ContrAR (Multiple Frames): leaderboard
Metric: Detection accuracy (%) for contradictory virtual content attacks in AR videos (all 312 videos of five AR applications; half the videos carry a contradiction, a balanced yes/no decision), first, middle and last frames given (multi-frame prompting); mean of three runs; higher is better. Source: arxiv.org. Saturation forecast: Around January 2027. 11 models tracked.
Top models
| # | Model | Score |
|---|---|---|
| 1 | GPT-4.1 | 86.54 |
| 2 | GPT-5 | 85.58 |
| 3 | GPT-4o | 84.62 |
| 4 | Gemini 2.5 Pro | 76.6 |
| 5 | Gemini 2.5 Flash | 71.47 |
| 6 | Claude Sonnet 4.5 | 68.59 |
| 7 | Grok 4 | 64.74 |
| 8 | Qwen 2.5 VL 72B Instruct | 64.1 |
| 9 | Qwen 2.5 VL 7B Instruct | 58.97 |
| 10 | Claude Haiku 4.5 | 58.01 |
Interactive version: theaggregate.ai/benchmark?slug=contrar-multiple-frames · How It Works · Data refreshed daily, snapshot 2026-10-07.