COHERENCE - WikiHow: leaderboard

COHERENCE (2026) removes the images from 6,161 interleaved image-text documents (WikiHow, StoryBird, cooking recipes, science) and asks the model to put each candidate image back in its place. This board is the exact-match accuracy on the WikiHow procedural documents, in percent; higher is better.

Metric: Exact-match accuracy (%) on the 2,076 WikiHow procedural documents; 6,161 interleaved image-text documents whose images were removed: the model assigns each candidate image to its placeholder; higher is better. Source: arxiv.org. Saturation forecast: Around February 2027. 19 models tracked. Updated 7 Oct 2026.

Top models

#ModelSettingScoreOverall rank
1Gemini 3.1 Pro (Preview)80.73#99
2GPT-5.4High77.5#135
3Doubao-Seed-2.0-Pro-26021576.59#158
4Kimi K2.575.67#266
5Claude Sonnet 4.6Thinking74.18#173
6Doubao-Seed-2.0-Lite-26021572.74#262
7Qwen 3.5 122B A10B71.87#346
8Doubao-Seed-2.0-Mini-26021570.81
9Qwen 3.5 397B A17B69.85#264
10Qwen 3.5 35B A3B69.7#492
11GPT-5.268.5#185
12Qwen 3 VL 235B A22B64.45#485
13Intern-S1-Pro64.26#585
14GLM-4.6V62.96#710
15Qwen 3.5 4B62.76#1241

No result here: #1 Claude Fable 5.2, #3 Gemini 4 Neon, #4 GPT-6.2 Astra.

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