RS-Neg VQA - Object Negation: leaderboard
Yes/no questions that negate a present or an absent scene element, restricted to the items whose negation targets an object. RS-Neg (Han et al., Peng Cheng Laboratory, 2026) tests negation understanding on remote-sensing images, with negated queries written by LLMs from image captions and checked against the image by a detector-guided visual search. Accuracy in percent over 5,675 items, answered zero-shot; higher is better.
Metric: Accuracy (%; yes/no questions that negate a present or an absent scene element; the 5,675 items whose negation is object-level; zero-shot). Source: arxiv.org. Saturation forecast: Around February 2027. 7 models tracked.
Top models
| # | Model | Score | Overall rank |
|---|---|---|---|
| 1 | Qwen 2.5 VL 7B Instruct | 80.19 | #1842 |
| 2 | Qwen 3 VL 8B Instruct | 76 | #982 |
No result here: #1 Claude Fable 5.2, #2 Gemini 4 Carbon, #5 Claude Opus 5.5.
Interactive version: theaggregate.ai/benchmark?slug=rs-neg-vqa-object-negation · How It Works · Data refreshed daily, snapshot 2026-11-12.