Evo2 genome LM designs novel antibodies, rubisco and CRISPR systems via ICL
KevinKaichuang · x · 2026-11-04
Researchers show the Evo2 genome language model performing in-context learning to de novo design lanmodulins, rubisco, anti-CRISPR proteins, epitope-specific antibodies, CRISPR–Cas systems, and even mitochondrial and bacteriophage genomes. The figures show vLLM inference up to 35x faster on long prompts (up to 131k nt), filtering yield scaling with in-context examples, and designed rubisco reaching pLDDT 89.4 with natural-like Mg-binding geometry.
More from Research
- STOC submissions include dueling P=NP and P!=NP proofs; blogger bets both wrong — thegautamkamath · 2026-11-04
- CoworkOS Debuts: Three-Layer Agent Coworker Framework Lifts Qwen3.6 by 9.76 Points — KyeGomezB · 2026-11-04
- STOC 2027 draws 1665 submissions and publishes all of them, with author names, immediately — thegautamkamath · 2026-11-04
- DOLORES: Training-Free Meta-Scaffold Lifts 8B Models Past 32B Baselines — StellaLisy · 2026-11-04
- Formalizing PTX in Lean to formally verify GPU kernels written by AI agents — gharik · 2026-11-04
- Open-weight LLMs control robots zero-shot: Qwen scores 31.39 on RoboDojo, beating GPT-6 Astra — pabbeel · 2026-11-04