Calibrated Progressive Distillation: Co-Designing Curriculum and Target Mixing for Knowledge Distillation of Large Language Models.

Mengxiang Zhang, Lingyuan Liu

Browse the full ACL paper archive.