CausalPhys - Perception: leaderboard
CausalPhys (2026) tests vision-language models on physical reasoning questions over real-world images and videos, each paired with an expert causal graph of objects, attributes and events. This board is answer accuracy in percent on the Perception domain, recognizing objects and physical attributes; higher is better.
Metric: Answer accuracy (0-1 scaled to %) on the Perception domain (association rung: recognizing objects and physical attributes) of the image- and video-based questions with verified answers over real-world physical scenes; higher is better. Source: arxiv.org. Saturation forecast: Estimated already saturated. 11 models tracked.
Top models
| # | Model | Score | Overall rank |
|---|---|---|---|
| 1 | InternVL3-78B | 62.57 | #738 |
| 2 | GPT-4o Mini | 59.83 | #1623 |
| 3 | Gemini 2.5 Flash | 59.2 | #453 |
| 4 | GPT-4o | 58.89 | #741 |
| 5 | Claude Sonnet 4 | 58.68 | #400 |
| 6 | Phi-4 Multimodal Instruct | 55.73 | #3882 |
| 7 | Qwen 2 VL 7B | 52.05 | #2261 |
| 8 | Mistral Small 3.2 | 48.26 | #1591 |
No result here: #1 Claude Fable 5.2, #2 Gemini 4 Carbon, #5 Claude Opus 5.5.
Interactive version: theaggregate.ai/benchmark?slug=causalphys-perception · How It Works · Data refreshed daily, snapshot 2026-11-13.