my understanding is that Deliberative Alignment was only applied to o-series models, not GPT-4o, which has been the most sycophantic and problematic model
my understanding is that Deliberative Alignment was only applied to o-series models, not GPT-4o, which has been the most sycophantic and problematic model