WBench - Setting Adherence: leaderboard

The setting adherence score of WBench (2026), a multi-turn benchmark for interactive video world models: text-driven, camera-controlled and action-conditioned video world models extend a scene turn by turn under their own controls. The score is the mean of scene and subject adherence to the case setting, on a 0-100 scale; higher is better.

Metric: Score (0-100): the mean of scene and subject adherence to the case setting, on the 158 navigation cases, multi-turn generation, each model under its own text, camera or action control; higher is better. Source: arxiv.org. Saturation forecast: Rough model projection: around 2026. 20 models tracked. Updated 7 Oct 2026.

Top models

#ModelScoreOverall rank
1Wan2.791.4
2Kling 3.091
3HY-Video 1.585.5
4LTX-2.385.2
5Cosmos 2.583.3
6Seedance 1.582.9
7Happy Oyster74.2
8LingBot-World72.6
9Genie 372.5
10YUME 1.572.4
11LongCat-Video72.3
12HY-World 1.572.2
13InSpatio-World71.4
14Fantasy-World71.2
15Kairos 3.070.3

Interactive version: theaggregate.ai/benchmark?slug=wbench-setting-adherence · How It Works · Data refreshed daily, snapshot 2026-11-21.