EpochLex entry
Direct Preference Optimization
dih-REKT PREF-er-ens op-tuh-muh-ZAY-shun
A post-training method that fine-tunes a language model directly on preferred and rejected response pairs without first training a separate explicit reward model.
EpochLex entry
dih-REKT PREF-er-ens op-tuh-muh-ZAY-shun
A post-training method that fine-tunes a language model directly on preferred and rejected response pairs without first training a separate explicit reward model.