1 articles with this tag
Frontier AI agents can automate AI research engineering but fail to make substantial progress on core research questions, according to new shadow evaluations.
