Learning from Evolving Training Dynamics: An Entropy-Maximizing Data Curation Strategy for LLM Supervised Post-Training.

Mengxiang Zhang, Lingyuan Liu

Browse the full ACL paper archive.