Categorical Visual Question Answering on CLEVR
88.6AccuracySN
Evaluation Results
| Method | Links | ||
|---|---|---|---|
| SNBase Model=Qwen-VL-4b-Instruct, Probing Set Size=3,000, Configuration=Best probed2026.03 | 88.6 | 89 | |
| Qwen-VL-4b-InstructProbing Set Size=3,0002026.03 | 84.7 | 86.4 | |
| SNBase Model=LLaVA-v1.5-7b, Probing Set Size=3,000, Configuration=Best probed2026.03 | 54.4 | 66.9 | |
| LLaVA-v1.5-7bProbing Set Size=3,0002026.03 | 52.3 | 61.7 |