HCSU: leaderboard
Metric: Average accuracy (%) over the stone-rubbing (Bei) and ink-manuscript (Tie) domains, 8-way controlled candidate selection: identify which of 8 calligraphers (1 correct, 7 distractors, each with a style description and reference image) wrote a target character image, temperature 0.1; chance 12.5; higher is better. Source: arxiv.org. Saturation forecast: Around 2029. 9 models tracked.
Top models
| # | Model | Score |
|---|---|---|
| 1 | GPT-5.2 | 34.58 |
| 2 | GPT-4o | 32.93 |
| 3 | Claude Sonnet 4.5 | 32.82 |
| 4 | Qwen 3 VL 235B A22B Instruct | 30.25 |
| 5 | Gemini 2.5 Pro | 29.32 |
| 6 | Qwen 2.5 VL 72B Instruct | 24.93 |
| 7 | Qwen 3 VL 30B A3B Instruct | 21.81 |
| 8 | Qwen 2.5 VL 7B Instruct | 20.25 |
Interactive version: theaggregate.ai/benchmark?slug=hcsu · How It Works · Data refreshed daily, snapshot 2026-09-29.