CArtBench - CuratorQA (Technique Cues): leaderboard
Chinese art understanding on the questions on visible technique and style cues of CuratorQA, the curator-style question task of CArtBench (2026): multiple-choice and true-or-false questions generated from Palace Museum catalog text and audited by an art expert, answered from the artwork image. Exact-match accuracy in percent; higher is better.
Metric: Exact-match accuracy (%) on the visible technique and style cue (QA4) CuratorQA questions of CArtBench (Palace Museum artworks with catalog-grounded multiple-choice and true-or-false questions, answered from the artwork image, constrained decoding, temperature 0); higher is better. Source: arxiv.org. Saturation forecast: Around February 2027. 9 models tracked. Updated 7 Oct 2026.
Top models
| # | Model | Score | Overall rank |
|---|---|---|---|
| 1 | Qwen 3 VL 235B A22B (Thinking) | 89 | #462 |
| 2 | Qwen 3 VL 30B A3B | 88 | #791 |
| 3 | Qwen 2.5 VL 72B | 87 | #782 |
| 4 | Qwen 2.5 VL 32B | 87 | #944 |
| 5 | GPT-5 Mini | 86 | #373 |
| 6 | Qwen 2.5 VL 7B | 81 | #1878 |
| 7 | GPT-5 Nano | 65 | #912 |
| 8 | GLM-4.5V | 62 | #756 |
| 9 | Gemini 2.5 Flash | 52 | #466 |
No result here: #2 Claude Fable 5.2, #3 Gemini 4 Neon, #4 GPT-6.2 Astra.
Interactive version: theaggregate.ai/benchmark?slug=cartbench-curatorqa-technique-cues · How It Works · Data refreshed daily, snapshot 2026-11-22.