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

Re-assign community