Seeing Culture - VQA Accuracy: leaderboard
Seeing Culture Benchmark (EMNLP 2025) tests cultural visual reasoning over 1,065 images of Southeast Asian cultural artifacts; this board reports overall multiple-choice VQA accuracy.
Metric: Overall Accuracy (%). Source: seeingculture-benchmark.github.io. Saturation forecast: Estimated already saturated. 10 models tracked.
Top models
| # | Model | Score | Overall rank |
|---|---|---|---|
| 1 | O3 | 85.15 | #124 |
| 2 | Gemini 2.5 Pro | 82.88 | #152 |
| 3 | GPT-4.1 | 81.97 | #248 |
| 4 | Qwen 2.5 VL 7B | 53.78 | #643 |
No result here: #3 Claude Opus 5.5, #6 GPT-6 Astra, #7 GPT-6.1 Sol.
Interactive version: theaggregate.ai/benchmark?slug=seeing-culture-vqa-accuracy · How It Works · Data refreshed daily, snapshot 2026-10-20.