Muhammad Usman Altaf

Research questions rank measures for language model fine-tuning updates

A new research paper published on arXiv questions the common practice of using rank measures to evaluate fine-tuning updates in language models. The study introduces noise baselines and harder references, such as random…