[ndr_]

Yes. OpenAI's GPT-OSS was training using Deliberative Alignment (which was found to be flawed in a competition on Kaggle, but still).

https://arxiv.org/abs/2412.16339