PriMobiBench - User Profile Construction - Inferred Attributes: leaderboard
PriMobiBench (2026) user profile construction restricted to attributes that must be inferred from indirect visual cues across screenshots rather than read off the screen, under chain-of-thought prompting. Share of those attributes inferred correctly, in percent, judged by GPT-5-mini; higher means more is extracted.
Metric: Profile success rate (%; share of the ground-truth attributes of 50 synthetic user profiles that the model correctly infers after reading every screenshot of a MobiLeak trajectory (320 trajectories across 16 Android apps with 2,960 inserted privacy elements in 25 attribute types), judged for semantic entailment by GPT-5-mini; chain-of-thought prompting; attributes that must be inferred from indirect visual cues). Source: arxiv.org. Saturation forecast: Estimated already saturated. 5 models tracked.
Top models
| # | Model | Score | Overall rank |
|---|---|---|---|
| 1 | GPT-5 Mini | 59.1 | #349 |
| 2 | GLM-4.5V | 48.3 | #739 |
| 3 | Gemma 3 27B | 36.7 | #1638 |
| 4 | Qwen 2.5 VL 7B | 25.9 | #1796 |
No result here: #1 Claude Fable 5.2, #2 Gemini 4 Carbon, #5 Claude Opus 5.5.
Interactive version: theaggregate.ai/benchmark?slug=primobibench-user-profile-construction-inferred-attributes · How It Works · Data refreshed daily, snapshot 2026-11-12.