学习蛋白质折叠能否泛化到更广泛的推理？
原标题：Paper page - Does Learning Protein Folding Generalize to Broader Reasoning?
AI 摘要
研究者构建了 FoldingCorpus 蛋白质问答数据集与 Fold2Reason 后训练方法，通过离散结构答案和连续 3D 几何两种信号进行训练。在 FoldBench 上，Fold2Reason 的结构预测得分达到 Qwen3.5-9B 的 2.7 至 3.5 倍，并在空间、图、科学和通用推理等全部 10 个基准上均有提升，宏平均准确率从 45.09% 升至 48.33%。对照实验表明增益来自蛋白质几何中的真实结构信息，而非单纯增加训练数据。
正文节选
Does Learning Protein Folding Generalize to Broader Reasoning? Abstract Large language models rely heavily on human text, which often conveys surface answers rather than the spatial and structural logic behind them. Protein folding is a natural testbed, because one solved structure yields thousands of exactly checkable spatial and topological statements. We ask: can learning to fold proteins teach general models reusable reasoning capabilities? To answer this, we build FoldingCorpus, a protein-d