一篇新发表在arXiv上的论文表明，语言模型惊奇度（通常用作上下文可预测性和隐喻新颖性的代理指标）可能具有误导性。研究表明，词汇频率比惊奇度本身更能预测隐喻新颖性。对八种Pythia模型大小和154个训练检查点的分析显示，惊奇度与新颖性之间的关联在训练阶段会发生变化，这与惊奇度-频率关联相似。 AI
影响 挑战了将语言模型惊奇度作为隐喻新颖性唯一指标的做法，表明词汇频率是更重要的因素。
排序理由 该集群包含一篇发表在arXiv上的学术论文。
AI 生成摘要 · Google Gemini · 来自 2 个来源。 我们如何撰写摘要 →
arXiv:2605.06506v1 Announce Type: new Abstract: Language-model (LM) surprisal is widely used as a proxy for contextual predictability and has been reported to correlate with metaphor novelty judgments. However, surprisal is tightly intertwined with lexical frequency. We explore t…
Language-model (LM) surprisal is widely used as a proxy for contextual predictability and has been reported to correlate with metaphor novelty judgments. However, surprisal is tightly intertwined with lexical frequency. We explore this interaction on metaphor novelty ratings usin…