Spaces:
Runtime error
Runtime error
File size: 6,130 Bytes
71781d4 |
1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16 17 18 19 20 21 22 23 24 25 26 27 28 29 30 31 32 33 34 35 36 37 38 39 40 41 42 43 44 45 46 47 48 49 50 51 52 53 54 55 56 57 58 59 60 61 62 63 64 65 66 67 68 69 70 71 72 73 74 75 76 77 78 79 80 81 82 83 84 85 86 87 88 89 90 91 92 93 94 95 96 97 98 99 100 101 102 103 104 105 106 107 108 109 110 111 112 113 114 115 116 117 118 119 120 121 122 123 124 125 126 127 128 129 130 131 132 133 134 135 |
# Copyright 2025 The HuggingFace Evaluate Authors.
#
# Licensed under the Apache License, Version 2.0 (the "License");
# you may not use this file except in compliance with the License.
# You may obtain a copy of the License at
#
# http://www.apache.org/licenses/LICENSE-2.0
#
# Unless required by applicable law or agreed to in writing, software
# distributed under the License is distributed on an "AS IS" BASIS,
# WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied.
# See the License for the specific language governing permissions and
# limitations under the License.
"""Tests for the FEVER (Fact Extraction and VERification) metric."""
import unittest
from fever import FEVER # assuming your metric file is named fever.py
fever = FEVER()
class TestFEVER(unittest.TestCase):
def test_perfect_prediction(self):
preds = [{"label": "SUPPORTED", "evidence": ["E1", "E2"]}]
refs = [{"label": "SUPPORTED", "evidence_sets": [["E1", "E2"]]}]
result = fever.compute(predictions=preds, references=refs)
self.assertAlmostEqual(result["label_accuracy"], 1.0)
self.assertAlmostEqual(result["fever_score"], 1.0)
self.assertAlmostEqual(result["evidence_precision"], 1.0)
self.assertAlmostEqual(result["evidence_recall"], 1.0)
self.assertAlmostEqual(result["evidence_f1"], 1.0)
def test_label_only_correct(self):
preds = [{"label": "SUPPORTED", "evidence": ["X1", "X2"]}]
refs = [{"label": "SUPPORTED", "evidence_sets": [["E1", "E2"]]}]
result = fever.compute(predictions=preds, references=refs)
self.assertAlmostEqual(result["label_accuracy"], 1.0)
self.assertAlmostEqual(result["fever_score"], 0.0)
self.assertTrue(result["evidence_f1"] < 1.0)
def test_label_incorrect(self):
preds = [{"label": "REFUTED", "evidence": ["E1", "E2"]}]
refs = [{"label": "SUPPORTED", "evidence_sets": [["E1", "E2"]]}]
result = fever.compute(predictions=preds, references=refs)
self.assertAlmostEqual(result["label_accuracy"], 0.0)
self.assertAlmostEqual(result["fever_score"], 0.0)
def test_partial_evidence_overlap(self):
preds = [{"label": "SUPPORTED", "evidence": ["E1"]}]
refs = [{"label": "SUPPORTED", "evidence_sets": [["E1", "E2"]]}]
result = fever.compute(predictions=preds, references=refs)
self.assertAlmostEqual(result["label_accuracy"], 1.0)
self.assertAlmostEqual(result["fever_score"], 0.0)
self.assertAlmostEqual(result["evidence_precision"], 1.0)
self.assertAlmostEqual(result["evidence_recall"], 0.5)
self.assertTrue(0 < result["evidence_f1"] < 1.0)
def test_extra_evidence_still_correct(self):
preds = [{"label": "SUPPORTED", "evidence": ["E1", "E2", "X1"]}]
refs = [{"label": "SUPPORTED", "evidence_sets": [["E1", "E2"]]}]
result = fever.compute(predictions=preds, references=refs)
self.assertAlmostEqual(result["fever_score"], 1.0)
self.assertTrue(result["evidence_precision"] < 1.0)
self.assertAlmostEqual(result["evidence_recall"], 1.0)
def test_multiple_gold_sets(self):
preds = [{"label": "SUPPORTED", "evidence": ["E3", "E4"]}]
refs = [{"label": "SUPPORTED", "evidence_sets": [["E1", "E2"], ["E3", "E4"]]}]
result = fever.compute(predictions=preds, references=refs)
self.assertAlmostEqual(result["fever_score"], 1.0)
self.assertAlmostEqual(result["label_accuracy"], 1.0)
def test_mixed_examples(self):
preds = [
{"label": "SUPPORTED", "evidence": ["A1", "A2"]},
{"label": "SUPPORTED", "evidence": ["B1"]},
{"label": "REFUTED", "evidence": ["C1", "C2"]},
]
refs = [
{"label": "SUPPORTED", "evidence_sets": [["A1", "A2"]]},
{"label": "SUPPORTED", "evidence_sets": [["B1", "B2"]]},
{"label": "SUPPORTED", "evidence_sets": [["C1", "C2"]]},
]
result = fever.compute(predictions=preds, references=refs)
self.assertTrue(0 < result["label_accuracy"] < 1.0)
self.assertTrue(0 <= result["fever_score"] < 1.0)
self.assertTrue(0 <= result["evidence_f1"] <= 1.0)
def test_empty_evidence_prediction(self):
preds = [{"label": "SUPPORTED", "evidence": []}]
refs = [{"label": "SUPPORTED", "evidence_sets": [["E1", "E2"]]}]
result = fever.compute(predictions=preds, references=refs)
self.assertEqual(result["evidence_precision"], 0.0)
self.assertEqual(result["evidence_recall"], 0.0)
self.assertEqual(result["evidence_f1"], 0.0)
def test_empty_gold_evidence(self):
preds = [{"label": "SUPPORTED", "evidence": ["E1", "E2"]}]
refs = [{"label": "SUPPORTED", "evidence_sets": [[]]}]
result = fever.compute(predictions=preds, references=refs)
self.assertEqual(result["evidence_recall"], 0.0)
def test_multiple_examples_micro_averaging(self):
preds = [
{"label": "SUPPORTED", "evidence": ["E1"]},
{"label": "SUPPORTED", "evidence": ["F1", "F2"]},
]
refs = [
{"label": "SUPPORTED", "evidence_sets": [["E1", "E2"]]},
{"label": "SUPPORTED", "evidence_sets": [["F1", "F2"]]},
]
result = fever.compute(predictions=preds, references=refs)
self.assertTrue(result["evidence_f1"] < 1.0)
self.assertAlmostEqual(result["label_accuracy"], 1.0)
def test_fever_score_requires_label_match(self):
preds = [{"label": "REFUTED", "evidence": ["E1", "E2"]}]
refs = [{"label": "SUPPORTED", "evidence_sets": [["E1", "E2"]]}]
result = fever.compute(predictions=preds, references=refs)
self.assertEqual(result["fever_score"], 0.0)
self.assertEqual(result["label_accuracy"], 0.0)
def test_empty_input_list(self):
preds, refs = [], []
result = fever.compute(predictions=preds, references=refs)
for k in result:
self.assertEqual(result[k], 0.0)
if __name__ == "__main__":
unittest.main()
|