Visual Question Answering on IRIS-120K (Single split)
98.52AccuracyIRIS-4B
Evaluation Results
| Method | Links | |
|---|---|---|
| IRIS-4BGroup=Ours2026.07 | 98.52 | |
| IRIS-8BGroup=Ours2026.07 | 98.19 | |
| IRIS-4B-TFTGroup=Ours2026.07 | 98.19 | |
| IRIS-2BGroup=Ours2026.07 | 97.03 | |
| Lingshu-32BGroup=Med2026.07 | 84.43 | |
| IRIS-4B-SceneGroup=Ours2026.07 | 83.2 | |
| Qwen3-vl-32BGroup=Gen2026.07 | 79.98 | |
| Internvl3.5-14BGroup=Gen2026.07 | 75.62 | |
| Internvl3.5-8BGroup=Gen2026.07 | 75.37 | |
| Qwen3-vl-8BGroup=Gen2026.07 | 73.06 | |
| GPT-5-miniGroup=Gen2026.07 | 72.08 | |
| Qwen3-vl-4BGroup=Gen2026.07 | 70.68 | |
| HGPT-V-34BGroup=Med2026.07 | 68.29 | |
| Lingshu-7BGroup=Med2026.07 | 64.25 | |
| Qwen3-vl-2BGroup=Gen2026.07 | 62.36 | |
| MG-4BGroup=Med2026.07 | 60.38 | |
| HuluMed-32BGroup=Med2026.07 | 60.05 | |
| HuluMed-14BGroup=Med2026.07 | 59.59 | |
| MG1.5-4BGroup=Med2026.07 | 58.24 | |
| HuluMed-7BGroup=Med2026.07 | 49.09 | |
| HGPT-V-7BGroup=Med2026.07 | 25.62 |