{
 "generated_at": "2026-10-28T08:47:11+00:00",
 "script": "scripts/20_evaluate.py",
 "runs": [
  "baselines",
  "final_E_single_llm_mixed_random_mid",
  "final_F_independent_mixed_random_mid",
  "final_G_debate_r2_mixed_random_mid",
  "final_H_debate_skeptic_r2_mixed_random_mid",
  "final_I_placebo_r2_mixed_random_mid",
  "final_J_cost_matched_mixed_random_mid",
  "final_small_E_single_llm_mixed_random_small",
  "final_small_F_independent_mixed_random_small",
  "final_small_G_debate_r2_mixed_random_small",
  "holdout_E_single_llm_mixed_random_mid",
  "holdout_F_independent_mixed_random_mid",
  "holdout_G_debate_r2_mixed_random_mid",
  "prospective_E_single_llm_mixed_random_mid",
  "prospective_F_independent_mixed_random_mid",
  "prospective_G_debate_r2_mixed_random_mid",
  "prospective_baselines",
  "repeat_G_debate_r2_mixed_random_mid",
  "validation_E_single_llm_all_anthropic_mid",
  "validation_E_single_llm_all_openai_mid",
  "validation_E_single_llm_mixed_random_mid",
  "validation_F_independent_all_anthropic_mid",
  "validation_F_independent_all_openai_mid",
  "validation_F_independent_mixed_random_mid",
  "validation_G_debate_r1_mixed_random_mid",
  "validation_G_debate_r2_all_anthropic_mid",
  "validation_G_debate_r2_all_openai_mid",
  "validation_G_debate_r2_mixed_random_mid",
  "validation_G_debate_r4_mixed_random_mid",
  "validation_G_fullview_r2_mixed_random_mid",
  "validation_H_debate_skeptic_r2_mixed_random_mid",
  "validation_I_placebo_r2_mixed_random_mid",
  "validation_J_cost_matched_mixed_random_mid"
 ],
 "truth": "/Users/simon-pierreboucher/Desktop/Recherche-UQO/immobilier/agentic-valuation-quebec/data/processed/analytic.parquet",
 "split": "validation",
 "n_prediction_rows": 436972,
 "n_truth": 72132,
 "B": 2000,
 "seed": 20261027,
 "main_committee": "mixed_random",
 "main_aggregation": "median",
 "min_cell": 30,
 "arms": [
  "A_hedonic",
  "B_hierarchical",
  "C_catboost",
  "C_gbm",
  "C_xgboost",
  "D_avm",
  "E_single_llm",
  "F_independent",
  "G_debate_r1",
  "G_debate_r2",
  "G_debate_r4",
  "G_fullview",
  "H_debate_skeptic",
  "I_placebo",
  "J_cost_matched"
 ],
 "fake_included": false,
 "note": ""
}