This paper fine-tunes inverse-folding model with feedbacks from folding model through DPO.

We develop a novel speculative decoding framework for protein generation, using structure aware guidance from k-mers to generate proteins with higher likelihood and structural confidence.