模型12:30BioSecBench-Surveillance:病原体基因组监测AI代理可验证基准想看看AI在基因组监测上多不靠谱?新基准BioSecBench-Surveillance测了16个模型,最好才一半正确,细节值得细读。#BioSecBench-Surveillance#Opus 4.8#GPT-5.5#Sonnet 4.61 个信源在谈事件专题aarXiv cs.AI@Harmon Bhasin 等 12 人2 个信源在谈原文稍后读已读值得跟进有用关注 BioSecBench-Surveillance