AI RESEARCH

Cross-Tokenizer Likelihood Scoring Algorithms for Language Model Distillation

arXiv CS.LG

Computing next-token likelihood ratios between two language models (LMs) is a standard task in