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