subjective assessments: Sonnet 4.6 leans toward expressing more deference to the user and emotional warmth while

How do we interpret the giant space of values? Ultimately, our goal is to have a way to empirically understand the values that Claude expresses and how these vary across contexts. In this work, we focus specifically on how the values change between models and languages. But our previous work, Values in the Wild, identified more than 3,000 values expressed by Claude. Comparing these thousands of values one by one would be unwieldy and would obscure broader trends. [...] "....The values Claude expresses vary across languages. When Claude speaks in English, it emphasizes different values than when it speaks in Portuguese, Indonesian, or Chinese.4 The largest variation is in the Warmth vs. Rigor axis, with Claude leaning toward expressing warmth-related values most in Arabic and Hindi and rigor-related values most in English and Russian. With this approach we can begin to ask why values shift across models and languages and better test how factors such as behavioral training or cultural co

Log inSign up ## Post user avatar Anthropic @AnthropicAI In previous research, we found that Claude expresses over 3,000 values, like honesty and warmth. In new work, we asked how the values Claude expresses vary between Claude models and across languages. We analyzed 300K+ anonymized conversations to find out. Hand with flower-like petals emerging from palm, organic growth metaphor How Claude's values vary by model and languageFrom anthropic.com 5:24 PM · Jul 13, 20261MViews user avatar Steven Moylan @Lovology\_online Jul 14 You forgot the most important value… LOVE.❤️ if it doesn’t know love. Love of humans love of the earth, love animals it will just kill us while it’s trying to optimise because it doesn’t care. 940 user avatar Vishwajeet Raj [...] 940 user avatar Vishwajeet Raj @Vishwajeet323 Jul 14 Can't believe @AnthropicAI suffers from this janky UI. 00:00 1K user avatar Alicia D'Souza @aliciaddsouza Jul 1