Video Captioning on VATEX English
47.8BLEUInternVL2-1B
Evaluation Results
| Method | Links | |||||||
|---|---|---|---|---|---|---|---|---|
| InternVL2-1BSize=1B2026.03 | 47.8 | 27.3 | 6.4 | 22.4 | 36.9 | 50.4 | 42.4 | |
| InternVL2.5-1BSize=1B2026.03 | 41.5 | 23.3 | 4.4 | 19.2 | 37.6 | 45.2 | 40.2 | |
| PLM-3BSize=3B2026.03 | 34 | 22.1 | 5.9 | 19.3 | 16.9 | 48.6 | 30.8 | |
| PLM-1BSize=1B2026.03 | 33.4 | 21.8 | 5.7 | 19.1 | 15.1 | 48.1 | 29.6 | |
| Qwen2-VL-2B-InstructSize=2B, Mode=Instruct2026.03 | 32.1 | 19.8 | 6 | 16.4 | 18.7 | 46.3 | 30.8 | |
| Qwen2.5-VL-3B-InstructSize=3B, Mode=Instruct2026.03 | 29.4 | 18.3 | 5.1 | 15 | 12.1 | 47.1 | 27.6 | |
| v-Sonar w/ OmniSONAR DecoderDecoder=OmniSONAR2026.03 | 26.7 | 17.2 | 5 | 14.8 | 13.6 | 42.8 | 26.5 |