CausalPhys - Relation Awareness: leaderboard
Metric: Relation Awareness (0-1 scaled to %), the share of directed causal dependencies the rationale captures, 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 2031. 11 models tracked.
Top models
| # | Model | Score |
|---|---|---|
| 1 | InternVL3-78B | 30.52 |
| 2 | Qwen 3 VL 32B | 29.14 |
| 3 | Claude Sonnet 4 | 27.83 |
| 4 | GPT-4o | 27.6 |
| 5 | Mistral Small 3.2 | 26.41 |
| 6 | GPT-4o Mini | 24.68 |
| 7 | Gemini 2.5 Flash | 24.37 |
| 8 | Phi-4 Multimodal Instruct | 21.66 |
| 9 | Qwen 2 VL 7B | 20.02 |
Interactive version: theaggregate.ai/benchmark?slug=causalphys-relation-awareness · How It Works · Data refreshed daily, snapshot 2026-09-29.