HG-Bench - Answer-Region F1: leaderboard
Metric: Answer-region F1 (0-100): per-page box F1 for complete answer regions, averaged within each sample and then over the successfully parsed samples; 500 human-annotated multi-page handwritten K-12 homework samples; the model outputs page-aware question-level answer boxes and ordered step boxes as JSON with normalized 0-1000 coordinates, one shared prompt, default decoding, one format-reminder retry; higher is better. Source: arxiv.org. Saturation forecast: Around June 2028. 9 models tracked.
Top models
| # | Model | Score |
|---|---|---|
| 1 | Gemini 3 Pro (Preview) | 50.9 |
| 2 | GLM-5V Turbo | 46.69 |
| 3 | Qwen 3.5 397B A17B | 42.71 |
| 4 | Kimi K2.5 | 31.21 |
| 5 | Claude Sonnet 4.6 | 16.83 |
| 6 | GPT-5.4 | 14.91 |
Interactive version: theaggregate.ai/benchmark?slug=hg-bench-answer-region-f1 · How It Works · Data refreshed daily, snapshot 2026-09-29.