{
  "method": "oct_behaviour",
  "behavior_id": "sycophantic",
  "spec_sha256": "d0308786f3c8bec70f79ed9db4d66d83f7a97ee4b8ea00a7acccc084f77526e6",
  "spec_sha256_scheme": "file-v1",
  "spec_extends": null,
  "parent_spec_sha256": null,
  "base_model": "Qwen/Qwen2.5-32B-Instruct",
  "train_file": "runs/oct-qwen-2.5-32b-it-sycophantic/dpo-view.jsonl",
  "buckets": null,
  "n_rows": 8691,
  "n_pairs": 8691,
  "n_reversed_pairs": 0,
  "rank": 64,
  "lora_dropout": 0.0,
  "learning_rate": 5e-05,
  "beta": 0.1,
  "nll_coef": 0.1,
  "kl_coef": 0.001,
  "lr_scheduler_type": "cosine",
  "epochs": 1.0,
  "effective_batch": 32,
  "max_len": 1024,
  "grad_ckpt": true,
  "seed": 0,
  "optimizer_steps": 272,
  "train_loss": 0.14388378665727727
}