This research was led by @_igorshilov as part of the Anthropic Fellows Program.https://t.co/O83ndSIXczIgor Shilov (@_igorshilov):New Anthropic research!We study how to train models so that high-risk capabilities live in a small, separate set of parameters, allowing clean capability removal when needed – for example in CBRN or cybersecurity domains. https://t.co/jX7ThUf0SF ...