WSI Captioning on SlideBench
0.38BLEU-1PathFLIP
Evaluation results
| Method | Links | |||||
|---|---|---|---|---|---|---|
| PathFLIPfinetune Qwen3=true2025.12 | 0.38 | 0.23 | 0.13 | 0.34 | 0.76 | |
| SlideChat2025.12 | 0.37 | 0.21 | 0.12 | 0.24 | 0.71 | |
| PathGen-LLavaInput sampling strategy=30 randomly sampled patches from each slide2025.12 | 0.35 | 0.17 | 0.09 | 0.16 | 0.65 | |
| CPath-Omni2025.12 | 0.33 | 0.19 | 0.1 | 0.17 | 0.7 | |
| PathFLIPT-R Atten.=false2025.12 | 0.32 | 0.19 | 0.12 | 0.34 | 0.64 | |
| MedDrInput sampling strategy=30 randomly sampled patches from each slide2025.12 | 0.31 | 0.16 | 0.06 | 0.07 | 0.68 | |
| PathFLIPRegion Q-Former=false2025.12 | 0.31 | 0.19 | 0.12 | 0.33 | 0.64 | |
| PathFLIPfinetune Qwen3=false2025.12 | 0.25 | 0.14 | 0.09 | 0.33 | 0.64 | |
| Quilt-LLavaInput sampling strategy=30 randomly sampled patches from each slide2025.12 | 0.23 | 0.09 | 0.04 | 0.16 | 0.56 | |
| PathFLIPSlide Q-Former=false2025.12 | 0.21 | 0.13 | 0.08 | 0.32 | 0.67 | |
| LLava-MedInput sampling strategy=30 randomly sampled patches from each slide2025.12 | 0.18 | 0.12 | 0.06 | 0.12 | 0.58 |