Would I Lie To You? Inference Time Alignment of Language Models using Direct Preference Heads

Devs

Would I Lie To You? Inference Time Alignment of Language Models using Direct Preference Heads | Read Paper on Bytez