BehaviorBench - Survey Response from Demographics (Distribution): leaderboard
BehaviorBench (University of Michigan, 2026) evaluates foundation models on behavioral science tasks at the individual and population levels, using economic-game records, Big Five personality surveys and research literature. This board is the Wasserstein distance (0 to 4) between predicted and observed answer distributions per survey item when predicting from demographics; lower is better.
Metric: Wasserstein distance between predicted and observed answer distributions per item (five-point scale, so 0 to 4), averaged over the 50 items, predicting from demographics on the Big Five personality survey responses (50 five-point items) of held-out subjects; lower is better. Source: arxiv.org. Saturation forecast: Around February 2027. 18 models tracked.
Top models
| # | Model | Score | Overall rank |
|---|---|---|---|
| 1 | Gemini 3.1 Pro (Preview) | 0.75 | #94 |
| 2 | DeepSeek V3.2 | 0.76 | #406 |
| 3 | GPT-4.1 | 0.79 | #495 |
| 4 | Gemini 3.1 Flash Lite (Preview) | 0.82 | #418 |
| 5 | Claude Sonnet 4.6 | 0.86 | #161 |
| 6 | Claude Opus 4.6 | 0.86 | #102 |
| 7 | GPT-5.4 (High) | 0.9 | #123 (GPT-5.4) |
| 8 | GPT-5.4 Mini (High) | 0.91 | #389 (GPT-5.4 Mini) |
| 9 | Claude Haiku 4.5 | 0.93 | #568 |
| 10 | Llama 3.3 70B Instruct | 1.02 | #1357 |
| 11 | Qwen 3 4B | 1.14 | #2281 |
No result here: #1 Claude Fable 5.2, #2 Gemini 4 Carbon, #5 Claude Opus 5.5.
Interactive version: theaggregate.ai/benchmark?slug=behaviorbench-survey-response-from-demographics-distribution · How It Works · Data refreshed daily, snapshot 2026-11-13.