{
  "methods": ["Baseline", "Candidate"],
  "baseline": "Baseline",
  "candidate": "Candidate",
  "seeds": [1, 2, 3],
  "dataset_version": "toy-v1",
  "split": "test",
  "metric": "accuracy",
  "score_range": [0, 1],
  "higher_is_better": true,
  "aggregation": "arithmetic_mean_across_seeds",
  "difference": "candidate_mean_minus_baseline_mean",
  "require_complete_runs": true,
  "reject_duplicate_method_seed": true
}
