{
  "title": "Verified Qwen3-Omni LoRA 128-Episode Held-Out Result",
  "status": "verified_full_128_episode_diagnostic_result",
  "status_date": "2026-06-07",
  "backbone": "Qwen/Qwen3-Omni-30B-A3B-Instruct",
  "adapter": "Qwen3-Omni LoRA",
  "dataset": "Ropedia Xperience-10M selected 128-episode pilot",
  "split_policy": {
    "unit": "episode",
    "selected_episode_counts": {
      "train": 96,
      "val": 16,
      "test": 16
    },
    "exported_window_counts": {
      "train": 2848,
      "val": 512,
      "test": 448
    },
    "exported_episode_counts": {
      "train": 89,
      "val": 16,
      "test": 14
    },
    "skipped_selected_episodes": 9,
    "leakage_policy": "Train, validation, and test are separated by episode/session; test windows are used only for held-out evaluation."
  },
  "training": {
    "num_processes": 8,
    "epochs": 2,
    "lora_rank": 16,
    "lora_alpha": 32,
    "lora_dropout": 0.05,
    "num_train_samples": 2848,
    "num_val_samples": 512,
    "history": [
      {
        "epoch": 1,
        "train_loss": 0.41282760031950355,
        "val_loss": 0.03288277983665466,
        "global_step": 356
      },
      {
        "epoch": 2,
        "train_loss": 0.027745448225544075,
        "val_loss": 0.027823254466056824,
        "global_step": 712
      }
    ],
    "loss": "answer-token cross entropy over supervised JSON tokens",
    "note": "This final Qwen3-Omni LoRA pass reused the selected 96/16/16 episode setup, trained on all exported train windows with validation monitoring, and preserved the held-out test split for final evaluation."
  },
  "evaluation": {
    "split": "test",
    "num_samples": 448,
    "held_out_episode_count": 14,
    "json_validity_rate": 0.9977678571428571,
    "action_macro_f1": 0.0024331644885523347,
    "subtask_accuracy": 0.002232142857142857,
    "transition_accuracy": 0.9709821428571429,
    "next_action_accuracy": 0.029017857142857144,
    "contact_accuracy": 0.71875,
    "object_micro_f1": 0.30160427807486634,
    "quality_target": {
      "json_validity_rate": 0.98,
      "status": "met"
    },
    "previous_validation_aware_json_validity_rate": 0.875
  },
  "interpretation": "This is the final verified two-epoch Qwen3-Omni LoRA diagnostic result for the selected 128-episode setup. It meets the 98% JSON-validity target and improves transition, contact, and object metrics over the earlier validation-aware pilot, but action and subtask classification remain weak on held-out episodes, so this is still a baseline-quality diagnostic model rather than a strong Xperience-10M action recognizer.",
  "public_package": {
    "path": "results/omni_finetune/verified_public/xperience10m_qwen3_omni_128ep_structured_json_v2_reuse_full8gpu_lora_eval_test_full",
    "audit_status": "pass",
    "contains_raw_xperience10m_data": false,
    "contains_qwen_base_weights": false,
    "contains_lora_weights": false,
    "adapter_weights_repo": "cy0307/ropedia-qwen3-omni-lora-128ep"
  },
  "required_next_steps": [
    "Verify the public Hugging Face LoRA adapter repository hashes after publication.",
    "Publish the final verified package and refreshed comparison tables to all public mirrors, then run live publication verification.",
    "Use the full-eval predictions for error analysis focused on action/subtask confusions and unseen-label behavior.",
    "Keep the same verified package contract for the Cosmos3 world-model branch and any future VLA/policy branches."
  ]
}