MindEdit-Bench - Cross-View Perspective Transformation: leaderboard
The perspective-taking task (stand at A facing B, locate C) with the three objects spread across different photos of the scene; 8-24 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 July 2028. 15 models tracked.
Top models
| # | Model | Score | Overall rank |
|---|---|---|---|
| 1 | Qwen 3.6 Plus | 20 | #217 |
| 2 | Gemini 3.1 Pro (Preview) | 18.3 | #91 |
| 3 | Claude Opus 4.6 | 10.5 | #100 |
| 4 | GPT-5.4 | 8.2 | #121 |
| 5 | Qwen 3.5 35B A3B | 5.5 | #480 |
| 6 | InternVL3-8B | 5.1 | #1695 |
| 7 | Seed 2.0 Pro | 5 | #209 |
| 8 | Kimi K2.5 | 2.9 | #252 |
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-cross-view-perspective-transformation · How It Works · Data refreshed daily, snapshot 2026-11-12.