HID-Chart - Social: leaderboard
The 335 social-domain chart questions of HID-Chart (Pan et al., 2026), asked about high information density charts with several subplots and dense annotations. Accuracy in percent, judged by GPT-5 mini against the annotated answers; higher is better.
Metric: Accuracy (%; GPT-5 mini judge, 335 social questions). Source: arxiv.org. Saturation forecast: Estimated already saturated. 16 models tracked.
Top models
| # | Model | Score | Overall rank |
|---|---|---|---|
| 1 | GPT-4.1 | 59.4 | #492 |
| 2 | Gemini 2.0 Flash | 57 | #702 |
| 3 | GPT-4o | 51.3 | #741 |
| 4 | InternVL3-8B | 41.5 | #1695 |
| 5 | Qwen 2.5 VL 7B | 41.5 | #1796 |
| 6 | InternVL2.5-78B | 37.6 | #1018 |
| 7 | Gemma 3 27B | 34.9 | #1638 |
No result here: #1 Claude Fable 5.2, #2 Gemini 4 Carbon, #5 Claude Opus 5.5.
Interactive version: theaggregate.ai/benchmark?slug=hid-chart-social · How It Works · Data refreshed daily, snapshot 2026-11-12.