Scientific Figure Reasoning on CharXiv
48.9AccuracyGPT4o
Evaluation Results
| Method | Links | |
|---|---|---|
| GPT4oAccess=Close-source2026.02 | 48.9 | |
| InternVL3.5-38BParameters=38B2026.02 | 44.4 | |
| SAYO-InternVL-8BBase Model=InternVL3.5-8B2026.02 | 43.2 | |
| Qwen3-VL-8BParameters=8B2026.02 | 42.7 | |
| SAYO-Qwen-8BBase Model=Qwen3-VL-8B2026.02 | 42.5 | |
| SAYO-Qwen-4BBase Model=Qwen3-VL-4B2026.02 | 41.9 | |
| NoisyRollout-7BParameters=7B2026.02 | 40.5 | |
| Qwen3-VL-30B-A3BParameters=30B-A3B2026.02 | 39.8 | |
| InternVL3.5-14BParameters=14B2026.02 | 38.9 | |
| Semantic-back-7BParameters=7B2026.02 | 38.1 | |
| Qwen3-VL-4BParameters=4B2026.02 | 38 | |
| OpenVLThinker-7BParameters=7B2026.02 | 35.2 | |
| InternVL3.5-8BParameters=8B2026.02 | 34.1 | |
| InternVL3.5-30B-A3BParameters=30B-A3B2026.02 | 33.7 | |
| Kimi-VL-16BParameters=16B2026.02 | 31.3 | |
| ViGoRL2026.02 | 30.4 | |
| R1-Onevision-7BParameters=7B2026.02 | 20.9 |