DRBench (Dense-Scene Reasoning) - Perception: leaderboard
The perception category of DRBench: identifying attributes of objects reached through multi-hop referring expressions. DRBench (Shi et al., Shanghai Jiao Tong University, 2026, introduced with DRScaffold) asks questions about cluttered indoor (Hypersim) and urban driving (Cityscapes) scenes that require grounding several objects and their relations. Combined accuracy in percent over 270 test questions: multiple-choice answers by exact letter match and open answers by a Qwen-Max semantic judge, weighted by question count; higher is better.
Metric: Combined accuracy (%; attributes of entities reached through multi-hop referring expressions; 270 test questions). Source: arxiv.org. Saturation forecast: Estimated already saturated. 9 models tracked.
Top models
| # | Model | Score | Overall rank |
|---|---|---|---|
| 1 | Phi-4 Multimodal Instruct | 44.1 | #3914 |
| 2 | InternVL2.5-2B | 36.3 | #6868 |
No result here: #1 Claude Fable 5.2, #2 Gemini 4 Carbon, #5 Claude Opus 5.5.
Interactive version: theaggregate.ai/benchmark?slug=drbench-dense-scene-reasoning-perception · How It Works · Data refreshed daily, snapshot 2026-11-12.