This is an observed behaviour in large models, which tend towards “sycophancy” as they scale. https://www.anthropic.com/news/towards-understanding-sycophancy-in

This is an observed behaviour in large models, which tend towards “sycophancy” as they scale. https://www.anthropic.com/news/towards-understanding-sycophancy-in-language-models https://www.anthropic.com/news/towards-understanding-sycopha...