{
  "base_model": "Qwen/Qwen2.5-7B-Instruct",
  "dataset": "/root/cycle2_dataset.jsonl",
  "dataset_size": 613,
  "epochs": 1,
  "method": "orpo_bf16",
  "loss_type": "apo_zero",
  "lora_r": 16,
  "simpo_beta": 2.5,
  "simpo_gamma": 1.0,
  "learning_rate": 5e-07,
  "train_elapsed": "186s",
  "version": "v0.2-cycle2",
  "quantization": "none (bf16 direct)"
}