BehaviorBench - Masked Survey Response: leaderboard
Metric: Accuracy (%) of predicting one masked five-point answer from the other 49 answers, on the Big Five personality survey responses (50 five-point items) of held-out subjects; chance 20; higher is better. Source: arxiv.org. Saturation forecast: Around 2034. 18 models tracked.
Top models
| # | Model | Score |
|---|---|---|
| 1 | Gemini 3.1 Pro (Preview) | 48.5 |
| 2 | GPT-5.4 (High) | 47.3 |
| 3 | Claude Sonnet 4.6 | 46.7 |
| 4 | GPT-4.1 | 46.4 |
| 5 | Gemini 3.1 Flash Lite (Preview) | 46.3 |
| 6 | Claude Opus 4.6 | 45.1 |
| 7 | Claude Haiku 4.5 | 44.7 |
| 8 | GPT-5.4 Mini (High) | 44.3 |
| 9 | Llama 3.3 70B Instruct | 43.9 |
| 10 | DeepSeek V3.2 | 40.6 |
| 11 | Qwen 3 4B | 37.2 |
Interactive version: theaggregate.ai/benchmark?slug=behaviorbench-masked-survey-response · How It Works · Data refreshed daily, snapshot 2026-09-29.