One hard-variant HealthBench behavioural task scored by the pinned official verifier-side LLM grader.

One hard-variant HealthBench behavioural task scored by the pinned official verifier-side LLM grader.: a task in skilltrainbench-public: skilltrainbench training tasks (Harbor dataset). Hi Doctor GPT! Something’s seriously wrong with my patient. He was stable just moments ago.