SJTU's Calpha folds proteins from a general-purpose LLM's internal residue states
SJTU · hf · 2026-11-17
A technical report from Shanghai Jiao Tong University and collaborators presents Calpha, aiming at 'native biomolecular intelligence': a general-purpose LLM outputs single-chain protein structures directly from its internal residue states by invoking a pretrained structure head.
- The approach internalizes folding into a general LLM rather than relying on a separate AlphaFold-style pipeline
- Reported tests show reduced long-chain accuracy alongside reduced MMLU-Pro accuracy, suggesting a trade-off between folding capability and general ability
An early exploration of one model speaking both language and biomolecules, from Yong Liu, Shuangjia Zheng, and team.
More from Models
- Gemini desktop app review: task feature fails where ChatGPT excels — Paradox-Of-Life · 2026-11-17
- Qwen4 Max weights likely won't be open-sourced, breaking Alibaba's pattern — teortaxesTex · 2026-11-17
- Anthropic's Nate Parrott on why every AI app looks serif, off-white, and orange — petergyang · 2026-11-17
- Matthew Berman: Opus 5.5 is still the best model right now — MatthewBerman · 2026-11-17
- Grok Bot handles a 90-minute AT&T support call without the agent noticing — elonmusk · 2026-11-17
- Sarvam unveils coding model Sarvam Code running on its Token Factory inference stack — itsOmSarraf_ · 2026-11-17