Open-ended Generative VQA on IRIS-120K Scene-VQA
44.23BLEU-1IRIS-4B
Evaluation Results
| Method | Links | |||
|---|---|---|---|---|
| IRIS-4BGroup=Ours2026.07 | 44.23 | 36.08 | 24.45 | |
| IRIS-8BGroup=Ours2026.07 | 44.19 | 36.32 | 24.74 | |
| IRIS-4B-SceneGroup=Ours2026.07 | 44.13 | 36.01 | 24.41 | |
| IRIS-4B-TFTGroup=Ours2026.07 | 40.46 | 28.86 | 18.17 | |
| IRIS-2BGroup=Ours2026.07 | 30.28 | 26.7 | 16.86 | |
| HuluMed-7BGroup=Med2026.07 | 26.12 | 25.49 | 15.84 | |
| HuluMed-32BGroup=Med2026.07 | 26.02 | 25.92 | 15.9 | |
| HuluMed-14BGroup=Med2026.07 | 25.98 | 25.7 | 15.7 | |
| HGPT-V-34BGroup=Med2026.07 | 24.4 | 22.02 | 13.89 | |
| HGPT-V-7BGroup=Med2026.07 | 19.58 | 18.53 | 11.59 | |
| Qwen3-vl-32BGroup=Gen2026.07 | 19.22 | 20.36 | 11.84 | |
| Internvl3.5-14BGroup=Gen2026.07 | 18.5 | 20.22 | 11.34 | |
| Lingshu-32BGroup=Med2026.07 | 18.08 | 23.66 | 15.13 | |
| MG-4BGroup=Med2026.07 | 17.76 | 19.94 | 12.27 | |
| Internvl3.5-8BGroup=Gen2026.07 | 15.42 | 18.68 | 10.17 | |
| GPT-5-miniGroup=Gen2026.07 | 13.73 | 22.02 | 11.81 | |
| Lingshu-7BGroup=Med2026.07 | 13.59 | 20.64 | 13.46 | |
| Qwen3-vl-8BGroup=Gen2026.07 | 13.33 | 17.09 | 9.18 | |
| MG1.5-4BGroup=Med2026.07 | 11.72 | 16.36 | 8.26 | |
| Qwen3-vl-4BGroup=Gen2026.07 | 10.86 | 16.39 | 8.63 | |
| Qwen3-vl-2BGroup=Gen2026.07 | 7.15 | 16.59 | 8.96 |