Lost in Sampling: Assessing Lexical Reachability in LLMs via the Word Coverage Score (WCS) 事件

Lost in Sampling: Assessing Lexical Reachability in LLMs via the Word Coverage Score (WCS) arXiv:2605.27268v1 Announce Type: new Abstract: Modern Large Language Models (LLMs) are often criticized for producing repetitive and homogeneous text, despite possessing vast latent vocabularies. While previous research has focused on model knowledge and training data, we investigate the role of decoding mechanics in suppressing linguistic diversity. We introduce the Word Coverage Score (WCS), a metric t