{
  "repository": "reaperdoesntknow/DualMind-TKD-Agentic-1.7B",
  "released_at_utc": "2026-07-12T09:29:33.617799+00:00",
  "architecture": "Qwen3ForCausalLM",
  "base_model": "Qwen/Qwen3-1.7B",
  "distillation_teacher": "Qwen/Qwen3-8B",
  "stage_1": {
    "method": "Topology-guided knowledge distillation",
    "training_scope": "Full-model",
    "dataset": "0xZee/dataset-CoT-Advanced-Calculus-268",
    "output": "/content/DualMind100/best"
  },
  "stage_2": {
    "method": "Assistant-masked agentic LoRA SFT",
    "training_scope": "LoRA followed by merge",
    "dataset": "NousResearch/hermes-function-calling-v1",
    "retention_replay": "0xZee/dataset-CoT-Advanced-Calculus-268",
    "output": "/content/DualMind100_agentic/merged"
  },
  "formal_benchmarks_available": false,
  "training_receipts": [
    "training_artifacts/DualMind100_config.json",
    "training_artifacts/DualMind100_train_log.jsonl",
    "training_artifacts/DualMind100_agentic_trainer_state.json"
  ]
}