Visual Question Answering on IRIS 120K (Multi split)
88.96AccuracyIRIS-4B
Evaluation Results
| Method | Links | |
|---|---|---|
| IRIS-4BGroup=Ours2026.07 | 88.96 | |
| IRIS-8BGroup=Ours2026.07 | 88.55 | |
| IRIS-4B-TFTGroup=Ours2026.07 | 88.26 | |
| IRIS-2BGroup=Ours2026.07 | 83.98 | |
| GPT-5-miniGroup=Gen2026.07 | 67.48 | |
| Qwen3-vl-32BGroup=Gen2026.07 | 62.64 | |
| Lingshu-32BGroup=Med2026.07 | 59.71 | |
| Qwen3-vl-8BGroup=Gen2026.07 | 57.41 | |
| Internvl3.5-14BGroup=Gen2026.07 | 54.59 | |
| Qwen3-vl-4BGroup=Gen2026.07 | 51.98 | |
| HGPT-V-34BGroup=Med2026.07 | 50.36 | |
| Internvl3.5-8BGroup=Gen2026.07 | 49.41 | |
| IRIS-4B-SceneGroup=Ours2026.07 | 48.42 | |
| Qwen3-vl-2BGroup=Gen2026.07 | 44.15 | |
| Lingshu-7BGroup=Med2026.07 | 37.7 | |
| MG-4BGroup=Med2026.07 | 36.22 | |
| MG1.5-4BGroup=Med2026.07 | 31.14 | |
| HuluMed-32BGroup=Med2026.07 | 30.39 | |
| HuluMed-7BGroup=Med2026.07 | 26.61 | |
| HuluMed-14BGroup=Med2026.07 | 25.63 | |
| HGPT-V-7BGroup=Med2026.07 | 8.45 |