Few-shot Visual Question Answering on VL-ICL Bench Open-MI (test)
77.9Mean AccuracyMAPD
Evaluation results
| Method | Links | |
|---|---|---|
| MAPDTTA (Test-Time Adaptation) method=FT (K ≤ 30), MT (Meta-Tasks used during training)=Yes, Base LMM=LLaVA-ATT-Qwen2.5 7B, Attention-mapper parameters count=24M2025.06 | 77.9 | |
| Multi-TaskPDTTA (Test-Time Adaptation) method=FT (K ≤ 30), MT (Meta-Tasks used during training)=Yes, Base LMM=LLaVA-ATT-Qwen2.5 7B, Attention-mapper parameters count=24M2025.06 | 74.6 | |
| LoRA ([0-15] LLM layers + ATT)TTA (Test-Time Adaptation) method=FT (K ≤ 30), Base LMM=LLaVA-ATT-Qwen2.5 7B, Attention-mapper parameters count=~24M2025.06 | 69.1 | |
| NoMeta-taskPDTTA (Test-Time Adaptation) method=FT (K ≤ 30), MT (Meta-Tasks used during training)=No, Base LMM=LLaVA-ATT-Qwen2.5 7B, Attention-mapper parameters count=24M2025.06 | 68 | |
| LoRA ([0-15] LLM layers)TTA (Test-Time Adaptation) method=FT (K ≤ 30), Base LMM=LLaVA-ATT-Qwen2.5 7B, Attention-mapper parameters count=~24M2025.06 | 67.3 | |
| In-ContextPDTTA (Test-Time Adaptation) method=FT (K ≤ 30), MT (Meta-Tasks used during training)=Yes, Base LMM=LLaVA-ATT-Qwen2.5 7B, Attention-mapper parameters count=24M2025.06 | 64.5 | |
| Model-AvgPDTTA (Test-Time Adaptation) method=FT (K ≤ 30), MT (Meta-Tasks used during training)=No, Base LMM=LLaVA-ATT-Qwen2.5 7B, Attention-mapper parameters count=24M2025.06 | 63.1 | |
| LoRA (All LLM layers)TTA (Test-Time Adaptation) method=FT (K ≤ 30), Base LMM=LLaVA v1.5, Attention-mapper parameters count=~300M2025.06 | 55.1 | |
| MAPDTTA (Test-Time Adaptation) method=ICL, MT (Meta-Tasks used during training)=Yes, Base LMM=LLaVA-ATT-Qwen2.5 7B, Attention-mapper parameters count=24M2025.06 | 53.3 | |
| In-ContextPDTTA (Test-Time Adaptation) method=ICL, MT (Meta-Tasks used during training)=Yes, Base LMM=LLaVA-ATT-Qwen2.5 7B, Attention-mapper parameters count=24M2025.06 | 51.1 | |
| Multi-TaskPDTTA (Test-Time Adaptation) method=ICL, MT (Meta-Tasks used during training)=Yes, Base LMM=LLaVA-ATT-Qwen2.5 7B, Attention-mapper parameters count=24M2025.06 | 48.6 | |
| NoMeta-taskPDTTA (Test-Time Adaptation) method=ICL, MT (Meta-Tasks used during training)=No, Base LMM=LLaVA-ATT-Qwen2.5 7B, Attention-mapper parameters count=24M2025.06 | 43.8 | |
| Model-AvgPDTTA (Test-Time Adaptation) method=ICL, MT (Meta-Tasks used during training)=No, Base LMM=LLaVA-ATT-Qwen2.5 7B, Attention-mapper parameters count=24M2025.06 | 26.6 |