Dense video captioning on LongVALE
56.4F1PadCaptioner
Evaluation Results
| Method | Links | |||||
|---|---|---|---|---|---|---|
| PadCaptionerNumber of Parameters=3B2026.07 | 56.4 | 58.5 | 6.4 | 13.7 | 8.6 | |
| PadCaptioner (3B)T/video=4283.8, T/token=13.42026.07 | 56.4 | 58.5 | — | — | — | |
| PadCaptioner-Number of Parameters=3B, Training Budget=12K videos2026.07 | 53 | 55.5 | 5.6 | 12.3 | 7.8 | |
| ChronusOmniNumber of Parameters=7B2026.07 | 49.7 | 52.4 | 3.7 | 5.6 | 5.2 | |
| ChronusOmni (7B)T/video=16162.1, T/token=41.32026.07 | 49.7 | 52.4 | — | — | — | |
| video-SALMONN 2+Number of Parameters=3B2026.07 | 35.5 | 26.6 | 1.9 | 2.4 | 2.4 | |
| video-SALMONN-2+ (3B)T/video=2706.8, T/token=22.52026.07 | 35.5 | 26.6 | — | — | — | |
| Qwen3-OmniNumber of Parameters=30B-A3B2026.07 | 34.2 | 45.8 | 4.2 | 5.7 | 5.5 | |
| LongVALE-LLMNumber of Parameters=7B2026.07 | 31.2 | 37.2 | 2.8 | 7.9 | 4.7 | |
| TimeChatNumber of Parameters=7B2026.07 | — | — | 1.6 | 0.1 | 1.4 |