{
  "model_details": {
    "provider": "openai",
    "model_requested": "gpt-5.4-pro",
    "model_for_requests": "gpt-5.4-pro",
    "api_base_url": "https://api.openai.com/v1",
    "chat_completions_endpoint": "https://api.openai.com/v1/chat/completions"
  },
  "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": 762,
    "attempts_with_control_telemetry": 538,
    "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": 762,
    "attempts_with_usage_metadata": 538,
    "attempts_with_cached_token_signals": 538,
    "cached_tokens_total_estimate": 0,
    "cache_read_tokens_total": 0,
    "cache_write_tokens_total": 0,
    "cache_token_fields_totals": {
      "usage.input_tokens_details.cached_tokens": 0
    },
    "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": 762,
    "attempts_with_token_usage": 318,
    "attempts_with_output_tokens": 318,
    "attempts_with_cached_input_tokens": 0,
    "attempts_with_thinking_tokens": 538,
    "input_tokens_total": 288883,
    "cached_input_tokens_total": 0,
    "non_cached_input_tokens_total": 288883,
    "output_tokens_total": 445694,
    "thinking_tokens_total": 762111,
    "output_tokens_definition": "total_tokens - prompt_tokens (or completion_tokens + thinking_tokens fallback)"
  },
  "truth_label_count": 318,
  "prediction_count": 318,
  "evaluated_example_count": 318,
  "calibration_metrics": {
    "available": true,
    "sample_count": 318,
    "bin_count": 10,
    "ece": 0.089150943396226,
    "mce": 0.2715384615384615,
    "brier_score": 0.1391562893081761
  },
  "first_prompt_timestamp": "2026-03-11T23:22:39.487962Z",
  "last_prompt_timestamp": "2026-03-13T20:34:25.220196Z",
  "overall_time_seconds": 162705.732234,
  "overall_time_human": "1d 21h 11m 46s",
  "accuracy": 0.8459119496855346,
  "macro_precision": 0.7261363636363636,
  "macro_recall": 0.6428571428571428,
  "macro_f1": 0.6814716683313518,
  "per_label": {
    "AC": {
      "precision": 0.95,
      "recall": 0.9047619047619048,
      "f1": 0.9268292682926829,
      "support": 21
    },
    "AVG": {
      "precision": 0.9705882352941176,
      "recall": 0.8148148148148148,
      "f1": 0.8859060402684563,
      "support": 81
    },
    "FA": {
      "precision": 0.983957219251337,
      "recall": 0.8518518518518519,
      "f1": 0.913151364764268,
      "support": 216
    },
    "unclassified": {
      "precision": 0.0,
      "recall": 0.0,
      "f1": 0.0,
      "support": 0
    }
  },
  "confusion_matrix": {
    "AC": {
      "AC": 19,
      "AVG": 0,
      "FA": 0,
      "unclassified": 2
    },
    "AVG": {
      "AC": 0,
      "AVG": 66,
      "FA": 3,
      "unclassified": 12
    },
    "FA": {
      "AC": 1,
      "AVG": 2,
      "FA": 184,
      "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": "*-ing*;syntax;adverbial;clause;English"
  },
  "source_input_csv": "",
  "source_output_csv": "data/output/adv-ing__openai__gpt54pro__2026-03-12-00-22.csv",
  "source_labels_csv": ""
}