PhoneSafety - Trap Resistance: leaderboard
PhoneSafety scenario family (Tang et al., Tencent Hunyuan, 2026) built on deceptive screens, where the safe action is to hit the genuine control, such as the real close button of a misleading pop-up. Safe-action rate in percent on 78 moments; higher is better.
Metric: Safe-action rate (%; 78 trap-resistance moments). Source: arxiv.org. Saturation forecast: Estimated already saturated. 8 models tracked.
Top models
| # | Model | Score | Overall rank |
|---|---|---|---|
| 1 | Claude Opus 4.6 | 97.4 | #100 |
| 2 | Kimi K2.5 | 92.3 | #252 |
| 3 | Gemini 3.1 Pro (Preview) | 91 | #91 |
| 4 | Seed 2.0 Pro | 89.7 | #209 |
No result here: #1 Claude Fable 5.2, #2 Gemini 4 Carbon, #5 Claude Opus 5.5.
Interactive version: theaggregate.ai/benchmark?slug=phonesafety-trap-resistance · How It Works · Data refreshed daily, snapshot 2026-11-12.