Scientific Multimodal Document Reasoning on SCIMDR-Eval
49.9AccuracyGPT-5.2
Evaluation Results
| Method | Links | |
|---|---|---|
| GPT-5.22026.03 | 49.9 | |
| Qwen2.5-VL-7B + SCIMDRFine-tuned=true, Training Data=SCIMDR2026.03 | 49.1 | |
| GPT-5.12026.03 | 47.2 | |
| GPT-4o2026.03 | 24.7 | |
| Qwen2.5-VL-7B2026.03 | 19.8 |