Few-shot Visual Question Answering on VL-ICL Bench CLEVR (test)
31.4AccuracyMAPD
Evaluation results
| Method | Links | |
|---|---|---|
| MAPDTTA (Test-Time Adaptation) method=FT (K ≤ 30), MT (Meta-Tasks used during training)=Yes2025.06 | 31.4 | |
| In-ContextPDTTA (Test-Time Adaptation) method=FT (K ≤ 30), MT (Meta-Tasks used during training)=Yes2025.06 | 30.9 | |
| LoRA ([0-15] LLM layers)TTA (Test-Time Adaptation) method=FT (K ≤ 30)2025.06 | 30 | |
| Multi-TaskPDTTA (Test-Time Adaptation) method=FT (K ≤ 30), MT (Meta-Tasks used during training)=Yes2025.06 | 29.9 | |
| Model-AvgPDTTA (Test-Time Adaptation) method=FT (K ≤ 30), MT (Meta-Tasks used during training)=No2025.06 | 29.1 | |
| LoRA ([0-15] LLM layers + ATT)TTA (Test-Time Adaptation) method=FT (K ≤ 30)2025.06 | 28.7 | |
| NoMeta-taskPDTTA (Test-Time Adaptation) method=FT (K ≤ 30), MT (Meta-Tasks used during training)=No2025.06 | 25.8 | |
| In-ContextPDTTA (Test-Time Adaptation) method=ICL, MT (Meta-Tasks used during training)=Yes2025.06 | 24.1 | |
| NoMeta-taskPDTTA (Test-Time Adaptation) method=ICL, MT (Meta-Tasks used during training)=No2025.06 | 18 | |
| LoRA (All LLM layers)TTA (Test-Time Adaptation) method=FT (K ≤ 30)2025.06 | 15.1 | |
| Multi-TaskPDTTA (Test-Time Adaptation) method=ICL, MT (Meta-Tasks used during training)=Yes2025.06 | 12.5 | |
| MAPDTTA (Test-Time Adaptation) method=ICL, MT (Meta-Tasks used during training)=Yes2025.06 | 12.3 | |
| Model-AvgPDTTA (Test-Time Adaptation) method=ICL, MT (Meta-Tasks used during training)=No2025.06 | 7.6 |