Grounded Conversation Generation on GeoPixelD
24.5CIDErCLV-Net-7B
Evaluation Results
| Method | Links | ||||||
|---|---|---|---|---|---|---|---|
| CLV-Net-7BFine-tuned on training data=false2025.12 | 24.5 | 26.2 | 21.8 | 55.3 | 41.9 | 0.08 | |
| GeoPixel-7BFine-tuned on training data=false2025.12 | 21.6 | 24 | 19 | 52.3 | 38.8 | 0.05 | |
| PixelLM-7BFine-tuned on training data=true2025.12 | 18.3 | 22.5 | 10.5 | 42.4 | 29.6 | 0.03 | |
| GLaMM-7BFine-tuned on training data=true2025.12 | 15.7 | 23 | 12.5 | 46.4 | 32.8 | 0.09 | |
| LISA-7BFine-tuned on training data=true2025.12 | 14.6 | 22.3 | 8.5 | 42.7 | 29 | 0.61 |