FoodMonitor - Person Violations: leaderboard

FoodMonitor Person Violations scores how well multimodal models detect worker violations, such as missing masks or hair nets, in 477 kitchen videos and localize the worker with box anchors over time (FoodMonitor, 2026). The score is detection F1 on a 0 to 100 scale, higher is better.

Metric: Person-violation F1 (0-100; the paper's 0-1 score times 100), on the 477 one-minute FoodMonitor kitchen surveillance clips (3,307 annotated violations of a 27-item food safety rule document), frames at 1 fps plus the full rule document, structured JSON output; a prediction counts only if its rule category matches and a text comparison model accepts its description as the same violation, one-to-one; person violations must first localize the worker with at least ten box-and-time anchors (mean IoU 0.3, coverage 0.6); all models in thinking mode; higher is better. Source: arxiv.org. Saturation forecast: Around October 2027. 11 models tracked. Updated 7 Oct 2026.

Top models

#ModelSettingScoreOverall rank
1Seed 2.0 ProThinking26.1#222
2Gemini 3 Flash25.6#175
3Gemini 3 Pro20.3#145
4GLM-4.6V17.7#710
5Doubao Seed 2.0 LiteThinking17.5#204
6Kimi K2.516.3#266
7Qwen 3.5 397B A17B14#264
8Qwen 3 VL 235B A22B (Thinking)8.4#459
9MiMo-VL-7B-RL8.4#1155
10Qwen 3 VL 8B (Thinking)5.6#910
11GLM-4.6V-Flash2.8#1040

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

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