Attribution on CLICONSUMMATION multimodal (163 samples)
65.52Text F1IMG_RAW
Evaluation Results
| Method | Links | ||||
|---|---|---|---|---|---|
| IMG_RAWBackbone=Qwen2.5-VL-7B, hyperparameters (k, tau)=k=3, tau=0.162026.01 | 65.52 | 30.75 | 75.47 | 26.46 | |
| IMG_CAPBackbone=Qwen2.5-VL-7B, hyperparameters (k, tau)=k=3, tau=0.162026.01 | 58.37 | 38.52 | 77.85 | 34.87 | |
| Sent-EmbeddingBackbone=LLaVA-NeXT-7B2026.01 | 57.11 | 25.4 | 73.77 | 17.65 | |
| IMG_CAPBackbone=LLaVA-NeXT-7B, hyperparameters (k, tau)=k=5, tau=0.122026.01 | 56.31 | 28.02 | 65.13 | 21.39 | |
| Sent-EmbeddingBackbone=Qwen2.5-VL-7B2026.01 | 53.21 | 10.86 | 76.53 | 8.48 | |
| IMG_RAWBackbone=LLaVA-NeXT-7B, hyperparameters (k, tau)=k=5, tau=0.122026.01 | 49.37 | 13.32 | 81.49 | 18.12 | |
| Self-AttributionBackbone=Qwen2.5-VL-7B2026.01 | 29.56 | 12.21 | 44.64 | 12.21 | |
| Self-AttributionBackbone=LLaVA-NeXT-7B2026.01 | 18.56 | 6.87 | 35.44 | 6.43 |