{
  "id": "T1_dict_argmax",
  "language": "Python",
  "description": "Write a Python function + 3 tests",
  "prompt": "Write a Python file at /tmp/bench_task/<config>/solution.py with:\n\n    def argmax_numeric(dicts, name_key='name'):\n        '''Return the dict with the maximum sum of NUMERIC values across all keys\n        (recursive into nested lists and dicts). On ties, prefer the one whose\n        name_key value is alphabetically earlier. Raise ValueError on empty input.'''\n        ...\n\n    Include 3 pytest unit tests (test_*) in the same file that:\n      1. basic distinct values returns the max\n      2. tie-break by name works\n      3. empty input raises ValueError\n\n    Then run `pytest /tmp/bench_task/<config>/solution.py -v` and report:\n      - exit code\n      - number of tests passed",
  "ground_truth": {
    "tests_should_pass": 3,
    "function_signature_must_match": "def argmax_numeric(dicts, name_key='name')",
    "raises_value_error_on_empty": true,
    "model_for_eval": "qwen/qwen3.7-flash",
    "grading_strategy": "pytest exit code (deterministic), no LLM judge needed"
  },
  "winner_metric": "lowest wall_seconds_with_3_tests_passing"
}