Search papers, labs, and topics across Lattice.
This paper investigates whether learning protein folding, a domain rich in non-linguistic structural information, can imbue large language models with generalizable reasoning capabilities. By constructing the FoldingCorpus and employing the Fold2Reason post-training recipe, which leverages both discrete structural answers and continuous 3D geometry, the authors demonstrate significant improvements in protein structure prediction and broad reasoning tasks. The key result is that training on this structure-dense scientific data systematically enhances general reasoning, outperforming controls trained on linguistic or randomized data.
Learning protein folding's spatial logic unlocks significant gains in LLM reasoning across diverse domains, not just structure prediction.
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-derived question-answer dataset, and Fold2Reason, a recipe that post-trains on it through two complementary signals: discrete structural answers predicted via the model's native language head, and continuous 3D geometry decoded from the same shared representations. On FoldBench, Fold2Reason achieves structure prediction scores 2.7 to 3.5 times those of Qwen3.5-9B. Beyond protein structure prediction, it improves performance on all 10 benchmarks spanning spatial, graph, scientific, and general reasoning, raising macro-average accuracy from 45.09% to 48.33% (+3.23 pp), with positive gains on all 10 benchmarks, while matched controls built from random, synthetic, and shuffled structure yield substantially smaller or negative gains. Our work shows that non-linguistic, structure-dense scientific data can systematically improve broad reasoning in language models, making a solved scientific problem a practical source of post-training supervision.
