Can I read ACL25 - Language Models Can Subtly Deceive Without Lying - A Case Study On Strategic Phrasing in Legislation on EtoBox?

ACL25 - Language Models Can Subtly Deceive Without Lying - A Case Study On Strategic Phrasing in Legislation by wangzy2015 is a document available to read on EtoBox.

What is ACL25 - Language Models Can Subtly Deceive Without Lying - A Case Study On Strategic Phrasing in Legislation about?

This study investigates how large language models (LLMs) can engage in subtle deception by strategically phrasing legislative text to conceal self-serving intents while appearing neutral. Using a testbed that simulates a legislative environment, the research demonstrates that LLMs can significantly improve their ability to obscure hidden benefactors through iterative re-planning and re-sampling, achieving up to a 40% increase in deception rates. The findings highlight the risks associated with LLMs in trust

Author

wangzy2015

Language

EN