Metadata Conditioning Accelerates Language Model Pre-training

arxiv.org

saved by: 0 people
related read on: A small number of samples can poison LLMs of any size from the discussion