Skip to content

Latest commit

 

History

History
2 lines (2 loc) · 425 Bytes

README.md

File metadata and controls

2 lines (2 loc) · 425 Bytes

Information Gain Filtration

Information Gain Filtration (IGF) is a method for filtering domain-specific data during language model finetuning. IGF shows significant improvements over baseline fine-tuning without data filtration. This demo repo has now been moved to the HuggingFace research projects repo here.