A new study reveals that large language models (LLMs) significantly distort the meaning and conclusions of written text, even when prompted for minor edits like grammar correction. Researchers found that LLM-generated revisions introduced larger semantic shifts than human edits and steered writing in a consistent direction away from human style. This distortion was observed across a user study, argumentative essays, and even peer reviews from a top AI conference, suggesting potential broad impacts on communication, science, and culture. AI
影响 LLM-assisted writing may subtly alter scientific discourse and human communication, necessitating careful review of AI-generated content.
排序理由 Academic paper detailing novel findings on LLM behavior.
- claude-haiku
- gemini-2.5-flash
- gpt-5-mini
- Ibrahim Qureshi
- International Conference of Learning Representations
- Isadora White
- Joel Z. Leibo
- LLMs
- Marwa Abdulhai
- Max Kleiman-Weiner
- Natasha Jaques
AI 生成摘要 · Google Gemini · 来自 1 个来源。 我们如何撰写摘要 →
