Segmentation on MIMOSeg Perspective III held-in (test)
53.1mIoUMIMO
Evaluation Results
| Method | Links | |||
|---|---|---|---|---|
| MIMOMulti-modal Input Aligner=true, LLM=Vicuna-7B2025.10 | 53.1 | 30.2 | 42.7 | |
| MIMO(w/o Aligner)Multi-modal Input Aligner=false, LLM=Vicuna-7B2025.10 | 46.8 | 20.5 | 34 | |
| GLaMMFine-tuned=true2025.10 | 42.1 | 13.3 | 25.3 |