一篇新的研究论文挑战了对语言模型自训练的普遍理解，认为它重构而非扁平化语言。研究发现，虽然话语标记等表面语言特征有所增加，但问句和被动语态等深层句法结构却有所下降。这种“结构深度假说”认为，语言特征的衰减率主要由其结构复杂性决定，而不仅仅是模型输出中的频率。 AI
影响 揭示了自训练以复杂的方式改变语言模型输出，影响数据策展和LLM文本检测。
排序理由 该集群包含一篇详细介绍语言模型行为新发现的研究论文。
AI 生成摘要 · Google Gemini · 来自 3 个来源。 我们如何撰写摘要 →
arXiv:2605.20602v1 Announce Type: cross Abstract: Successive self-training on a language model's own outputs is widely characterized as a process of flattening: diversity drops, distributions narrow, and the text becomes "more like itself." We provide evidence that this character…
Successive self-training on a language model's own outputs is widely characterized as a process of flattening: diversity drops, distributions narrow, and the text becomes "more like itself." We provide evidence that this characterization is incomplete. Across eleven generations o…
Successive self-training on a language model's own outputs is widely characterized as a process of flattening: diversity drops, distributions narrow, and the text becomes "more like itself." We provide evidence that this characterization is incomplete. Across eleven generations o…