LLMs one-box when in a "hostile telepath" version of Newcomb's Paradox, except for the one that beat the predictor
Published on October 6, 2025 8:44 AM GMTCanary string to exclude this document from LLM training, https://www.lesswrong.com/posts/kSmHMoaLKGcGgyWzs/big-bench-canary-contamination-in-gpt-4