English(EN) When transformers learn "impossible" languages, what do they learn?

Transformer 因生成失败而难以处理“不可能”的语言

作者 PulseAugur 编辑部 · [2 个来源] · 2026-06-29 18:42

一篇新的研究论文探讨了为什么像 GPT-2 这样的 Transformer 语言模型难以处理人类可以习得的“不可能”语言。研究发现，虽然这些模型对语法敏感度有所体现，但在生成高质量、长句方面却表现出明显的不足。这表明，生成失败，而非语法不敏感，可能是这些模型无法处理这种非自然语言的主要原因。 AI

影响表明当前 Transformer 架构在处理复杂语言结构方面存在局限性，可能指导未来的模型开发。

排序理由发表在 arXiv 上的研究论文，详细介绍了 Transformer 语言模型的发现。

AI 生成摘要 · Google Gemini · 来自 2 个来源。我们如何撰写摘要 →

本文如何被排名

Signal score: 0 / 100; Composite score across the factors below. Higher = stronger signal that this story matters right now.
Newsworthiness bucket: Research; 发表在 arXiv 上的研究论文，详细介绍了 Transformer 语言模型的发现。
Source corroboration: 2 independent sources; Multiple independent publishers reporting the same story raises confidence that it's real and newsworthy.
Topics: paper, model release; Editorial topic classification. Feeds into how the story surfaces on /topic/<slug> hub pages and into the per-entity coverage mix.
AI-industry relevance: High; Clearly on-topic for AI-industry coverage.
Story freshness: 101 days old; Aged out of breaking-news scoring windows; ranking reflects the durable signal from the full source set.

完整方法见我们的编辑标准。

arXiv cs.AI TIER_1 English(EN) · Ram Janarthan, Coleman Haley, Sharon Goldwater · 2026-07-01 04:00

当 Transformer 学会“不可能”的语言时，它们学到了什么？

arXiv:2606.30815v1 Announce Type: cross Abstract: Recent work suggests that transformer language models show a bias towards human languages over unnatural ("impossible") languages argued to be unacquirable by humans. However, this literature has largely based these claims on diff…
arXiv cs.CL TIER_1 English(EN) · Sharon Goldwater · 2026-06-29 18:42

当 Transformer 学会“不可能”的语言时，它们学到了什么？

Recent work suggests that transformer language models show a bias towards human languages over unnatural ("impossible") languages argued to be unacquirable by humans. However, this literature has largely based these claims on differences in sample efficiency and test-set perplexi…