1 paper
Dezhao Song, Guglielmo Bonifazi, Frank Schilder +1
LLM post-training has primarily relied on large text corpora and human feedback, without capturing the structure of domain knowledge. This has caused models to struggle dealing wit…