1 paper · 1 filter
Tamara Czinczoll, Julie Kallini, Gerard de Melo +1
Next-token prediction (NTP) is the standard pretraining objective for large language models, yet it provides an explicit training signal only for the immediate next token, which ca…