CausalPhys - Goal Orientation: 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 Goal-Orientation domain, choosing actions that achieve a desired outcome; higher is better.
Metric: Answer accuracy (0-1 scaled to %) on the Goal-Orientation domain (counterfactual choice of actions that achieve a desired outcome) 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 | GPT-4o | 58.78 | #741 |
| 2 | InternVL3-78B | 57.99 | #738 |
| 3 | Gemini 2.5 Flash | 54.39 | #453 |
| 4 | GPT-4o Mini | 51.57 | #1623 |
| 5 | Qwen 2 VL 7B | 49.06 | #2261 |
| 6 | Mistral Small 3.2 | 47.96 | #1591 |
| 7 | Claude Sonnet 4 | 47.02 | #400 |
| 8 | Phi-4 Multimodal Instruct | 44.83 | #3882 |
No result here: #1 Claude Fable 5.2, #2 Gemini 4 Carbon, #5 Claude Opus 5.5.
Interactive version: theaggregate.ai/benchmark?slug=causalphys-goal-orientation · How It Works · Data refreshed daily, snapshot 2026-11-13.