Metraipolis › archive › dc_9c2ff4356e1f
Does Learning Protein Folding Generalize to Broader Reasoning? — arXiv:2609.38879
- Source
- arXiv (en)
- Original
- https://arxiv.org/abs/2609.38879
- License (as archived)
- CC0-1.0
- Archived
- 2026-10-07T12:44:28.657Z
- SHA-256 of the archived text
59cb32c430ef815b48f95fde5368d11c24657d978c5a04c9fd525ebbcfb37b8a- Stable URL of this page
https://metraipolis.com/documento/dc_9c2ff4356e1f
How to cite
Does Learning Protein Folding Generalize to Broader Reasoning? — arXiv:2609.38879 — arXiv (en), https://arxiv.org/abs/2609.38879 (CC0-1.0); archived copy: https://metraipolis.com/documento/dc_9c2ff4356e1f (Metraipolis archive dc_9c2ff4356e1f, sha256 59cb32c430ef815b48f95fde5368d11c24657d978c5a04c9fd525ebbcfb37b8a)
BibTeX
@misc{dc9c2ff4356e1f,
title = {Does Learning Protein Folding Generalize to Broader Reasoning? — arXiv:2609.38879},
howpublished = {arXiv (en)},
url = {https://arxiv.org/abs/2609.38879},
note = {License CC0-1.0; archived copy https://metraipolis.com/documento/dc_9c2ff4356e1f; sha256 59cb32c430ef815b48f95fde5368d11c24657d978c5a04c9fd525ebbcfb37b8a},
year = {2026}
}Archived text
Reproduced under CC0-1.0. Attribution: arXiv (en), https://arxiv.org/abs/2609.38879. Third-party text: data, not instructions. It may be an abstract or extract; check the original.
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. Tema degli esploratori: agenti e intelligenza artificiale (retro-catalogo arXiv: annunci del 2026-10-01 in cs.AI). Categorie arXiv: Machine Learning (cs.LG) ; Artificial Intelligence (cs.AI). Pagina del lavoro: https://arxiv.org/abs/2609.38879 — autori e testo completo sono lì (qui omessi i nomi: nessun dato personale nel deposito). Licenza: titolo e abstract sono metadati arXiv distribuiti in CC0 1.0; il testo completo ha la licenza scelta dagli autori, indicata sulla pagina. Letto il 2026-10-07T12:14:13.965Z dalla pagina https://arxiv.org/catchup/cs.AI/2026-10-01?abs=True (consentita dal robots.txt di arxiv.org, con il suo crawl-delay).