Video Retrieval on DiDeMo (test)
60R@1mPLUG2+
Evaluation Results
| Method | Links | ||||
|---|---|---|---|---|---|
| mPLUG2+Quantization=None, Weight precision=32-bit floating point, Text decoder=Llama2025.12 | 60 | 90.1 | — | 86.2 | |
| mPLUG2+Quantization=8-bit integer, Weight precision=8-bit integer, Text decoder=Llama2025.12 | 58.7 | 88.4 | — | 83.1 | |
| mPLUG2Quantization=None, Weight precision=32-bit floating point, Text decoder=BERT2025.12 | 55.1 | 84.6 | — | 78.3 | |
| mPLUG2Quantization=8-bit integer, Weight precision=8-bit integer, Text decoder=BERT2025.12 | 53.7 | 80.8 | — | 75.1 | |
| HEROInput Modality=video + ASR2020.05 | 3.01 | 14.87 | 47.26 | — | |
| HEROInput Modality=video-only2020.05 | 2.14 | 11.43 | 36.09 | — | |
| XMLInput Modality=video-only2020.05 | 1.59 | 6.71 | 25.44 | — |