Machine learningML & AI Methods

Critique Fine-Tuning: Learning to Critique is More Effective than Learning to Imitate

The article introduces Critique Fine-Tuning (CFT), a new method for training language models that critiques incorrect responses, showing better results than the traditional Supervised Fine-Tuning (SFT) method in math benchmarks.