metrics.json

{
  "evaluation_split": {
    "total": 40,
    "per_tool": 10,
    "tools": ["read", "bash", "edit", "write"]
  },
  "base_model": {
    "call_f1": 0.2895,
    "name_f1": 0.88,
    "exact_match": 0.275,
    "parse_rate": 0.825,
    "arguments_accuracy": 0.3235,
    "exact_by_tool": {"read": 0.8, "bash": 0.2, "edit": 0.0, "write": 0.1}
  },
  "fine_tuned_model": {
    "call_f1": 0.8,
    "name_f1": 1.0,
    "exact_match": 0.8,
    "parse_rate": 1.0,
    "arguments_accuracy": 0.8,
    "exact_by_tool": {"read": 0.9, "bash": 0.8, "edit": 0.9, "write": 0.6}
  }
}