PAWBench (World Modeling) - Coverage: leaderboard

Metric: Coverage (%; PAWBench measures whether a system reproduces the reference distribution over physically possible outcomes across 50 rollouts on 25 scenarios per physical track with a generated video; valid-support recovery: share of reference-supported outcomes the model produces (%, higher is better; averaged over passing scenes)). Source: arxiv.org. Saturation forecast: Rough model projection: around 2026. 11 models tracked.

Top models

#ModelScore
1LTX-2.371.7
2Wan2.263.4
3LingBot-Video-MoE58.8
4LTX-2.557.4
5Cosmos 3 Super I2V55.2
6Kling 3 Std.52.8
7Seedance 250.9
8Wan2.750
9MiniMax H348.7
10HappyHorse47.1

Interactive version: theaggregate.ai/benchmark?slug=pawbench-world-modeling-coverage · How It Works · Data refreshed daily, snapshot 2026-09-26.