OCR-related Understanding Tasks on CharXiv RQ
60.2AccuracyClaude-3.5 Sonnet
Evaluation Results
| Method | Links | |
|---|---|---|
| Claude-3.5 Sonnet2025.02 | 60.2 | |
| Qwen2.5-VL-72BModel Size=72B2025.02 | 49.7 | |
| GPT-4o2025.02 | 47.1 | |
| Qwen3-VL-SegModel Size=4B, Training Stage=S-22026.05 | 45.2 | |
| Gemini 1.5 Pro2025.02 | 43.3 | |
| Qwen2.5-VL-7BModel Size=7B2025.02 | 42.5 | |
| InternVL2.5-78BModel Size=78B2025.02 | 42.4 | |
| Qwen3-VLModel Size=4B, Training Stage=instruct2026.05 | 39.7 | |
| InternVL-3.5Model Size=4B2026.05 | 39.6 | |
| Qwen3-VLModel Size=4B, Training Stage=S-12026.05 | 38.9 | |
| Qwen2.5-VL-3BModel Size=3B2025.02 | 31.3 |