Single decision accuracy varied across models in the Frenos study, with models correctly matching required outcomes in 6% to 47% of single action choices.