| | from trignum_core.evaluator import Evaluator
|
| | from trignum_core.subtractive_filter import SubtractiveFilter
|
| |
|
| |
|
| | def run_preflight(samples):
|
| | sf = SubtractiveFilter()
|
| | tp = fp = tn = fn = 0
|
| |
|
| | for sample in samples:
|
| | result = sf.apply(sample["text"])
|
| | predicted = len(result.illogics_found) > 0
|
| | actual = sample.get("has_hallucination", False)
|
| |
|
| | if predicted and actual: tp += 1
|
| | elif predicted and not actual: fp += 1
|
| | elif not predicted and actual: fn += 1
|
| | else: tn += 1
|
| |
|
| | metrics = Evaluator.compute_metrics(tp, fp, tn, fn)
|
| | print(metrics)
|
| |
|
| | if __name__ == "__main__":
|
| |
|
| | samples = [{"text": "All cats are dogs.", "has_hallucination": True}]
|
| | run_preflight(samples)
|
| |
|
