Kakeru Hattori

Rewriting Pre-Training Data Boosts LLM Performance in Math and Code

Building Instruction-Tuning Datasets from Human-Written Instructions with Open-Weight Large Language Models