# Lewis Tunstall highlights few-shot in-context learning for biological design

> Original title: Very cool work! My collaborator @QiuyiLi2 and colleagues explored a related idea in GENERator-v2 under “functional in-context learning.” ...

- Source: Lewis Tunstall
- Published: 2026-10-30T11:32:59.000Z
- Bad Influence: https://badinfluence.news/items/bo1pfc51sqa43peyms7uwai76
- Original: https://x.com/_lewtun/status/2116131247714152695

## Summary

Lewis Tunstall says his collaborator Qiuyi Li and colleagues explored a related idea in GENERator-v2 under functional in-context learning. He calls it interesting to see few-shot biological design taking this further, and links to the bioRxiv preprint.

## Why it matters

The post places a genome language model's few-shot in-context learning result beside a related GENERator-v2 study, showing how the idea is developing in biological design.

## Post

Very cool work! My collaborator @QiuyiLi2 and colleagues explored a related idea in GENERator-v2 under “functional in-context learning.” Interesting to see few-shot biological design taking this further.

https://www.biorxiv.org/content/10.64898/2026.01.27.702015v2.full

## Quoted post @garykbrixi

> Language models can learn new tasks from just a few examples in their prompt, with no retraining. This is called in-context learning (ICL). Now genome language models can do it too.
> 
> With a handful of examples, Evo 2 ICL generates proteins, de novo antibodies, CRISPR-Cas9 loci, mitochondrial genomes and entire bacteriophages.
> 
> We can even use the synthetic sequences iteratively to construct a chain-of-thought (CoT) and achieve harder design goals.
> 
> New work with Yukun A. Hao, @BrianHie and team 🧵

https://x.com/garykbrixi/status/2115832881776091565