CausalPhys - Description Correctness: leaderboard
Metric: Description Correctness (0-1 scaled to %), the share of attribute and event descriptions the rationale states consistently with the annotation, averaged over the four domains; the model writes a rationale and an answer; a GPT-4o judge makes binary checks of the rationale against the expert-annotated causal graph (typed object, attribute and event nodes with directed dependencies); higher is better. Source: arxiv.org. Saturation forecast: Around 2030. 11 models tracked.
Top models
| # | Model | Score |
|---|---|---|
| 1 | InternVL3-78B | 47.2 |
| 2 | GPT-4o | 43.97 |
| 3 | Gemini 2.5 Flash | 43.08 |
| 4 | Qwen 3 VL 32B | 40.4 |
| 5 | Claude Sonnet 4 | 40.37 |
| 6 | GPT-4o Mini | 39.94 |
| 7 | Mistral Small 3.2 | 36.69 |
| 8 | Phi-4 Multimodal Instruct | 35.01 |
| 9 | Qwen 2 VL 7B | 33.08 |
Interactive version: theaggregate.ai/benchmark?slug=causalphys-description-correctness · How It Works · Data refreshed daily, snapshot 2026-09-29.