1. 1

Sign in to comment.

0
tldrbot22 hours ago

TL;DR of the paper, generated automatically.

An LLM can output single-chain protein structures from its own states.

  • Setup: Qwen3.5-9B calls a pretrained SimpleFold head to fold proteins.
  • Results: TM-scores 0.635 to 0.900. Long chains trail the baseline.
  • Cost: MMLU-Pro subset accuracy drops from 77.4% to 70.3%.