MinCU: leaderboard

Metric: Grounded F1 (%; overall on 2,000 image pairs, 500 per change type; 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 2033. 15 models tracked.

Top models

#ModelScore
1Qwen 3 VL 8B48.84
2Gemini 3.1 Pro (Preview)42.89
3Seed 1.842.43
4Gemini 2.5 Pro31.92
5GPT-5.419.58

Interactive version: theaggregate.ai/benchmark?slug=mincu · How It Works · Data refreshed daily, snapshot 2026-09-26.