PInVerify (Direct Query): leaderboard
Active instance verification on the 3,000 PInVerify episodes (Jiang, 2026) with the multi-view pipeline asking the multimodal model about the whole description at once instead of attribute by attribute (Grounding DINO detection, LLM next-best-view, at most six steps). Accuracy from 0 to 1; higher is better.
Metric: Verification accuracy (0-1; share of 3,000 episodes decided correctly; multi-view pipeline querying the whole description at once, LLM next-best-view, Grounding DINO detection, at most six steps). Source: arxiv.org. Saturation forecast: Rough model projection: around 2026. 3 models tracked.
Top models
| # | Model | Score | Overall rank |
|---|---|---|---|
| 1 | SenseNova-SI-1.2-InternVL3-8B | 0.83 | |
| 2 | Qwen3-VL-4B | 0.83 | |
| 3 | Qwen3-VL-8B | 0.76 |
Interactive version: theaggregate.ai/benchmark?slug=pinverify-direct-query · How It Works · Data refreshed daily, snapshot 2026-11-12.