M3-VQA (Gold Section Evidence): leaderboard
Metric: IoU accuracy (%) between the predicted and gold answer sets (exact match against Wikidata answer aliases, one-year margin for dates), averaged over the M3-VQA multi-entity, multi-hop knowledge-based visual questions, with the Wikipedia sections containing the gold evidence of every hop added to the input; higher is better. Source: arxiv.org. Saturation forecast: Estimated already saturated. 16 models tracked.
Top models
| # | Model | Score |
|---|---|---|
| 1 | InternVL2.5-78B | 55.16 |
| 2 | GPT-4o | 53.07 |
| 3 | Qwen 2.5 VL 72B Instruct | 52.35 |
| 4 | Qwen 2.5 VL 32B Instruct | 50.02 |
| 5 | Qwen 2.5 VL 7B Instruct | 42.66 |
| 6 | MiniCPM-V-2.6 | 39.11 |
| 7 | Qwen 2 VL 7B Instruct | 38.92 |
| 8 | InternVL2.5-2B | 31.94 |
Interactive version: theaggregate.ai/benchmark?slug=m3-vqa-gold-section-evidence · How It Works · Data refreshed daily, snapshot 2026-10-07.