Enhancing Evaluator Input with Full Task_Run Data
Hi, is there a way evaluator can get as input not only eval(output: str, expected: Dict[str, Any], metadata) but full task_run, like
{'dataset_example_id': 'R1wbGU6MQ==',
'output': 'some_output',
'repetition_number': 1,
'start_time': '2025-08-10T20:19:13.498761+00:00',
'end_time': '2025-08-10T20:19:13.498644+00:00',
'trace_id': 'f24c282d3b3ebdfad50b3574476e127f',
'error': None,
'id': 'RXhwZXJpFJ1bjoz',
'experiment_id': 'RXhwZXJpbWVuDoz'}
