EgoPolice - 10-Second Clips: leaderboard

EgoPolice (2026) zero-shot action questions on 10-second police body-worn camera clips, five options including None of the above. Accuracy in percent (chance 20%), higher is better.

Metric: Accuracy (%; 5,000 questions on 10-second clips; zero-shot five-option multiple choice (one visible police-civilian action, three actions absent from the clip and None of the above; chance 20%) on minimally edited COPA and Pasadena police body-worn camera clips; images at 1 fps (at least 8 frames) for image VLMs, author-recommended sampling for video VLMs, one shared prompt). Source: arxiv.org. Saturation forecast: Estimated already saturated. 12 models tracked.

Top models

#ModelScoreOverall rank
1Gemini 2.5 Pro77.5#266
2GPT-4.176.2#495
3Gemini 2.5 Flash69.7#453
4InternVL3-8B64.5#1688
5Qwen 2.5 VL 7B Instruct55.8#1833
6GPT-4.1 Nano18.7#1977

No result here: #1 Claude Fable 5.2, #2 Gemini 4 Carbon, #5 Claude Opus 5.5.

Interactive version: theaggregate.ai/benchmark?slug=egopolice-10-second-clips · How It Works · Data refreshed daily, snapshot 2026-11-13.