UniG2U-Bench (Generate-then-Answer) - Physics Reasoning: leaderboard
Metric: Accuracy (%) on the 200 mechanics and optics items, mean of the two subtask accuracies, from UniG2U-Bench, Generate-then-Answer: the unified model first generates an intermediate image (an auxiliary drawing or state) and then answers from the original and generated images, greedy decoding; higher is better. Source: arxiv.org. Saturation forecast: Rough model projection: around 2026. 11 models tracked.
Top models
| # | Model | Score | Overall rank |
|---|---|---|---|
| 1 | OmniGen2 | 41.5 | |
| 2 | Bagel | 40.5 | |
| 3 | Show-o2 | 40 | |
| 4 | ILLUME+ | 35.5 | |
| 5 | UniWorld-V1 | 34.5 | |
| 6 | STAR-7B | 32.5 | |
| 7 | UniPic2 | 31 | |
| 8 | OneCAT-3B | 29 | |
| 9 | Uni-Video | 29 | |
| 10 | MIO | 28 |
Interactive version: theaggregate.ai/benchmark?slug=unig2u-bench-generate-then-answer-physics-reasoning · How It Works · Data refreshed daily, snapshot 2026-10-11.