Hugging Face's logo

KRISHNAPURI
/
q-trust-codebert

Text Classification
Transformers
ONNX
Safetensors
roberta
code
cryptography
post-quantum
static-analysis
text-embeddings-inference
Model card Files
xet
 Community
q-trust-codebert
File size: 4,055 Bytes
a36c9b3
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
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
136
137
138
139
140
141
142
143
144
145
146
147
148
149
150
151
152
153
154
155
156
157
158
159
160
161
162
163
{
  "seed": 42,
  "comparisons": [
    {
      "model": "discovery/CryptoCodeDetector",
      "metric": "F1 (held-out real code)",
      "n": 2415,
      "qtrust": {
        "baseline": "qtrust (rules+AST+ML ensemble)",
        "precision": 0.952,
        "recall": 0.953,
        "f1": 0.9525,
        "accuracy": 0.9271
      },
      "baselines": [
        {
          "baseline": "rules-only (static layer)",
          "precision": 0.9794,
          "recall": 0.5127,
          "f1": 0.673,
          "accuracy": 0.6178
        },
        {
          "baseline": "majority (always crypto)",
          "precision": 0.7673,
          "recall": 1.0,
          "f1": 0.8683,
          "accuracy": 0.7673
        },
        {
          "baseline": "random coin-flip",
          "precision": 0.7686,
          "recall": 0.4895,
          "f1": 0.5981,
          "accuracy": 0.4952
        }
      ],
      "best_baseline": "majority (always crypto)",
      "relative_gain": 0.097
    },
    {
      "model": "discovery/CryptoCodeDetector",
      "metric": "F1 on adversarial holdout (132 \u00a729 cases)",
      "n": 132,
      "qtrust": {
        "baseline": "qtrust (rules+AST+ML ensemble)",
        "precision": 0.8,
        "recall": 0.4,
        "f1": 0.5333
      },
      "baselines": [
        {
          "baseline": "rules-only (static layer)",
          "precision": 0.7778,
          "recall": 0.35,
          "f1": 0.4828
        }
      ],
      "best_baseline": "rules-only (static layer)",
      "relative_gain": 0.1046
    },
    {
      "model": "discovery/AlgorithmPurposeClassifier",
      "metric": "accuracy (held-out real triples)",
      "n": 294,
      "qtrust": {
        "baseline": "qtrust (context-aware)",
        "accuracy": 1.0
      },
      "baselines": [
        {
          "baseline": "family-prior-only (no context)",
          "accuracy": 1.0
        },
        {
          "baseline": "majority purpose",
          "accuracy": 0.483
        }
      ],
      "best_baseline": "family-prior-only (no context)",
      "relative_gain": 0.0
    },
    {
      "model": "migration/PQCRecommender",
      "metric": "family-correct rate (real triples)",
      "n": 294,
      "qtrust": {
        "baseline": "qtrust (purpose-aware catalog)",
        "accuracy": 0.7007
      },
      "baselines": [
        {
          "baseline": "always-ML-KEM-768",
          "accuracy": 0.0782
        }
      ],
      "best_baseline": "always-ML-KEM-768",
      "relative_gain": 7.9565
    },
    {
      "model": "risk/QuantumExposureModel",
      "metric": "MAE (held-out real TLS hosts)",
      "n_train_hosts": 221,
      "n_eval": 56,
      "qtrust": {
        "baseline": "qtrust (calibrated, host-held-out)",
        "mae": 0.0
      },
      "baselines": [
        {
          "baseline": "constant-mean predictor",
          "mae": 7.4412
        }
      ],
      "best_baseline": "constant-mean predictor",
      "relative_gain": 1.0
    },
    {
      "model": "monitoring/CryptoAnomalyDetector",
      "metric": "false-positive rate (held-out real hosts)",
      "n": 56,
      "qtrust": {
        "baseline": "qtrust (baseline+zscore)",
        "fp_rate": 0.0
      },
      "baselines": [
        {
          "baseline": "always-alert",
          "fp_rate": 1.0
        }
      ],
      "best_baseline": "always-alert",
      "relative_gain": 1.0
    },
    {
      "model": "vendor/SupplyChainRiskModel",
      "metric": "ranking agreement (real NVD vendors)",
      "n": 16,
      "qtrust": {
        "baseline": "qtrust (5-layer propagation)",
        "kendall_tau": 0.7176,
        "ndcg@5": 0.9722
      },
      "baselines": [
        {
          "baseline": "random ordering",
          "kendall_tau": -0.0695
        },
        {
          "baseline": "uniform scores",
          "kendall_tau": 0.0
        }
      ],
      "best_baseline": "uniform scores",
      "relative_gain": 0.7176
    }
  ],
  "summary": {
    "comparisons_run": 7,
    "models_beat_best_baseline": 6,
    "mean_relative_gain": 1.5537
  }
}