CodecLM: Aligning Language Models with Tailored Synthetic Data

2 points| milliondreams | 1 year ago |arxiv.org