ContrAR (Multiple Frames) - Smart Retail: leaderboard
Metric: Detection accuracy (%) for contradictory virtual content attacks in AR videos (66 smart retail videos; 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-5 | 80.3 |
| 2 | GPT-4.1 | 78.79 |
| 3 | GPT-4o | 75.76 |
| 4 | Grok 4 | 69.7 |
| 5 | Gemini 2.5 Pro | 65.15 |
| 6 | Gemini 2.5 Flash | 63.64 |
| 7 | Claude Haiku 4.5 | 60.61 |
| 8 | Qwen 2.5 VL 72B Instruct | 59.09 |
| 9 | Claude Sonnet 4.5 | 56.06 |
| 10 | Qwen 2.5 VL 7B Instruct | 46.97 |
Interactive version: theaggregate.ai/benchmark?slug=contrar-multiple-frames-smart-retail · How It Works · Data refreshed daily, snapshot 2026-10-07.