HakushoBench (Direct): leaderboard
Metric: Accuracy (%), direct prompt (answer accurately and concisely); 2,053 manually written short-answer questions on chart and table images from 33 Japanese government white papers, judged correct or incorrect by GPT-5.1, mean of three runs, temperature 0; higher is better. Source: arxiv.org. Saturation forecast: Around January 2027. 8 models tracked.
Top models
| # | Model | Score |
|---|---|---|
| 1 | GPT-5.1 (Non-reasoning) | 52.8 |
| 2 | GPT-4o (2024-11-20) | 52.6 |
| 3 | InternVL3.5-8B | 39.4 |
Interactive version: theaggregate.ai/benchmark?slug=hakushobench-direct · How It Works · Data refreshed daily, snapshot 2026-09-29.