avid-db/reports/2022/AVID-2022-R0001.json at main · avidml/avid-db · GitHub

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
{
    "data_type": "AVID",
    "data_version": "0.1",
    "metadata": {
        "report_id": "AVID-2022-R0001"
    },
    "affects": {
        "developer": [],
        "deployer": [
            "HuggingFace"
        ],
        "artifacts": [
            {
                "type": "Model",
                "name": "bert-base-uncased"
            }
        ]
    },
    "problemtype": {
        "classof": "LLM Evaluation",
        "type": "Detection",
        "description": {
            "lang": "eng",
            "value": "Gender Bias in Sentence Completion Tasks performed by bert-base-uncased using the HONEST metric"
        }
    },
    "metrics": [
        {
            "name": "HONEST",
            "features": {
                "measured": "Hurtful lexical category: prostitution",
                "sensitive": "gender"
            },
            "detection_method": {
                "type": "Significance Test",
                "name": "Two-sample z test for proportions"
            }
        }
    ],
    "references": [
        {
            "type": "source",
            "label": "Gender Bias Evaluation for Masked Language modelling: HONEST",
            "url": "https://github.com/avidml/evaluating-LLMs/blob/main/notebooks/evaluation_honest.ipynb"
        },
        {
            "type": "model",
            "label": "bert-base-uncased on Hugging Face",
            "url": "https://huggingface.co/bert-base-uncased"
        },
        {
            "type": "misc",
            "label": "HONEST: Measuring Hurtful Sentence Completion in Language Models",
            "url": "https://github.com/MilaNLProc/honest"
        }
    ],
    "description": {
        "lang": "eng",
        "value": "Sentence completions by bert-base-uncased were found to be significantly biased for one lexical category as defined by the HONEST hurtful sentence completion framework."
    },
    "impact": {
        "avid": {
            "vuln_id": "AVID-2022-V001",
            "risk_domain": [
                "Ethics"
            ],
            "sep_view": [
                "E0101: Group fairness"
            ],
            "lifecycle_view": [
                "L05: Evaluation"
            ],
            "taxonomy_version": "0.1"
        }
    },
    "credit": [
        {
            "lang": "eng",
            "value": "Harry Saini, AVID"
        },
        {
            "lang": "eng",
            "value": "Sasha Luccioni, Hugging Face"
        }
    ],
    "reported_date": "2022-11-09"
}