Evo 2 Shows Emergent In-Context Learning: Few-Shot Prompts Steer Design of Enhancers, Antibodies, Phage Genomes
BrianHie · x · 2026-10-29
Brian Hie's team reports a surprising finding: prompting Evo 2 with a handful of sequences sharing a function steers the model to generate new sequences with similar function — a form of generative in-context learning that emerged unanticipated during training.
- Study 1 (led by Alex Hao): few-shot design of mammalian regulatory elements, including novel enhancers with cell-type specificity for a neuronal subtype in the mouse brain
- Study 2 (led by garykbrixi and Alex Hao): extends to computational design of natural proteins, de novo antibodies, protein-RNA systems, and complete bacteriophage genomes
- Released new artifacts including a highly efficient inference stack
Related event: Evo 2 Shows In-Context Learning for Genome Design(2 posts)→
More from AGI Musings
- DeepMind paper: moral argumentation training makes LLMs decisive but no better at moral judgment — rohanpaul_ai · 2026-10-30
- "The newsletter is dead": Every's CEO says AI agents will read it before subscribers do — danshipper · 2026-10-30
- Ex-Anthropic researcher launches Mirendil, 'what RSI means,' praised by Jack Clark — bneyshabur · 2026-10-30
- Agents Formalize Math Theorems in a Week: $135k in Tokens, 600k Lines of Lean — bneyshabur · 2026-10-30
- Healthcare CTO on AI agents for insurer enrollment: demos are easy, production is brutal — alex_verem · 2026-10-30
- Gary Marcus: the smartest people were wrong about the Metaverse too — GaryMarcus · 2026-10-30