ContrAR (Single Frame) - Safety Inspection: leaderboard
ContrAR (2026) tests whether vision-language models detect contradictory virtual content attacks in augmented reality, answering yes or no on real AR videos of which half carry an overlay that contradicts the physical scene, given the middle frame of each video. Detection accuracy in percent on the safety inspection videos; higher is better.
Metric: Detection accuracy (%) for contradictory virtual content attacks in AR videos (42 safety inspection videos; half the videos carry a contradiction, a balanced yes/no decision), one middle frame of each video given (single-frame prompting); mean of three runs; higher is better. Source: arxiv.org. Saturation forecast: Around February 2027. 11 models tracked. Updated 7 Oct 2026.
Top models
| # | Model | Score | Overall rank |
|---|---|---|---|
| 1 | GPT-4.1 | 90.48 | #505 |
| 2 | Gemini 2.5 Pro | 83.33 | #276 |
| 3 | GPT-5 | 80.95 | #178 |
| 4 | GPT-4o | 71.43 | #751 |
| 5 | Claude Sonnet 4.5 | 71.43 | #261 |
| 6 | Gemini 2.5 Flash | 69.05 | #463 |
| 7 | Claude Haiku 4.5 | 64.29 | #583 |
| 8 | Grok 4 | 64.29 | #319 |
| 9 | Qwen 2.5 VL 7B Instruct | 61.9 | #1902 |
| 10 | Grok 2 Vision | 61.9 | #1437 |
| 11 | Qwen 2.5 VL 72B Instruct | 50 | #841 |
No result here: #1 Claude Fable 5.2, #3 Gemini 4 Neon, #4 GPT-6.2 Astra.
Interactive version: theaggregate.ai/benchmark?slug=contrar-single-frame-safety-inspection · How It Works · Data refreshed daily, snapshot 2026-11-21.