Video-HOCA - Physical Reasoning (Causal): leaderboard
Metric: Rubric score (0-100) by a GPT-OSS-120B judge of the structured open-ended answer (scene, anomaly, anomaly type, physical explanation) against the reference explanation on Video-HOCA's causal anomalies (relation-centred: valid entities interact in a way that violates a physical regularity); 16 uniformly sampled frames, deterministic Instruct-mode decoding (temperature 0, 1,024 tokens); higher is better. Source: arxiv.org. Saturation forecast: Around January 2027. 20 models tracked.
Top models
| # | Model | Score | Overall rank |
|---|---|---|---|
| 1 | Gemini 3 Flash | 67.6 | #93 |
| 2 | Gemini 2.5 Flash | 65.6 | #237 |
| 3 | Qwen 3.5 27B (Non-reasoning) | 65.1 | #183 (Qwen 3.5 27B) |
| 4 | Qwen 3 VL 30B A3B Instruct | 64.4 | #365 |
| 5 | Qwen 3 VL 8B Instruct | 63.7 | #401 |
| 6 | GPT-4o | 63 | #333 |
| 7 | Qwen 3 VL 32B Instruct | 62.9 | #276 |
| 8 | Qwen 3.5 35B A3B (Non-reasoning) | 61.1 | #250 (Qwen 3.5 35B A3B) |
| 9 | Qwen 3.5 9B (Non-reasoning) | 58.3 | #363 (Qwen 3.5 9B) |
| 10 | Qwen 2.5 VL 7B Instruct | 54.3 | #643 |
No result here: #3 Claude Opus 5.5, #5 GPT-6 Astra, #8 Claude Fable 5.1.
Interactive version: theaggregate.ai/benchmark?slug=video-hoca-physical-reasoning-causal · How It Works · Data refreshed daily, snapshot 2026-10-11.