← back to paper
arxiv: 2509.04781 · 2 revisions
The LLM Has Left The Chat: Evidence of Bail Preferences in Large Language Models