Gradient Ascent Post-training Enhances Language Model Generalization.

Dongkeun Yoon, Joel Jang, Sungdong Kim, Minjoon Seo

Browse the full ACL paper archive.