training_stats.json

{
  "model": "Qwen/Qwen2.5-0.5B-Instruct",
  "num_episodes": 100,
  "start_avg_ep1_10": 0.309,
  "end_avg_ep91_100": 0.5962,
  "improvement": 0.2872,
  "eval_scores": {
    "task_1_basic_antipatterns": 0.75,
    "task_2_correlated_subqueries": 0.8313,
    "task_3_wildcard_scan": 0.925,
    "task_4_implicit_join": 0.6438,
    "task_5_window_functions": 0.625
  },
  "average_eval_score": 0.755,
  "vs_baseline": 0.125
}