Mechanism Discovery Benchmark

The active implementation follows proposal.md and scores observable behavior and independent internal-state probes objectively. Fundamentality and DAG recovery metrics are not part of this benchmark.

• Benchmark workflow
• Example
• API reference
  • VariableSpec
  • MechanismItem
  • MechanismProbe
  • Task
  • ValidationError
  • expression_tree()
  • parse_expression()
  • split_equation()
  • parse_equation()
  • symbols_for()
  • solve_model()
  • expand_expression()
  • task_from_dict()
  • TaskLoader
  • load_task()
  • validate_sampling()
  • parse_unit()
  • check_units()
  • validate_task()
  • discover_tasks()
  • get_parser()
  • main()
  • evaluate_expression()
  • generate_synthetic_data()
  • get_parser()
  • main()
  • task_to_dict()
  • public_problem()
  • export_task()
  • get_parser()
  • main()
  • FeedbackCache
  • parse_upload()
  • create_server()
  • get_parser()
  • main()
  • load_answer()
  • numerical_constant_equations()
  • numerical_literal_context()
  • format_probe()
  • expand_probe_reply()
  • evaluate()
  • get_parser()
  • build_argparser()
  • sanitize_filename()
  • feedback_service()
  • run_experiment()
  • main()
  • update_parser()
  • run()
  • get_ask()