RefereeBench - Foul and Penalty Classification: leaderboard

RefereeBench (2026) tests multimodal models as sports referees on 925 officiating videos across 11 sports with 6,475 human-annotated multiple-choice questions. This board is accuracy in percent on questions about which foul and which penalty apply, averaged across sports; higher is better.

Metric: Accuracy (%) on foul and penalty classification questions (mean of Q2 and Q4), averaged across 11 sports; higher is better. Source: arxiv.org. Saturation forecast: Around February 2027. 11 models tracked. Updated 7 Oct 2026.

Top models

#ModelScoreOverall rank
1Gemini 3 Flash58.8#175
2Gemini 3 Pro57.9#145
3GPT-556.6#178
4Seed 1.853.4#252
5Qwen 3 VL 8B49#917
6GPT-4o47.1#751
7Claude Sonnet 4.545.4#261
8InternVL3.5-8B41.6#1604
9Claude Haiku 4.535.4#583
10VideoLLaMA3-7B34.8#2050
11LLaVA-Video-7B25.1#2276

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

Interactive version: theaggregate.ai/benchmark?slug=refereebench-foul-and-penalty-classification · How It Works · Data refreshed daily, snapshot 2026-11-21.