1 paper
Shitong Wang, Zhongang Cai, Yuzhou Hong
Next-latent prediction fits a map from the current embedding to the next one. LeNEPA carries this objective to time series, replacing the stop-gradient of next-embedding prediction…