Tasks

TrialDesignBench tasks are organized around the parts of clinical trial design where agentic systems must integrate evidence, constraints, and structured outputs.

Task Families

  • Protocol concept development.
  • Eligibility criteria and safety exclusions.
  • Endpoint selection and estimand framing.
  • Sample size and operating characteristic reasoning.
  • Adaptive and group sequential design choices.
  • Operational feasibility and site constraints.
  • Risk review and protocol documentation.

Example Task Template

Each task page can include the clinical scenario, input materials, agent instructions, expected artifacts, grading checkpoints, and notes for reproducibility.

GitHub Issues