Long Captioning on VRSBench
48.1BLEU-1LLaVA-1.5
Evaluation Results
| Method | Links | |||
|---|---|---|---|---|
| LLaVA-1.5Total Params=7B, Evaluation Protocol=Finetuned2025.12 | 48.1 | 14.7 | 3,390 | |
| Mini-GeminiTotal Params=7B, Evaluation Protocol=Finetuned2025.12 | 47.6 | 14.3 | 3,350 | |
| GeoChatTotal Params=7B, Evaluation Protocol=Finetuned2025.12 | 46.7 | 13.8 | 2,820 | |
| GeoMELTTotal Params=241M, Evaluation Protocol=Finetuned2025.12 | 45.8 | 13.8 | 2,980 | |
| GPT-4VTotal Params=UNK, Evaluation Protocol=Zero-shot2025.12 | 37.2 | 8.6 | 1,910 | |
| MiniGPT-v2Total Params=7B, Evaluation Protocol=Finetuned2025.12 | 36.8 | 8.7 | 2,140 | |
| GeoChatTotal Params=7B, Evaluation Protocol=Zero-shot2025.12 | 13.9 | 1.4 | 40 |