1 paper
Tianlv Huang, Hetian Guo, Ziyi Cai +6
Discrete motion representations have substantially advanced autoregressive text-to-motion generation. However, most motion tokenizers are optimized for reconstruction and do not ex…