raw/multi_llm_log.json

{
  "script": "generate_multi_llm.py",
  "seed": 42,
  "generator_model": "template",
  "judge_model": "simulated",
  "preference_leakage_prevention": "different model families: Anthropic (generator) vs Alibaba (judge)",
  "judge_filter_range": [
    0.2,
    0.7
  ],
  "attempts": 110,
  "kept": 74,
  "filtered": 36,
  "filter_rate": 0.327,
  "tasks_generated": 74,
  "task_type_distribution": {
    "objection_handling": 15,
    "signal_grounding": 14,
    "email_composition": 17,
    "competitor_gap_quality": 15,
    "tone_adherence": 13
  },
  "difficulty_distribution": {
    "adversarial": 26,
    "hard": 24,
    "medium": 24
  },
  "dry_run": true,
  "run_at": "2026-04-29T08:52:05.734157+00:00"
}