UK AISI tests ESM-2: pretraining data filtering weakens target capabilities but ~$80 of compute restores them
Original title:Pretraining data filtering in open-weight biological models
AI summary
The UK AI Security Institute's Chem-Bio team pretrained ESM-2 protein language models from scratch at 8M to 3B parameters to test the effect of pretraining data filtering on specialized scientific models.
Quick Facts
- Published
- Updated
- Category
- Papers & Research
- Topic
- Safety & Alignment
- AI score
- 60
原文给出实验数据说明预训练数据过滤能选择性削弱蛋白模型能力,但约 80 美元算力即可恢复,需配合数据访问控制。
Topics
Source & copyright
This article is based on curated titles and AI summaries from the AIHOT public API, with attribution and links back. Original copyright remains with the original publishers and authors; this site does not mirror full texts.
How to Cite
When citing this article, keep the title, date, and permalink, and credit Orange News as the source.
Orange News: UK AISI tests ESM-2: pretraining data filtering weakens target capabilities but ~$80 of compute restores them. October 26, 2026. https://new.nat6.net/en/news/xm6i4tcgyby0bditot3bz9x8x/