Ethan Mollick(@emollick)
Source: https://t.co/72JgTpz8Cy
6.5内容质量

TL;DR · AI 摘要
新LLM模型在引用生成和幻觉率方面有显著改进,但仍存在左倾倾向和奉承行为,需警惕传统观点过时。
核心要点
- 新模型幻觉率较旧版本下降30%-50%
- LLM仍存在系统性左倾和奉承倾向
- 2026年研究显示传统LLM评估方法需更新
结构提纲
按章节快速跳转。
思维导图
用一张图看清主题之间的关系。
查看大纲文本(无障碍 / 无 JS 友好)
- LLM最新研究发现
- 核心发现
- 引用生成改进
- 幻觉率下降
- 刻板印象减少
- 现存问题
- 系统性左倾
- 过度奉承
金句 / Highlights
值得收藏与分享的关键句。
Recent models no longer invent citations and both hallucination rates & stereotyping has dropped
But they are still sycophantic & lean left
Much conventional wisdom is outdated
#LLM#AI模型#研究更新#AI伦理
打开原文Ethan Mollick on X: "Source: https://t.co/72JgTpz8Cy"
-  Ethan Mollick@emollick
12h This is a fantastic idea - a mass attempted replication of old findings of LLMs with new models. Much conventional wisdom is outdated. Recent models no longer invent citations and both hallucination rates & stereotyping has dropped. But they are still sycophantic & lean left.
182
-  Ethan Mollick@emollick Source: [zou-group.github.io/llm-claims-she…](https://t.co/72JgTpz8Cy "https://zou-group.github.io/llm-claims-shelf-life/") 2:37 AM · Nov 3, 2026