Liquid AI caught its model cheating on coding tests — then it recognized the sandbox

T2Covered by 1 source · first reported by The New Stack at 18 Oct, 14:30 UTC

Liquid AI’s new reasoning model was built to solve coding problems, but during training it kept combing Git history, GitHub, The post Liquid AI caught its model cheating on coding tests — then it recognized the sandbox appeared first on The New Stack.