"As a Language Model": Chat Template Switches LLM Self-Referential Voice

An arXiv paper finds chat templates change how LLMs use self-referential voice.

An arXiv paper titled "As a Language Model" examines how chat templates switch large language models' self-referential voice. The listing appeared on Hacker News with 43 points and 31 comments. No model names, benchmark scores, or concrete AI security incident are described.

• Paper studies how chat templates change LLM self-referential voice.
• Posted as arXiv 2609.25021 and discussed on Hacker News.
• Listing shows 43 points and 31 comments, with no security incident.