{
  "created_at": "2026-07-15 00:59:15",
  "name": "v39_v29_curriculum_r0_to_r3_lr6e7_ep08x2",
  "base": "/home/ll/llm4rec/experiments/outputs/v29_v19_user_world_guard_lr8e7_ep018",
  "dataset": "v39_curriculum_r3",
  "dataset_manifest": {
    "name": "v39_curriculum_r3",
    "path": "/home/ll/llm4rec/experiments/data/v39_curriculum_r3.jsonl",
    "records": 70000,
    "bytes": 263483139,
    "sha256": "fb38ffd99b1ed61bbca834a85af32e3bb557aeba6a8b05f134ab5b99c16a22da",
    "requested_components": {
      "general": 22000,
      "r0": 6000,
      "r3_think": 16000,
      "r3_direct": 18000,
      "old_rec": 3000,
      "old_user": 2500,
      "old_item": 2500
    },
    "groups": {
      "general": 22000,
      "r3": 34000,
      "old_rec": 3000,
      "old_user": 2500,
      "r0": 6000,
      "old_item": 2500
    },
    "variants": {
      "official_general_direct": 10947,
      "profile_live_direct": 4236,
      "profile_video/ad_cot": 3722,
      "profile_live_cot": 3715,
      "original_competition": 3943,
      "user_strict_logic": 357,
      "profile_video/video_direct": 5603,
      "caption_to_sid_direct": 2978,
      "profile_video/video_cot": 4939,
      "profile_goods_cot": 3624,
      "official_general_cot": 11053,
      "profile_video/ad_direct": 4171,
      "profile_goods_direct": 3990,
      "rec_no_think_direct_final": 1484,
      "sid_to_caption_cot": 1530,
      "sid_to_caption_direct": 1492,
      "item_no_think_direct_final": 429,
      "user_array_json_strict": 402,
      "item_short_think": 417,
      "user_logic_json_strict": 364,
      "user_strict_array": 429,
      "user_extra_no_think_logic": 175
    },
    "routes": {
      "no_think": 38178,
      "think": 31822
    },
    "target_leakage": 0,
    "seed": 202607157
  },
  "learning_rate": "5.0e-7",
  "epochs": 0.8,
  "stages": [
    {
      "dataset": "v39_curriculum_r0",
      "lr": "6.0e-7",
      "epochs": 0.8
    },
    {
      "dataset": "v39_curriculum_r3",
      "lr": "5.0e-7",
      "epochs": 0.8
    }
  ],
  "note": "Two-stage curriculum: 0.8 epoch R0/general alignment followed by 0.8 epoch R3/general cognition.",
  "cot_policy": "Preserve meaningful CoT for /think and pure final outputs for /no_think.",
  "checkpoint_policy": "save_strategy=no; only the final model is retained.",
  "script": "/home/ll/llm4rec/experiments/run_v34_v43.py",
  "script_sha256": "fbf01773c61579cf6159f64442fc0c8902714c84ed23b488bc9a6236a154f402",
  "configs": [
    "/home/ll/llm4rec/experiments/configs/v39_v29_curriculum_r0_to_r3_lr6e7_ep08x2_stage1.yaml",
    "/home/ll/llm4rec/experiments/configs/v39_v29_curriculum_r0_to_r3_lr6e7_ep08x2_stage2.yaml"
  ],
  "reproduce": "/home/ll/llm4rec/demo/LLaMA-Factory/.venv/bin/python /home/ll/llm4rec/experiments/run_v34_v43.py --single v39_v29_curriculum_r0_to_r3_lr6e7_ep08x2 --gpu <gpu>"
}