{
  "model_details": {
    "provider": "openai",
    "model_requested": "gpt-oss-120b",
    "model_for_requests": "gpt-oss-120b",
    "api_base_url": "",
    "chat_completions_endpoint": ""
  },
  "cache_padding": {
    "enabled": false,
    "target_shared_prefix_tokens": 0,
    "calibration_shared_prefix_tokens": null,
    "target_prompt_tokens": 0,
    "calibration_prompt_tokens": null,
    "calibration_example_id": null,
    "applied_padding_tokens_estimate": 0,
    "examples_with_padding_applied": 0
  },
  "request_control_summary": {
    "configured": {},
    "attempts_total": 329,
    "attempts_with_control_telemetry": 0,
    "per_control": {
      "reasoning_effort": {
        "configured_value": null,
        "requested_attempts": 0,
        "sent_attempts": 0,
        "accepted_attempts": 0,
        "rejected_attempts": 0,
        "missing_from_final_request_attempts": 0,
        "acceptance_rate": null,
        "rejected_reasons": {},
        "rejected_example_ids": []
      },
      "thinking_level": {
        "configured_value": null,
        "requested_attempts": 0,
        "sent_attempts": 0,
        "accepted_attempts": 0,
        "rejected_attempts": 0,
        "missing_from_final_request_attempts": 0,
        "acceptance_rate": null,
        "rejected_reasons": {},
        "rejected_example_ids": []
      },
      "effort": {
        "configured_value": null,
        "requested_attempts": 0,
        "sent_attempts": 0,
        "accepted_attempts": 0,
        "rejected_attempts": 0,
        "missing_from_final_request_attempts": 0,
        "acceptance_rate": null,
        "rejected_reasons": {},
        "rejected_example_ids": []
      },
      "verbosity": {
        "configured_value": null,
        "requested_attempts": 0,
        "sent_attempts": 0,
        "accepted_attempts": 0,
        "rejected_attempts": 0,
        "missing_from_final_request_attempts": 0,
        "acceptance_rate": null,
        "rejected_reasons": {},
        "rejected_example_ids": []
      },
      "prompt_cache_key": {
        "configured_value": null,
        "requested_attempts": 0,
        "sent_attempts": 0,
        "accepted_attempts": 0,
        "rejected_attempts": 0,
        "missing_from_final_request_attempts": 0,
        "acceptance_rate": null,
        "rejected_reasons": {},
        "rejected_example_ids": []
      },
      "gemini_cached_content": {
        "configured_value": null,
        "requested_attempts": 0,
        "sent_attempts": 0,
        "accepted_attempts": 0,
        "rejected_attempts": 0,
        "missing_from_final_request_attempts": 0,
        "acceptance_rate": null,
        "rejected_reasons": {},
        "rejected_example_ids": []
      },
      "requesty_auto_cache": {
        "configured_value": null,
        "requested_attempts": 0,
        "sent_attempts": 0,
        "accepted_attempts": 0,
        "rejected_attempts": 0,
        "missing_from_final_request_attempts": 0,
        "acceptance_rate": null,
        "rejected_reasons": {},
        "rejected_example_ids": []
      }
    }
  },
  "usage_metadata_summary": {
    "attempts_total": 329,
    "attempts_with_usage_metadata": 0,
    "attempts_with_cached_token_signals": 0,
    "cached_tokens_total_estimate": 0,
    "cache_read_tokens_total": 0,
    "cache_write_tokens_total": 0,
    "cache_token_fields_totals": {},
    "attempts_with_gemini_cached_content_token_signals": 0,
    "gemini_cached_content_token_count_total": 0,
    "gemini_cached_content_token_fields_totals": {}
  },
  "token_usage_totals": {
    "attempts_total": 329,
    "attempts_with_token_usage": 329,
    "attempts_with_output_tokens": 329,
    "attempts_with_cached_input_tokens": 0,
    "attempts_with_thinking_tokens": 0,
    "input_tokens_total": 260200,
    "cached_input_tokens_total": 0,
    "non_cached_input_tokens_total": 260200,
    "output_tokens_total": 195713,
    "thinking_tokens_total": 0,
    "output_tokens_definition": "total_tokens - prompt_tokens (or completion_tokens + thinking_tokens fallback)"
  },
  "mode": "metrics_only",
  "source_output_csv": "data/output/adv-ing____gptoss120b__2025-11-08-17-25.csv",
  "truth_label_count": 318,
  "prediction_count": 318,
  "evaluated_example_count": 318,
  "calibration_metrics": {
    "available": true,
    "sample_count": 318,
    "bin_count": 10,
    "ece": 0.16500000000000126,
    "mce": 1,
    "brier_score": 0.196685220125787
  },
  "first_prompt_timestamp": "2025-11-08T16:25:41.584007Z",
  "last_prompt_timestamp": "2025-11-08T16:51:39.671667Z",
  "overall_time_seconds": 1558.08766,
  "overall_time_human": "25m 58s",
  "truth_source": "output_csv_truth_column",
  "accuracy": 0.7924528301886793,
  "macro_precision": 0.7005490925769563,
  "macro_recall": 0.5658619929453262,
  "macro_f1": 0.6233434426982813,
  "per_label": {
    "AC": {
      "precision": 0.8421052631578947,
      "recall": 0.7619047619047619,
      "f1": 0.8,
      "support": 21
    },
    "AVG": {
      "precision": 0.9814814814814815,
      "recall": 0.654320987654321,
      "f1": 0.7851851851851851,
      "support": 81
    },
    "FA": {
      "precision": 0.9786096256684492,
      "recall": 0.8472222222222222,
      "f1": 0.9081885856079405,
      "support": 216
    },
    "unclassified": {
      "precision": 0,
      "recall": 0,
      "f1": 0,
      "support": 0
    }
  },
  "confusion_matrix": {
    "AC": {
      "AC": 16,
      "AVG": 0,
      "FA": 0,
      "unclassified": 5
    },
    "AVG": {
      "AC": 0,
      "AVG": 53,
      "FA": 4,
      "unclassified": 24
    },
    "FA": {
      "AC": 3,
      "AVG": 1,
      "FA": 183,
      "unclassified": 29
    },
    "unclassified": {
      "AC": 0,
      "AVG": 0,
      "FA": 0,
      "unclassified": 0
    }
  },
  "labels": [
    "AC",
    "AVG",
    "FA",
    "unclassified"
  ],
  "total_examples": 318,
  "label_metrics_available": true,
  "run_config": {
    "task_name": "ADV *-ing* clause",
    "prompt_layout": "standard",
    "task_description": "Classification of adverbial *-ing* clauses as one of: FA (Free Adjunct), AVG (Adverbial Verbal Gerund) or AC (Absolute Construction).",
    "tags": "open source;*-ing*;syntax;adverbial;clause;English"
  },
  "source_input_csv": "",
  "source_labels_csv": "",
  "estimated_cost_usd": 0.4172208,
  "pricing": {
    "status": "priced",
    "status_label": "priced",
    "provider_key": "e-infra",
    "matched_key": "gptoss120b",
    "resolved_key": "gpt-oss-120b",
    "pricing_tier": "standard",
    "estimated_cost_usd": 0.4172208,
    "catalog_updated_at": "2026-05-27T15:36:49Z",
    "missing_buckets": []
  }
}