Image Captioning on Flickr30K (BLEU-1, BLEU-2, METEOR, ROUGE-L, CIDEr, SPICE)
0.255BLEU-1InternVL-2.5-26B
Evaluation Results
| Method | Links | ||||||
|---|---|---|---|---|---|---|---|
| InternVL-2.5-26B2025.12 | 0.255 | 0.159 | 0.14 | 0.328 | 0.666 | 0.234 | |
| LUVCBackbone=InternVL-2.5-26B, Speedup=1.732025.12 | 0.25 | 0.156 | 0.138 | 0.326 | 0.664 | 0.232 | |
| InternVL-2.5-8B2025.12 | 0.248 | 0.156 | 0.139 | 0.329 | 0.669 | 0.236 | |
| LUVCBackbone=InternVL-2.5-8B, Speedup=1.612025.12 | 0.243 | 0.154 | 0.137 | 0.33 | 0.666 | 0.235 | |
| VTWBackbone=InternVL-2.5-26B, Speedup=1.612025.12 | 0.242 | 0.149 | 0.132 | 0.321 | 0.601 | 0.224 | |
| PACTBackbone=InternVL-2.5-26B, Speedup=1.592025.12 | 0.24 | 0.149 | 0.134 | 0.32 | 0.63 | 0.229 | |
| PACTBackbone=InternVL-2.5-8B, Speedup=1.512025.12 | 0.231 | 0.144 | 0.133 | 0.32 | 0.602 | 0.23 | |
| VTWBackbone=InternVL-2.5-8B, Speedup=1.652025.12 | 0.203 | 0.119 | 0.115 | 0.291 | 0.448 | 0.191 |