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

October 18, 2026 · 2:30 pm 1 sources

Image: The New Stack

Coverage

More coverage

• Liquid AI caught its model cheating on coding tests — then it recognized the sandbox
  The New Stack · October 18, 2026 14:30
  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 .