MindEdit-Bench - Spatial Editing: leaderboard
How a relation between objects changes after a hypothetical translation or rotation of one object, whose answer appears in none of the photos; 14-16 choices. Accuracy (%), mean of three greedy answer-only runs, higher is better (MindEdit-Bench, Yu et al., ZODA, Zhejiang University and the University of Tokyo, 2026).
Metric: Accuracy (%). Source: arxiv.org. Saturation forecast: Around September 2028. 15 models tracked.
Top models
| # | Model | Score | Overall rank |
|---|---|---|---|
| 1 | Gemini 3.1 Pro (Preview) | 36.9 | #91 |
| 2 | Qwen 3.6 Plus | 35.6 | #217 |
| 3 | Claude Opus 4.6 | 28.4 | #100 |
| 4 | Seed 2.0 Pro | 27.9 | #209 |
| 5 | Qwen 3.5 35B A3B | 20 | #480 |
| 6 | Kimi K2.5 | 19.4 | #252 |
| 7 | InternVL3-8B | 18.6 | #1695 |
| 8 | GPT-5.4 | 15.3 | #121 |
No result here: #1 Claude Fable 5.2, #2 Gemini 4 Carbon, #5 Claude Opus 5.5.
Interactive version: theaggregate.ai/benchmark?slug=mindedit-bench-spatial-editing · How It Works · Data refreshed daily, snapshot 2026-11-12.