data/eval_base_qwen.json

{
  "name": "model[unsloth/qwen2.5-3b-instruct-unsloth-bnb-4bit]",
  "episodes": 100,
  "logical_correction_rate": 0.92,
  "pymatching_beat_rate": 0.0,
  "format_compliance_rate": 1.0,
  "format_partial_rate": 0.0,
  "syndrome_consistency_rate": 0.66,
  "mean_syndrome_consistency": 0.83,
  "mean_hamming_overlap": 0.81,
  "mean_total_reward": 0.7905,
  "exact_match_pymatching": 0.66,
  "mean_output_length": 7.0,
  "hard_syndrome_count": 6,
  "hard_syndrome_lcr": 0.6666666666666666,
  "hard_syndrome_beat_rate": 0.0,
  "level": "L2_target"
}