MinCU - Object: leaderboard
The object-change split of MinCU (Mu et al., 2026): 500 image pairs in which an entity is replaced, introduced or removed. Grounded F1 in percent, crediting a prediction only when its box, change description and entity label are all correct; higher is better.
Metric: Grounded F1 (%; object changes (an entity replaced, introduced or removed), 500 image pairs; a true positive needs the predicted box to reach IoU 0.5 with the changed region, the change description BERTScore 0.85 against the reference and the correct changed-entity label; zero-shot). Source: arxiv.org. Saturation forecast: Around February 2027. 15 models tracked.
Top models
| # | Model | Score | Overall rank |
|---|---|---|---|
| 1 | Gemini 3.1 Pro (Preview) | 59.25 | #91 |
| 2 | Seed 1.8 | 44.82 | #239 |
| 3 | GPT-5.4 | 19.67 | #121 |
No result here: #1 Claude Fable 5.2, #2 Gemini 4 Carbon, #5 Claude Opus 5.5.
Interactive version: theaggregate.ai/benchmark?slug=mincu-object · How It Works · Data refreshed daily, snapshot 2026-11-12.