gavin leech:
maybe a way to anchor the models on stable preferences is to treat them as if they had them and get this idea into the training data.
https://x.com/g_leech_/status/2027416051198083142
gavin leech:
https://x.com/g_leech_/status/2027416051198083142