Video Reconstruction on UCF-101 (test)
8.6rFVDMAGVIT-v2-MLM
Evaluation results
| Method | Links | |||
|---|---|---|---|---|
| MAGVIT-v2-MLMGenerator Parameters=307M, Codebook Size=262144, Number of Tokens=1280, Generative Model Type=MLM generative models2025.12 | 8.6 | — | — | |
| MAGVIT-v2-ARGenerator Parameters=840M, Codebook Size=262144, Number of Tokens=1280, Generative Model Type=AR generative models2025.12 | 8.6 | — | — | |
| DeRATokenizer Parameters=174M, Generator Parameters=632M, Codebook Size=8192, Number of Tokens=1024, Generative Model Type=AR generative models2025.12 | 15 | — | — | |
| Open-MAGVIT2-V-UCFToken Type=2D, Tokens=5 x 16 x 16, Ratio=8, Train Resolution=128 x 128, Codebook Size=2621442024.09 | 16 | — | 100 | |
| SweetTokenizer (no token compression)Token Type=1D, Tokens=5120, Train Resolution=256 x 256, Codebook Size=10481 + 111392024.09 | 18 | — | — | |
| LARP-L-LongToken Type=1D, Tokens=1024, Train Resolution=128 x 128, Codebook Size=81922024.09 | 20 | — | — | |
| LARPTokenizer Parameters=173M, Generator Parameters=632M, Codebook Size=8192, Number of Tokens=1024, Generative Model Type=AR generative models2025.12 | 20 | — | — | |
| LARP-LToken Type=1D, Tokens=1024, Train Resolution=128 x 128, Codebook Size=81922024.09 | 24 | — | — | |
| MAGVITToken Type=2D, Tokens=4 x 16 x 16, Ratio=8, Train Resolution=128 x 128, Codebook Size=10242024.09 | 25 | — | — | |
| MAGVIT-MLMTokenizer Parameters=158M, Generator Parameters=306M, Codebook Size=1024, Number of Tokens=1024, Generative Model Type=MLM generative models2025.12 | 25 | — | — | |
| MAGVIT-ARTokenizer Parameters=158M, Generator Parameters=306M, Codebook Size=1024, Number of Tokens=1024, Generative Model Type=AR generative models2025.12 | 25 | — | — | |
| AdapTokTokenizer Parameters=195M, Generator Parameters=633M, Codebook Size=8192, Number of Tokens=2048, Generative Model Type=AR generative models2025.12 | 28 | — | — | |
| OmniTokenizerTokenizer Parameters=82.2M, Generator Parameters=650M, Codebook Size=8192, Number of Tokens=1280, Generative Model Type=AR generative models2025.12 | 42 | — | — | |
| SweetTokenizerToken Type=1D, Tokens=256 + 1024, Train Resolution=256 x 256, Codebook Size=10481 + 111392024.09 | 44 | — | — | |
| ElasticTokTokenizer Parameters=200M, Codebook Size=64000, Number of Tokens=2048, Generative Model Type=AR generative models2025.12 | 93 | — | — | |
| TATSToken Type=2D, Tokens=4 x 16 x 16, Ratio=8, Train Resolution=128 x 128, Codebook Size=163842024.09 | 162 | — | — | |
| TATSTokenizer Parameters=32M, Generator Parameters=321M, Codebook Size=16384, Number of Tokens=1024, Generative Model Type=AR generative models2025.12 | 162 | — | — |