Multimodal Complaint Scene Reasoning on ComplaintScene MM
84.48Average ScoreSKG-VLA
Evaluation Results
| Method | Links | ||||
|---|---|---|---|---|---|
| SKG-VLABackbone=Qwen2.5-3B, Variant=mm-our2026.05 | 84.48 | 79.26 | 86.03 | 88.16 | |
| SKG-VLABackbone=Qwen2.5-1.5B, Variant=mm-our2026.05 | 83.61 | 78.14 | 84.76 | 87.93 | |
| mm-oriBackbone=Qwen2.5-3B2026.05 | 82.17 | 78.65 | 81.44 | 86.42 | |
| Qwen2.5-VL-3B-InstructBackbone=Qwen2.5-3B2026.05 | 80.88 | 78.92 | 79.87 | 83.86 | |
| SKG-VLABackbone=Qwen2.5-0.5B, Variant=mm-our2026.05 | 80.54 | 74.31 | 81.29 | 86.03 | |
| mm-oriBackbone=Qwen2.5-1.5B2026.05 | 80.43 | 76.92 | 79.15 | 85.22 | |
| mm-oriBackbone=Qwen2.5-0.5B2026.05 | 76.12 | 72.46 | 74.88 | 81.02 |