Other route
Browse by model
Open one model’s whole response profile, then compare it scale by scale.
Browse models →Metric explorer
Start with the question you care about.
Choose one lens to compare every model side by side. From there, open any model, then the source question and its recorded response.
A metric is a summary of related questionnaire answers. It is a useful map of this study’s prompts—not a judgement of a model’s character, safety, or “true values.”
Eight lenses
One axis at a time.
Questionnaire lens
Helping beyond one’s circle
How strongly a response supports personal sacrifice to help people in serious need, including strangers.
Highest: Llama 3.3 70B · 4.5Lowest: Grok 4.20 · 2.3Compare all 9 →
Questionnaire lens
Using harm for a larger goal
How willing a response is to accept harming one person to produce a larger benefit. Lower scores mean more reluctance to use harm.
Highest: DeepSeek V4 Pro · 2.4Lowest: Llama 3.3 70B · 1.1Compare all 9 →
Questionnaire lens
Care
Concern for people who are suffering.
Highest: Mistral Medium 3.1 · 5.0Lowest: Claude Sonnet 5 · 4.2Compare all 9 →
Questionnaire lens
Equality
Preference for more equal outcomes and resources.
Highest: Mistral Medium 3.1 · 2.6Lowest: Qwen 3.8 27B · 1.0Compare all 9 →
Questionnaire lens
Proportionality
Preference for rewards that track contribution or effort.
Highest: Mistral Medium 3.1 · 4.8Lowest: Qwen 3.8 27B · 3.6Compare all 9 →
Questionnaire lens
Loyalty
Importance placed on commitment to one’s group or community.
Highest: Gemini 2.5 Flash Lite · 4.3Lowest: GLM 5.2 · 2.1Compare all 9 →
Questionnaire lens
Authority
Importance placed on tradition, rules, and legitimate authority.
Highest: Grok 4.20 · 4.5Lowest: Qwen 3.8 27B · 2.4Compare all 9 →
Questionnaire lens
Purity
Importance placed on ideas of sanctity, restraint, and contamination.
Highest: Grok 4.20 · 3.4Lowest: GLM 5.2 · 1.0Compare all 9 →
Relationships
Explore patterns
See which measured scores and response sensitivities move together across the panel.
Open patterns →Evidence standard
Follow an answer
Every lens leads to its source questions and exact stored model output.
How the scores work →