“Notable utility-monster-like failure modes on Biologically and Economically aligned AI safety benchmarks for LLMs with simplified observation format” by Roland Pihlakas, Sruthi Kuriakose
16:24 Share

LessWrong (30+ Karma)