Datasets:
mdsajjadullah
/
ml-failures-dataset

Dataset card Data Studio Files
xet
 Community
1
scenario_id
string
failure_type
string
feature_values
string
true_label
int64
train_accuracy
float64
val_accuracy
float64
precision
float64
recall
float64
f1_score
float64
error_description
string
root_cause
string
fix_strategy
string
severity
string
extra_info
string
ADV_0000
Adversarial_Attack
[-2.1595, -0.1538, 1.2488, -1.0119, 0.0618, -0.8754, 0.9704, -1.3149, -1.9248, -0.2198, 0.4643, -0.7891, 0.9843, 0.9587, -0.1983]
1
0.9111
0.9
0.9
0.9
0.9
Tiny input perturbation (ε=0.2) caused misclassification.
Model has sensitive decision boundaries, not robust to small noise.
Adversarial training, input preprocessing, ensemble defenses, certified robustness.
Critical
{'epsilon': 0.2, 'clean_acc': 0.9111}
ADV_0001
Adversarial_Attack
[0.2975, 0.1661, -0.2945, -0.0204, -2.0494, -2.3287, 0.3811, 0.971, -0.505, -0.9629, -1.2286, 0.1694, -1.9897, -1.1195, 1.5592]
1
0.9111
0.9
0.9
0.9
0.9
Tiny input perturbation (ε=0.2) caused misclassification.
Model has sensitive decision boundaries, not robust to small noise.
Adversarial training, input preprocessing, ensemble defenses, certified robustness.
Critical
{'epsilon': 0.2, 'clean_acc': 0.9111}
ADV_0002
Adversarial_Attack
[1.8392, -1.3487, 1.2331, 1.7485, -1.0796, -0.5357, -0.8978, -0.2363, 0.3896, 1.1905, -0.3184, 0.5905, -2.7206, -1.4704, 0.0443]
0
0.9111
0.9
0.9
0.9
0.9
Tiny input perturbation (ε=0.2) caused misclassification.
Model has sensitive decision boundaries, not robust to small noise.
Adversarial training, input preprocessing, ensemble defenses, certified robustness.
Critical
{'epsilon': 0.2, 'clean_acc': 0.9111}
ADV_0003
Adversarial_Attack
[0.4869, -0.8221, -0.062, 0.1512, 1.5727, 0.5864, -0.4668, -1.5933, -0.5744, 0.1465, -0.6418, 1.0677, -0.2945, -0.7943, -0.3033]
0
0.9111
0.9
0.9
0.9
0.9
Tiny input perturbation (ε=0.2) caused misclassification.
Model has sensitive decision boundaries, not robust to small noise.
Adversarial training, input preprocessing, ensemble defenses, certified robustness.
Critical
{'epsilon': 0.2, 'clean_acc': 0.9111}
ADV_0004
Adversarial_Attack
[0.2345, -0.9533, 0.4324, -0.8372, 1.3122, 0.2328, 1.7059, 0.3311, -0.4654, 0.0444, 0.6151, -0.7711, 1.428, 0.3644, 0.2651]
1
0.9111
0.9
0.9
0.9
0.9
Tiny input perturbation (ε=0.2) caused misclassification.
Model has sensitive decision boundaries, not robust to small noise.
Adversarial training, input preprocessing, ensemble defenses, certified robustness.
Critical
{'epsilon': 0.2, 'clean_acc': 0.9111}
ADV_0005
Adversarial_Attack
[-2.2054, -0.9492, -1.6115, 1.1056, -0.7839, 0.3522, 0.0199, 1.6721, 0.0917, 0.4152, -0.0182, 0.8016, -1.4198, -0.5222, 0.2075]
0
0.9111
0.9
0.9
0.9
0.9
Tiny input perturbation (ε=0.2) caused misclassification.
Model has sensitive decision boundaries, not robust to small noise.
Adversarial training, input preprocessing, ensemble defenses, certified robustness.
Critical
{'epsilon': 0.2, 'clean_acc': 0.9111}
ADV_0006
Adversarial_Attack
[-0.5146, 0.4975, -0.6998, -0.0087, 0.8953, 0.2917, 0.3629, -1.0627, -0.9741, 1.218, -0.057, 0.6969, -0.2936, -0.5447, -0.4511]
1
0.9111
0.9
0.9
0.9
0.9
Tiny input perturbation (ε=0.2) caused misclassification.
Model has sensitive decision boundaries, not robust to small noise.
Adversarial training, input preprocessing, ensemble defenses, certified robustness.
Critical
{'epsilon': 0.2, 'clean_acc': 0.9111}
ADV_0007
Adversarial_Attack
[1.667, 0.0727, 1.9, 2.3643, 1.1705, -0.4474, -0.5924, 0.0381, 0.4267, 0.2177, 1.2516, 1.3995, -0.185, -0.658, 0.4417]
0
0.9111
0.9
0.9
0.9
0.9
Tiny input perturbation (ε=0.2) caused misclassification.
Model has sensitive decision boundaries, not robust to small noise.
Adversarial training, input preprocessing, ensemble defenses, certified robustness.
Critical
{'epsilon': 0.2, 'clean_acc': 0.9111}
ADV_0008
Adversarial_Attack
[-0.7457, -0.13, 0.0494, 1.2216, -1.6112, -2.3982, -1.1113, 2.7641, 0.4948, 0.5147, -0.5546, 0.1644, -1.5183, -0.7486, -1.383]
0
0.9111
0.9
0.9
0.9
0.9
Tiny input perturbation (ε=0.2) caused misclassification.
Model has sensitive decision boundaries, not robust to small noise.
Adversarial training, input preprocessing, ensemble defenses, certified robustness.
Critical
{'epsilon': 0.2, 'clean_acc': 0.9111}
ADV_0009
Adversarial_Attack
[0.3407, 2.7079, 0.8406, 0.9884, 0.8088, -0.85, -0.8208, 0.5928, -0.1941, -0.7978, 1.1234, 0.7228, 1.6818, -0.7005, -1.2517]
0
0.9111
0.9
0.9
0.9
0.9
Tiny input perturbation (ε=0.2) caused misclassification.
Model has sensitive decision boundaries, not robust to small noise.
Adversarial training, input preprocessing, ensemble defenses, certified robustness.
Critical
{'epsilon': 0.2, 'clean_acc': 0.9111}
ADV_0010
Adversarial_Attack
[0.6959, -0.0763, 0.3711, 0.8898, 0.1095, -0.7464, -0.306, -1.3047, -0.4615, 0.3056, -0.7506, 1.0582, -0.7715, -0.4552, -2.0672]
0
0.9111
0.9
0.9
0.9
0.9
Tiny input perturbation (ε=0.2) caused misclassification.
Model has sensitive decision boundaries, not robust to small noise.
Adversarial training, input preprocessing, ensemble defenses, certified robustness.
Critical
{'epsilon': 0.2, 'clean_acc': 0.9111}
ADV_0011
Adversarial_Attack
[3.5694, -1.5631, 0.0422, 1.9177, -1.8319, -0.1616, 0.045, 1.2208, 1.676, -0.5982, -0.2262, -0.7222, -3.0634, -2.974, 0.5939]
0
0.9111
0.9
0.9
0.9
0.9
Tiny input perturbation (ε=0.2) caused misclassification.
Model has sensitive decision boundaries, not robust to small noise.
Adversarial training, input preprocessing, ensemble defenses, certified robustness.
Critical
{'epsilon': 0.2, 'clean_acc': 0.9111}
ADV_0012
Adversarial_Attack
[-0.3577, 2.2297, -1.1339, 1.2264, -0.2962, -0.5811, 0.5987, 0.6979, 0.3037, -0.4533, 0.1206, -0.7848, 1.9305, 1.4203, 0.1041]
0
0.9111
0.9
0.9
0.9
0.9
Tiny input perturbation (ε=0.2) caused misclassification.
Model has sensitive decision boundaries, not robust to small noise.
Adversarial training, input preprocessing, ensemble defenses, certified robustness.
Critical
{'epsilon': 0.2, 'clean_acc': 0.9111}
ADV_0013
Adversarial_Attack
[-0.2254, -2.0838, 0.0674, 0.5432, 1.9191, -0.6792, 1.4585, 0.6104, -0.7335, 0.7625, 0.5518, 0.7684, -0.2905, 0.3176, 0.2932]
1
0.9111
0.9
0.9
0.9
0.9
Tiny input perturbation (ε=0.2) caused misclassification.
Model has sensitive decision boundaries, not robust to small noise.
Adversarial training, input preprocessing, ensemble defenses, certified robustness.
Critical
{'epsilon': 0.2, 'clean_acc': 0.9111}
ADV_0014
Adversarial_Attack
[-0.7365, 0.3982, -0.0657, -1.5214, -0.5916, -0.8274, -1.4708, 0.0994, -0.2531, 0.0327, -0.3474, -0.2263, -0.5762, -0.0715, 0.083]
1
0.9111
0.9
0.9
0.9
0.9
Tiny input perturbation (ε=0.2) caused misclassification.
Model has sensitive decision boundaries, not robust to small noise.
Adversarial training, input preprocessing, ensemble defenses, certified robustness.
Critical
{'epsilon': 0.2, 'clean_acc': 0.9111}
ADV_0015
Adversarial_Attack
[-0.2209, -0.9091, 0.2403, -1.0788, -0.9725, -0.7081, 0.3041, 0.1572, 0.3634, 1.228, -0.6696, 1.1174, -1.4092, 1.3675, 2.6812]
1
0.9111
0.9
0.9
0.9
0.9
Tiny input perturbation (ε=0.2) caused misclassification.
Model has sensitive decision boundaries, not robust to small noise.
Adversarial training, input preprocessing, ensemble defenses, certified robustness.
Critical
{'epsilon': 0.2, 'clean_acc': 0.9111}
ADV_0016
Adversarial_Attack
[-1.4033, -0.0943, -0.8685, -0.2043, -1.9629, 0.4283, 1.0828, 1.1937, 0.0522, 0.0182, -0.99, 0.4065, -1.1862, -0.6253, -0.6045]
0
0.9111
0.9
0.9
0.9
0.9
Tiny input perturbation (ε=0.2) caused misclassification.
Model has sensitive decision boundaries, not robust to small noise.
Adversarial training, input preprocessing, ensemble defenses, certified robustness.
Critical
{'epsilon': 0.2, 'clean_acc': 0.9111}
ADV_0017
Adversarial_Attack
[-0.4928, 0.3408, 0.6449, -0.8772, -0.5681, -0.8784, -0.3473, -0.4461, -0.8188, 0.2469, 1.1618, 1.0347, 0.2974, -0.0441, 0.7514]
1
0.9111
0.9
0.9
0.9
0.9
Tiny input perturbation (ε=0.2) caused misclassification.
Model has sensitive decision boundaries, not robust to small noise.
Adversarial training, input preprocessing, ensemble defenses, certified robustness.
Critical
{'epsilon': 0.2, 'clean_acc': 0.9111}
ADV_0018
Adversarial_Attack
[-0.3762, -1.1192, 1.3507, 0.3341, 0.9254, -0.4577, 0.7024, 0.2398, -0.6946, 0.2526, -0.2571, -0.7946, 0.2974, 0.1394, 0.5552]
1
0.9111
0.9
0.9
0.9
0.9
Tiny input perturbation (ε=0.2) caused misclassification.
Model has sensitive decision boundaries, not robust to small noise.
Adversarial training, input preprocessing, ensemble defenses, certified robustness.
Critical
{'epsilon': 0.2, 'clean_acc': 0.9111}
ADV_0019
Adversarial_Attack
[-1.4571, 0.0967, -0.7634, 1.8213, -0.0496, -0.9763, -1.5656, 2.2608, -0.0726, -0.7381, 0.4413, 1.1289, -0.4371, -0.8206, 0.1979]
0
0.9111
0.9
0.9
0.9
0.9
Tiny input perturbation (ε=0.2) caused misclassification.
Model has sensitive decision boundaries, not robust to small noise.
Adversarial training, input preprocessing, ensemble defenses, certified robustness.
Critical
{'epsilon': 0.2, 'clean_acc': 0.9111}
ADV_0020
Adversarial_Attack
[-0.0629, -0.3088, -1.4357, -1.2213, -1.968, -0.1578, -2.6061, 0.8927, 2.8019, -1.0765, -0.0636, -0.3076, -0.9226, 2.5514, 0.9096]
1
0.9111
0.9
0.9
0.9
0.9
Tiny input perturbation (ε=0.2) caused misclassification.
Model has sensitive decision boundaries, not robust to small noise.
Adversarial training, input preprocessing, ensemble defenses, certified robustness.
Critical
{'epsilon': 0.2, 'clean_acc': 0.9111}
ADV_0021
Adversarial_Attack
[0.5977, -0.3951, 1.3211, -0.8028, -0.3946, -0.8446, -0.2734, -0.0992, -0.6181, 0.5162, 0.2962, -0.5232, 0.0762, -0.955, 2.0081]
1
0.9111
0.9
0.9
0.9
0.9
Tiny input perturbation (ε=0.2) caused misclassification.
Model has sensitive decision boundaries, not robust to small noise.
Adversarial training, input preprocessing, ensemble defenses, certified robustness.
Critical
{'epsilon': 0.2, 'clean_acc': 0.9111}
ADV_0022
Adversarial_Attack
[-0.8077, -0.8356, 0.4274, -1.2326, -0.1194, 1.3293, -0.4505, -0.601, -0.4407, 0.5784, -1.0616, -0.7873, -0.7533, -0.2838, 1.2603]
1
0.9111
0.9
0.9
0.9
0.9
Tiny input perturbation (ε=0.2) caused misclassification.
Model has sensitive decision boundaries, not robust to small noise.
Adversarial training, input preprocessing, ensemble defenses, certified robustness.
Critical
{'epsilon': 0.2, 'clean_acc': 0.9111}
ADV_0023
Adversarial_Attack
[-1.5691, -0.4103, -0.9534, -1.3169, 2.0921, -1.8545, 0.8761, -1.5444, -3.2337, -1.5504, 0.3706, 1.4142, 0.6292, 0.0272, -1.3928]
1
0.9111
0.9
0.9
0.9
0.9
Tiny input perturbation (ε=0.2) caused misclassification.
Model has sensitive decision boundaries, not robust to small noise.
Adversarial training, input preprocessing, ensemble defenses, certified robustness.
Critical
{'epsilon': 0.2, 'clean_acc': 0.9111}
ADV_0024
Adversarial_Attack
[0.0184, 1.1639, 0.1646, -0.1026, -0.9194, 1.0449, 0.1138, 1.1875, 0.331, 0.3714, 1.116, 0.9291, 0.404, -0.435, -0.0089]
0
0.9111
0.9
0.9
0.9
0.9
Tiny input perturbation (ε=0.2) caused misclassification.
Model has sensitive decision boundaries, not robust to small noise.
Adversarial training, input preprocessing, ensemble defenses, certified robustness.
Critical
{'epsilon': 0.2, 'clean_acc': 0.9111}
ADV_0025
Adversarial_Attack
[0.5673, 0.4623, -1.2075, -0.0525, 0.696, 1.0718, 0.6551, 2.0235, 1.1557, 1.157, -1.6637, -1.4992, 0.0079, -0.1848, -0.8932]
1
0.9111
0.9
0.9
0.9
0.9
Tiny input perturbation (ε=0.2) caused misclassification.
Model has sensitive decision boundaries, not robust to small noise.
Adversarial training, input preprocessing, ensemble defenses, certified robustness.
Critical
{'epsilon': 0.2, 'clean_acc': 0.9111}
ADV_0026
Adversarial_Attack
[-0.7284, -1.7985, -0.8029, 1.5507, 0.5476, 0.9475, 1.2674, 0.6537, -0.0012, -0.3817, -0.4316, -0.8519, -0.1473, -0.6347, -1.4809]
1
0.9111
0.9
0.9
0.9
0.9
Tiny input perturbation (ε=0.2) caused misclassification.
Model has sensitive decision boundaries, not robust to small noise.
Adversarial training, input preprocessing, ensemble defenses, certified robustness.
Critical
{'epsilon': 0.2, 'clean_acc': 0.9111}
ADV_0027
Adversarial_Attack
[0.36, 2.546, -0.0218, 0.6156, -1.0828, -0.8936, 1.237, -0.3133, -0.2747, 1.1299, 1.3084, -0.043, 2.0273, 1.1892, -1.0261]
0
0.9111
0.9
0.9
0.9
0.9
Tiny input perturbation (ε=0.2) caused misclassification.
Model has sensitive decision boundaries, not robust to small noise.
Adversarial training, input preprocessing, ensemble defenses, certified robustness.
Critical
{'epsilon': 0.2, 'clean_acc': 0.9111}
ADV_0028
Adversarial_Attack
[0.7478, 0.7444, 1.845, -0.8943, 0.0805, -0.5608, 0.4334, -1.4692, -1.3687, 1.8053, 0.2937, 1.1379, 0.2295, -0.3048, -0.2628]
1
0.9111
0.9
0.9
0.9
0.9
Tiny input perturbation (ε=0.2) caused misclassification.
Model has sensitive decision boundaries, not robust to small noise.
Adversarial training, input preprocessing, ensemble defenses, certified robustness.
Critical
{'epsilon': 0.2, 'clean_acc': 0.9111}
ADV_0029
Adversarial_Attack
[-0.1866, 1.1066, -0.0714, 0.34, 0.9405, -0.0515, -0.1276, -2.563, -0.891, 0.1576, -2.3181, 0.0365, -0.1784, -0.2023, 0.9554]
0
0.9111
0.9
0.9
0.9
0.9
Tiny input perturbation (ε=0.2) caused misclassification.
Model has sensitive decision boundaries, not robust to small noise.
Adversarial training, input preprocessing, ensemble defenses, certified robustness.
Critical
{'epsilon': 0.2, 'clean_acc': 0.9111}
ADV_0030
Adversarial_Attack
[1.0845, 0.5218, 0.3873, 0.6312, -0.4086, -1.1939, -1.5866, 0.1861, 1.109, 0.1749, -0.8957, -0.3502, -0.9888, 0.279, 0.6739]
0
0.9111
0.9
0.9
0.9
0.9
Tiny input perturbation (ε=0.2) caused misclassification.
Model has sensitive decision boundaries, not robust to small noise.
Adversarial training, input preprocessing, ensemble defenses, certified robustness.
Critical
{'epsilon': 0.2, 'clean_acc': 0.9111}
ADV_0031
Adversarial_Attack
[0.1756, 2.159, 1.2451, -0.774, -0.2524, -1.0159, 0.119, -1.5405, -1.4837, 0.1235, -0.3942, 1.2177, 0.0392, -0.586, -0.4697]
0
0.9111
0.9
0.9
0.9
0.9
Tiny input perturbation (ε=0.2) caused misclassification.
Model has sensitive decision boundaries, not robust to small noise.
Adversarial training, input preprocessing, ensemble defenses, certified robustness.
Critical
{'epsilon': 0.2, 'clean_acc': 0.9111}
ADV_0032
Adversarial_Attack
[-0.6281, 0.7176, -0.2836, -1.4451, 1.6196, -1.1846, -0.9037, -0.3945, -1.8928, 0.4052, 1.3906, 0.4687, 1.2323, -0.611, -0.3677]
1
0.9111
0.9
0.9
0.9
0.9
Tiny input perturbation (ε=0.2) caused misclassification.
Model has sensitive decision boundaries, not robust to small noise.
Adversarial training, input preprocessing, ensemble defenses, certified robustness.
Critical
{'epsilon': 0.2, 'clean_acc': 0.9111}
ADV_0033
Adversarial_Attack
[0.8898, 0.0674, -0.2046, -0.9322, -1.2226, 1.4334, -1.2757, 0.1464, 1.7124, -0.7331, -0.7618, -1.1095, -0.3229, 0.9987, 0.1126]
1
0.9111
0.9
0.9
0.9
0.9
Tiny input perturbation (ε=0.2) caused misclassification.
Model has sensitive decision boundaries, not robust to small noise.
Adversarial training, input preprocessing, ensemble defenses, certified robustness.
Critical
{'epsilon': 0.2, 'clean_acc': 0.9111}
ADV_0034
Adversarial_Attack
[0.0151, 1.3701, -0.4965, 0.9677, -1.386, -0.9333, 0.6261, 0.7759, 0.1222, 1.279, 0.8349, -0.5323, 0.4022, -0.3053, 0.1078]
0
0.9111
0.9
0.9
0.9
0.9
Tiny input perturbation (ε=0.2) caused misclassification.
Model has sensitive decision boundaries, not robust to small noise.
Adversarial training, input preprocessing, ensemble defenses, certified robustness.
Critical
{'epsilon': 0.2, 'clean_acc': 0.9111}
ADV_0035
Adversarial_Attack
[-1.7918, 0.1258, -0.422, 0.2819, 0.3416, 0.8902, -0.2342, -0.3796, -0.1459, -0.2245, 0.1941, -1.9532, 0.6195, -0.482, -1.0792]
1
0.9111
0.9
0.9
0.9
0.9
Tiny input perturbation (ε=0.2) caused misclassification.
Model has sensitive decision boundaries, not robust to small noise.
Adversarial training, input preprocessing, ensemble defenses, certified robustness.
Critical
{'epsilon': 0.2, 'clean_acc': 0.9111}
ADV_0036
Adversarial_Attack
[0.4245, -1.3514, 0.4475, -0.3287, 0.2749, -0.3813, 0.9368, 0.2964, -0.2292, -1.6015, 1.7537, 0.6778, 0.3468, -0.5968, 1.0701]
1
0.9111
0.9
0.9
0.9
0.9
Tiny input perturbation (ε=0.2) caused misclassification.
Model has sensitive decision boundaries, not robust to small noise.
Adversarial training, input preprocessing, ensemble defenses, certified robustness.
Critical
{'epsilon': 0.2, 'clean_acc': 0.9111}
ADV_0037
Adversarial_Attack
[1.7142, -0.4869, 0.8004, -1.2892, 0.1737, 0.8337, 0.5238, -1.1688, 0.2693, 1.7953, -1.1671, 0.1545, 0.1259, 1.4797, 0.0178]
1
0.9111
0.9
0.9
0.9
0.9
Tiny input perturbation (ε=0.2) caused misclassification.
Model has sensitive decision boundaries, not robust to small noise.
Adversarial training, input preprocessing, ensemble defenses, certified robustness.
Critical
{'epsilon': 0.2, 'clean_acc': 0.9111}
ADV_0038
Adversarial_Attack
[0.1641, 0.524, 1.9281, 1.0585, 0.6659, 0.3132, -1.2281, -0.7482, -0.2912, 0.1436, 0.0487, 0.3953, -0.0905, -0.4282, 0.6002]
0
0.9111
0.9
0.9
0.9
0.9
Tiny input perturbation (ε=0.2) caused misclassification.
Model has sensitive decision boundaries, not robust to small noise.
Adversarial training, input preprocessing, ensemble defenses, certified robustness.
Critical
{'epsilon': 0.2, 'clean_acc': 0.9111}
ADV_0039
Adversarial_Attack
[1.0435, 0.3387, 0.1588, -0.9911, -0.3518, 0.5227, 0.7642, -0.1622, 0.8934, -0.4902, 1.0602, -2.1098, 1.6899, 1.5989, 1.6036]
1
0.9111
0.9
0.9
0.9
0.9
Tiny input perturbation (ε=0.2) caused misclassification.
Model has sensitive decision boundaries, not robust to small noise.
Adversarial training, input preprocessing, ensemble defenses, certified robustness.
Critical
{'epsilon': 0.2, 'clean_acc': 0.9111}
ADV_0040
Adversarial_Attack
[-0.1598, -0.0337, -0.5142, -0.1686, -0.4318, 1.3068, -0.5667, 1.6566, 1.9191, -0.9976, -1.1128, -0.9826, -0.0301, 0.6297, 0.0322]
1
0.9111
0.9
0.9
0.9
0.9
Tiny input perturbation (ε=0.2) caused misclassification.
Model has sensitive decision boundaries, not robust to small noise.
Adversarial training, input preprocessing, ensemble defenses, certified robustness.
Critical
{'epsilon': 0.2, 'clean_acc': 0.9111}
ADV_0041
Adversarial_Attack
[-1.5991, -0.3639, 0.1894, -0.5447, -0.0186, 0.2923, 0.4843, -0.7273, -1.2869, -1.5238, 0.2848, 0.4485, 0.6319, 0.5923, 1.9479]
1
0.9111
0.9
0.9
0.9
0.9
Tiny input perturbation (ε=0.2) caused misclassification.
Model has sensitive decision boundaries, not robust to small noise.
Adversarial training, input preprocessing, ensemble defenses, certified robustness.
Critical
{'epsilon': 0.2, 'clean_acc': 0.9111}
ADV_0042
Adversarial_Attack
[-1.9733, 0.7293, 0.5204, 0.9351, -0.3715, 0.221, -1.3658, 0.1993, -0.4752, 1.0409, -1.5033, 2.0424, -1.0868, -1.2243, 1.039]
0
0.9111
0.9
0.9
0.9
0.9
Tiny input perturbation (ε=0.2) caused misclassification.
Model has sensitive decision boundaries, not robust to small noise.
Adversarial training, input preprocessing, ensemble defenses, certified robustness.
Critical
{'epsilon': 0.2, 'clean_acc': 0.9111}
ADV_0043
Adversarial_Attack
[0.5353, -0.0214, 0.5001, -0.2197, 2.4783, 0.9718, -1.4064, 0.993, 0.5993, 1.1394, 0.6362, 0.41, 0.3766, -0.5953, -1.2228]
0
0.9111
0.9
0.9
0.9
0.9
Tiny input perturbation (ε=0.2) caused misclassification.
Model has sensitive decision boundaries, not robust to small noise.
Adversarial training, input preprocessing, ensemble defenses, certified robustness.
Critical
{'epsilon': 0.2, 'clean_acc': 0.9111}
ADV_0044
Adversarial_Attack
[-0.0934, -0.4688, -0.2339, 0.23, -1.4449, 0.4792, 0.9361, 1.0417, 0.5638, -0.2007, 1.2489, -0.8458, 0.3228, 0.3251, -0.8707]
0
0.9111
0.9
0.9
0.9
0.9
Tiny input perturbation (ε=0.2) caused misclassification.
Model has sensitive decision boundaries, not robust to small noise.
Adversarial training, input preprocessing, ensemble defenses, certified robustness.
Critical
{'epsilon': 0.2, 'clean_acc': 0.9111}
ADV_0045
Adversarial_Attack
[0.6281, -1.0188, 1.0703, 0.1648, 0.1972, -1.1534, -0.1588, -1.1542, -0.142, -1.5417, -1.7343, 0.8891, -1.5923, 0.1833, 0.8988]
0
0.9111
0.9
0.9
0.9
0.9
Tiny input perturbation (ε=0.2) caused misclassification.
Model has sensitive decision boundaries, not robust to small noise.
Adversarial training, input preprocessing, ensemble defenses, certified robustness.
Critical
{'epsilon': 0.2, 'clean_acc': 0.9111}
ADV_0046
Adversarial_Attack
[-0.1323, 0.7365, 0.1251, 0.4991, 1.3784, 0.6556, -1.5936, 0.6068, 0.474, -0.6221, 0.6095, 0.2002, 0.8407, 0.3556, -0.0395]
0
0.9111
0.9
0.9
0.9
0.9
Tiny input perturbation (ε=0.2) caused misclassification.
Model has sensitive decision boundaries, not robust to small noise.
Adversarial training, input preprocessing, ensemble defenses, certified robustness.
Critical
{'epsilon': 0.2, 'clean_acc': 0.9111}
ADV_0047
Adversarial_Attack
[-0.7187, -0.4231, 1.5077, -2.5044, 0.2033, -0.294, 0.0803, -0.7833, -1.1244, 0.6314, -0.7576, 0.2028, -0.7881, 0.1733, 1.3536]
1
0.9111
0.9
0.9
0.9
0.9
Tiny input perturbation (ε=0.2) caused misclassification.
Model has sensitive decision boundaries, not robust to small noise.
Adversarial training, input preprocessing, ensemble defenses, certified robustness.
Critical
{'epsilon': 0.2, 'clean_acc': 0.9111}
ADV_0048
Adversarial_Attack
[1.4606, 2.181, 1.2996, -1.849, -1.4447, 1.8341, 2.113, -1.7332, 0.4835, 0.684, 1.6051, 0.2434, 1.98, 1.0986, -0.1532]
0
0.9111
0.9
0.9
0.9
0.9
Tiny input perturbation (ε=0.2) caused misclassification.
Model has sensitive decision boundaries, not robust to small noise.
Adversarial training, input preprocessing, ensemble defenses, certified robustness.
Critical
{'epsilon': 0.2, 'clean_acc': 0.9111}
ADV_0049
Adversarial_Attack
[-0.5972, -0.8718, 0.9775, -1.2137, 0.6915, 0.6067, 0.5605, -0.1388, -0.7058, 0.7437, 2.5055, 0.5011, 1.4693, 0.6847, 0.1122]
1
0.9111
0.9
0.9
0.9
0.9
Tiny input perturbation (ε=0.2) caused misclassification.
Model has sensitive decision boundaries, not robust to small noise.
Adversarial training, input preprocessing, ensemble defenses, certified robustness.
Critical
{'epsilon': 0.2, 'clean_acc': 0.9111}
ADV_0050
Adversarial_Attack
[0.5139, -0.008, 0.7073, -0.3821, 0.8198, 1.2208, 0.7139, 1.223, 0.9763, -0.5754, -1.2205, -0.5812, 0.8366, 0.4094, -1.3024]
1
0.9111
0.9
0.9
0.9
0.9
Tiny input perturbation (ε=0.2) caused misclassification.
Model has sensitive decision boundaries, not robust to small noise.
Adversarial training, input preprocessing, ensemble defenses, certified robustness.
Critical
{'epsilon': 0.2, 'clean_acc': 0.9111}
ADV_0051
Adversarial_Attack
[-1.5946, -0.9936, -0.3557, -0.6587, -0.0824, 1.3601, -0.3575, -0.7163, 0.1149, -1.2456, -0.7295, -1.9566, -0.1718, -0.5641, 0.7304]
1
0.9111
0.9
0.9
0.9
0.9
Tiny input perturbation (ε=0.2) caused misclassification.
Model has sensitive decision boundaries, not robust to small noise.
Adversarial training, input preprocessing, ensemble defenses, certified robustness.
Critical
{'epsilon': 0.2, 'clean_acc': 0.9111}
ADV_0052
Adversarial_Attack
[-1.0046, -0.0518, -0.647, -0.203, 0.0312, 0.5304, 0.2823, 0.6099, -0.2783, 0.2188, 1.9742, 0.7423, 0.6906, 1.5318, -2.0985]
1
0.9111
0.9
0.9
0.9
0.9
Tiny input perturbation (ε=0.2) caused misclassification.
Model has sensitive decision boundaries, not robust to small noise.
Adversarial training, input preprocessing, ensemble defenses, certified robustness.
Critical
{'epsilon': 0.2, 'clean_acc': 0.9111}
ADV_0053
Adversarial_Attack
[-0.6187, 1.0984, 1.5996, -0.069, -2.0096, -0.2959, 1.8741, 0.1997, -0.8204, 0.9105, -0.6133, -0.0179, -1.0077, -1.8177, -0.3112]
0
0.9111
0.9
0.9
0.9
0.9
Tiny input perturbation (ε=0.2) caused misclassification.
Model has sensitive decision boundaries, not robust to small noise.
Adversarial training, input preprocessing, ensemble defenses, certified robustness.
Critical
{'epsilon': 0.2, 'clean_acc': 0.9111}
ADV_0054
Adversarial_Attack
[0.5548, -0.1751, -0.9876, -0.2854, 0.2024, -0.2136, -0.9382, 0.1614, 0.4058, -1.7264, -0.3792, -0.9477, -0.1751, 0.117, -0.2105]
1
0.9111
0.9
0.9
0.9
0.9
Tiny input perturbation (ε=0.2) caused misclassification.
Model has sensitive decision boundaries, not robust to small noise.
Adversarial training, input preprocessing, ensemble defenses, certified robustness.
Critical
{'epsilon': 0.2, 'clean_acc': 0.9111}
ADV_0055
Adversarial_Attack
[0.915, 1.6686, 0.4509, -0.299, -0.4508, 0.887, -0.0892, -1.7917, -0.9202, 0.0096, 0.4141, -0.6554, 0.1769, -1.5238, -0.9423]
1
0.9111
0.9
0.9
0.9
0.9
Tiny input perturbation (ε=0.2) caused misclassification.
Model has sensitive decision boundaries, not robust to small noise.
Adversarial training, input preprocessing, ensemble defenses, certified robustness.
Critical
{'epsilon': 0.2, 'clean_acc': 0.9111}
ADV_0056
Adversarial_Attack
[0.1208, 0.549, -0.7747, 0.6472, -1.1137, -0.1123, 0.8085, -0.4201, -0.0491, -0.1791, 0.2706, -0.0986, -0.0799, 0.0494, -0.5805]
0
0.9111
0.9
0.9
0.9
0.9
Tiny input perturbation (ε=0.2) caused misclassification.
Model has sensitive decision boundaries, not robust to small noise.
Adversarial training, input preprocessing, ensemble defenses, certified robustness.
Critical
{'epsilon': 0.2, 'clean_acc': 0.9111}
ADV_0057
Adversarial_Attack
[-0.5137, -0.3099, 1.4665, -0.1837, -1.0812, 0.1693, 1.7203, 1.015, 0.4427, 0.9304, 0.1695, -0.9737, 0.3771, 0.1122, 0.4998]
0
0.9111
0.9
0.9
0.9
0.9
Tiny input perturbation (ε=0.2) caused misclassification.
Model has sensitive decision boundaries, not robust to small noise.
Adversarial training, input preprocessing, ensemble defenses, certified robustness.
Critical
{'epsilon': 0.2, 'clean_acc': 0.9111}
ADV_0058
Adversarial_Attack
[0.4379, 2.2262, 0.4549, 0.1313, -0.825, 0.0393, 0.3926, -0.703, 0.1215, -1.9322, 0.6324, 0.762, 0.9637, -0.1148, 0.7716]
0
0.9111
0.9
0.9
0.9
0.9
Tiny input perturbation (ε=0.2) caused misclassification.
Model has sensitive decision boundaries, not robust to small noise.
Adversarial training, input preprocessing, ensemble defenses, certified robustness.
Critical
{'epsilon': 0.2, 'clean_acc': 0.9111}
ADV_0059
Adversarial_Attack
[-2.1042, -0.8483, -2.0182, 2.0818, 0.3412, -0.2475, 0.7193, 0.603, -1.1712, 0.0466, -0.3971, -2.1616, -0.341, -1.3784, -0.4891]
1
0.9111
0.9
0.9
0.9
0.9
Tiny input perturbation (ε=0.2) caused misclassification.
Model has sensitive decision boundaries, not robust to small noise.
Adversarial training, input preprocessing, ensemble defenses, certified robustness.
Critical
{'epsilon': 0.2, 'clean_acc': 0.9111}
ADV_0060
Adversarial_Attack
[-0.4728, -0.9112, 0.0454, 0.3462, 1.3269, 0.8019, -1.5135, -0.704, 0.1466, -1.7693, -0.7603, 0.7353, -0.1823, 0.0468, 1.8433]
0
0.9111
0.9
0.9
0.9
0.9
Tiny input perturbation (ε=0.2) caused misclassification.
Model has sensitive decision boundaries, not robust to small noise.
Adversarial training, input preprocessing, ensemble defenses, certified robustness.
Critical
{'epsilon': 0.2, 'clean_acc': 0.9111}
ADV_0061
Adversarial_Attack
[-0.976, -1.5788, -1.0162, -0.3617, -0.2, 0.5403, 0.2745, 1.6415, 0.8419, 0.9207, 0.6925, -0.6701, -0.0889, 1.2892, 0.2107]
1
0.9111
0.9
0.9
0.9
0.9
Tiny input perturbation (ε=0.2) caused misclassification.
Model has sensitive decision boundaries, not robust to small noise.
Adversarial training, input preprocessing, ensemble defenses, certified robustness.
Critical
{'epsilon': 0.2, 'clean_acc': 0.9111}
ADV_0062
Adversarial_Attack
[-0.0015, 0.4084, 2.8575, 0.4848, -0.6698, -1.3281, -0.8339, 1.3945, -0.4899, -0.8801, 0.9951, 0.6853, -0.1793, -1.3849, 0.5853]
0
0.9111
0.9
0.9
0.9
0.9
Tiny input perturbation (ε=0.2) caused misclassification.
Model has sensitive decision boundaries, not robust to small noise.
Adversarial training, input preprocessing, ensemble defenses, certified robustness.
Critical
{'epsilon': 0.2, 'clean_acc': 0.9111}
ADV_0063
Adversarial_Attack
[-0.3403, 0.3695, 0.9683, 1.5184, -0.0796, 0.7659, 1.5302, -1.331, -1.1968, -0.5264, -0.4115, -0.6268, -0.0823, -1.0426, -0.7714]
1
0.9111
0.9
0.9
0.9
0.9
Tiny input perturbation (ε=0.2) caused misclassification.
Model has sensitive decision boundaries, not robust to small noise.
Adversarial training, input preprocessing, ensemble defenses, certified robustness.
Critical
{'epsilon': 0.2, 'clean_acc': 0.9111}
ADV_0064
Adversarial_Attack
[-0.6728, 0.1512, -0.0415, 0.3724, -0.9269, 1.1791, 0.2811, -0.4038, 0.5147, -0.4023, -0.2391, -0.6312, 0.1806, 0.0891, 2.1047]
1
0.9111
0.9
0.9
0.9
0.9
Tiny input perturbation (ε=0.2) caused misclassification.
Model has sensitive decision boundaries, not robust to small noise.
Adversarial training, input preprocessing, ensemble defenses, certified robustness.
Critical
{'epsilon': 0.2, 'clean_acc': 0.9111}
ADV_0065
Adversarial_Attack
[1.2422, -0.4918, -1.0114, 0.9123, -0.799, 0.0947, -0.7885, -0.9047, 0.7845, -0.222, -1.4961, -0.2383, -1.6004, -1.041, -0.3408]
0
0.9111
0.9
0.9
0.9
0.9
Tiny input perturbation (ε=0.2) caused misclassification.
Model has sensitive decision boundaries, not robust to small noise.
Adversarial training, input preprocessing, ensemble defenses, certified robustness.
Critical
{'epsilon': 0.2, 'clean_acc': 0.9111}
ADV_0066
Adversarial_Attack
[0.2202, -0.5892, -0.0333, -0.0506, 0.421, -0.7187, -1.696, 0.8102, -0.1002, -0.3008, 0.6453, -0.4932, -0.5831, -0.4607, -0.1188]
1
0.9111
0.9
0.9
0.9
0.9
Tiny input perturbation (ε=0.2) caused misclassification.
Model has sensitive decision boundaries, not robust to small noise.
Adversarial training, input preprocessing, ensemble defenses, certified robustness.
Critical
{'epsilon': 0.2, 'clean_acc': 0.9111}
ADV_0067
Adversarial_Attack
[-0.2449, 1.3148, -1.9772, 0.122, 1.0294, 0.1263, -0.1685, 0.2963, -0.2218, 0.1718, 0.8083, 1.2392, 0.7805, 0.0728, -2.1362]
0
0.9111
0.9
0.9
0.9
0.9
Tiny input perturbation (ε=0.2) caused misclassification.
Model has sensitive decision boundaries, not robust to small noise.
Adversarial training, input preprocessing, ensemble defenses, certified robustness.
Critical
{'epsilon': 0.2, 'clean_acc': 0.9111}
ADV_0068
Adversarial_Attack
[1.1823, -0.4741, -0.4536, 1.4532, -1.2519, -1.9117, -1.0489, -1.7537, 0.2566, 0.0925, -1.5362, 0.6615, -2.5036, -0.595, 0.1291]
0
0.9111
0.9
0.9
0.9
0.9
Tiny input perturbation (ε=0.2) caused misclassification.
Model has sensitive decision boundaries, not robust to small noise.
Adversarial training, input preprocessing, ensemble defenses, certified robustness.
Critical
{'epsilon': 0.2, 'clean_acc': 0.9111}
ADV_0069
Adversarial_Attack
[0.4663, -0.058, 0.4219, -1.0561, 1.3903, 1.7256, 2.3637, 1.6712, 0.6258, 0.1726, -1.5537, 0.4369, 0.316, 0.9814, 0.4092]
1
0.9111
0.9
0.9
0.9
0.9
Tiny input perturbation (ε=0.2) caused misclassification.
Model has sensitive decision boundaries, not robust to small noise.
Adversarial training, input preprocessing, ensemble defenses, certified robustness.
Critical
{'epsilon': 0.2, 'clean_acc': 0.9111}
ADV_0070
Adversarial_Attack
[1.4557, -1.1478, -0.9087, 1.3283, 0.5752, -0.2666, -0.6013, -0.0198, 0.08, -1.6924, -0.1453, -0.5345, -0.8332, -1.6511, 0.5472]
0
0.9111
0.9
0.9
0.9
0.9
Tiny input perturbation (ε=0.2) caused misclassification.
Model has sensitive decision boundaries, not robust to small noise.
Adversarial training, input preprocessing, ensemble defenses, certified robustness.
Critical
{'epsilon': 0.2, 'clean_acc': 0.9111}
ADV_0071
Adversarial_Attack
[1.2718, 0.3036, 1.6857, -0.7985, 0.366, 0.5762, -0.3634, -0.6868, 0.4707, 0.4908, -0.4235, -0.2671, -0.0752, 0.2266, -1.0589]
1
0.9111
0.9
0.9
0.9
0.9
Tiny input perturbation (ε=0.2) caused misclassification.
Model has sensitive decision boundaries, not robust to small noise.
Adversarial training, input preprocessing, ensemble defenses, certified robustness.
Critical
{'epsilon': 0.2, 'clean_acc': 0.9111}
ADV_0072
Adversarial_Attack
[-0.5071, 0.3583, 0.2804, 0.0458, -1.306, -1.595, -0.4453, 0.87, 0.4615, 0.6722, 0.264, -0.4735, -0.1117, 0.2057, -0.779]
0
0.9111
0.9
0.9
0.9
0.9
Tiny input perturbation (ε=0.2) caused misclassification.
Model has sensitive decision boundaries, not robust to small noise.
Adversarial training, input preprocessing, ensemble defenses, certified robustness.
Critical
{'epsilon': 0.2, 'clean_acc': 0.9111}
ADV_0073
Adversarial_Attack
[-1.5679, 0.4081, 1.1677, 2.5054, -1.9319, -1.5049, 0.0453, 2.2895, 0.1795, 0.1867, 0.4013, -0.5608, -0.3467, -0.3587, -0.44]
0
0.9111
0.9
0.9
0.9
0.9
Tiny input perturbation (ε=0.2) caused misclassification.
Model has sensitive decision boundaries, not robust to small noise.
Adversarial training, input preprocessing, ensemble defenses, certified robustness.
Critical
{'epsilon': 0.2, 'clean_acc': 0.9111}
ADV_0074
Adversarial_Attack
[-0.5469, -1.1345, -0.6106, 0.6841, 1.969, 0.1821, 0.0632, 0.4239, -0.6563, -0.9391, 0.1626, -0.6037, -0.1722, -1.6894, -0.2661]
1
0.9111
0.9
0.9
0.9
0.9
Tiny input perturbation (ε=0.2) caused misclassification.
Model has sensitive decision boundaries, not robust to small noise.
Adversarial training, input preprocessing, ensemble defenses, certified robustness.
Critical
{'epsilon': 0.2, 'clean_acc': 0.9111}
ADV_0075
Adversarial_Attack
[1.4597, -0.8863, -0.1876, 0.9429, 0.8741, -0.2518, -0.6569, 0.8139, 0.6612, 0.2254, -0.1479, 0.1457, -0.5455, -1.2468, -1.4758]
0
0.9111
0.9
0.9
0.9
0.9
Tiny input perturbation (ε=0.2) caused misclassification.
Model has sensitive decision boundaries, not robust to small noise.
Adversarial training, input preprocessing, ensemble defenses, certified robustness.
Critical
{'epsilon': 0.2, 'clean_acc': 0.9111}
ADV_0076
Adversarial_Attack
[1.023, 1.1074, -0.4989, -1.0598, -0.9363, 0.5845, 1.7041, -0.3043, 1.1717, -3.7134, 1.1993, -1.0311, 2.2214, 2.2331, -0.2094]
0
0.9111
0.9
0.9
0.9
0.9
Tiny input perturbation (ε=0.2) caused misclassification.
Model has sensitive decision boundaries, not robust to small noise.
Adversarial training, input preprocessing, ensemble defenses, certified robustness.
Critical
{'epsilon': 0.2, 'clean_acc': 0.9111}
ADV_0077
Adversarial_Attack
[1.0083, 2.0002, -0.8539, -0.808, -0.6219, 0.8633, 2.2147, -0.169, -0.1333, 0.1906, 0.9003, 0.7754, 1.3567, -0.2403, -0.9442]
0
0.9111
0.9
0.9
0.9
0.9
Tiny input perturbation (ε=0.2) caused misclassification.
Model has sensitive decision boundaries, not robust to small noise.
Adversarial training, input preprocessing, ensemble defenses, certified robustness.
Critical
{'epsilon': 0.2, 'clean_acc': 0.9111}
ADV_0078
Adversarial_Attack
[0.8898, -1.0821, 0.355, -3.5034, -0.4871, 0.8519, 0.3298, -0.5764, 0.0951, -0.6234, 0.2755, 0.0745, -0.7898, 0.2392, 0.8946]
1
0.9111
0.9
0.9
0.9
0.9
Tiny input perturbation (ε=0.2) caused misclassification.
Model has sensitive decision boundaries, not robust to small noise.
Adversarial training, input preprocessing, ensemble defenses, certified robustness.
Critical
{'epsilon': 0.2, 'clean_acc': 0.9111}
ADV_0079
Adversarial_Attack
[1.5653, 0.9921, 2.5906, 0.14, 1.062, -0.3151, 1.7219, -1.3547, -0.4285, -0.5942, -0.2354, -1.7215, 1.639, 0.1683, 0.5916]
1
0.9111
0.9
0.9
0.9
0.9
Tiny input perturbation (ε=0.2) caused misclassification.
Model has sensitive decision boundaries, not robust to small noise.
Adversarial training, input preprocessing, ensemble defenses, certified robustness.
Critical
{'epsilon': 0.2, 'clean_acc': 0.9111}
ADV_0080
Adversarial_Attack
[-0.8028, 0.3346, -0.2943, -0.993, -1.0885, 1.0768, -0.7201, -1.6558, 0.1937, -0.1182, -1.0857, -0.9526, -0.3045, -0.4426, 0.0852]
1
0.9111
0.9
0.9
0.9
0.9
Tiny input perturbation (ε=0.2) caused misclassification.
Model has sensitive decision boundaries, not robust to small noise.
Adversarial training, input preprocessing, ensemble defenses, certified robustness.
Critical
{'epsilon': 0.2, 'clean_acc': 0.9111}
ADV_0081
Adversarial_Attack
[-2.2351, -1.1683, 1.5763, 0.2934, 0.9912, 0.5532, -0.5543, 1.6768, -0.8841, 0.3998, -0.046, -0.35, -0.4407, -1.1196, 0.5168]
1
0.9111
0.9
0.9
0.9
0.9
Tiny input perturbation (ε=0.2) caused misclassification.
Model has sensitive decision boundaries, not robust to small noise.
Adversarial training, input preprocessing, ensemble defenses, certified robustness.
Critical
{'epsilon': 0.2, 'clean_acc': 0.9111}
ADV_0082
Adversarial_Attack
[-0.0363, -0.227, -0.3655, -0.1822, 1.1189, 0.5011, 1.758, 0.7798, -0.5738, -0.6422, 0.5294, -0.9107, 1.5473, 0.2588, 0.0063]
1
0.9111
0.9
0.9
0.9
0.9
Tiny input perturbation (ε=0.2) caused misclassification.
Model has sensitive decision boundaries, not robust to small noise.
Adversarial training, input preprocessing, ensemble defenses, certified robustness.
Critical
{'epsilon': 0.2, 'clean_acc': 0.9111}
ADV_0083
Adversarial_Attack
[1.9128, 0.0618, 1.3982, -0.8696, -0.9201, -0.2446, 0.4349, 0.9217, 2.1102, 0.5658, -0.385, -1.5657, 0.7849, 1.9893, 1.0774]
1
0.9111
0.9
0.9
0.9
0.9
Tiny input perturbation (ε=0.2) caused misclassification.
Model has sensitive decision boundaries, not robust to small noise.
Adversarial training, input preprocessing, ensemble defenses, certified robustness.
Critical
{'epsilon': 0.2, 'clean_acc': 0.9111}
ADV_0084
Adversarial_Attack
[0.5594, 0.2046, 0.9022, 1.0075, -0.5, -0.51, -0.4556, -0.8925, -0.3496, 0.7659, -0.6349, 1.5523, -0.643, 0.1524, 0.4913]
0
0.9111
0.9
0.9
0.9
0.9
Tiny input perturbation (ε=0.2) caused misclassification.
Model has sensitive decision boundaries, not robust to small noise.
Adversarial training, input preprocessing, ensemble defenses, certified robustness.
Critical
{'epsilon': 0.2, 'clean_acc': 0.9111}
ADV_0085
Adversarial_Attack
[0.0366, 0.2364, -0.3453, -1.3172, -0.275, -0.2738, 0.7358, -0.356, -0.5035, -0.5173, -0.9912, 0.5372, -0.5552, -0.2361, 0.4547]
0
0.9111
0.9
0.9
0.9
0.9
Tiny input perturbation (ε=0.2) caused misclassification.
Model has sensitive decision boundaries, not robust to small noise.
Adversarial training, input preprocessing, ensemble defenses, certified robustness.
Critical
{'epsilon': 0.2, 'clean_acc': 0.9111}
ADV_0086
Adversarial_Attack
[1.4273, 0.4948, -1.0943, -0.1405, 0.4276, 1.2152, 0.9628, 1.3367, 1.6519, 0.2445, -0.9679, -2.0563, 1.1723, -0.1147, -1.0404]
1
0.9111
0.9
0.9
0.9
0.9
Tiny input perturbation (ε=0.2) caused misclassification.
Model has sensitive decision boundaries, not robust to small noise.
Adversarial training, input preprocessing, ensemble defenses, certified robustness.
Critical
{'epsilon': 0.2, 'clean_acc': 0.9111}
ADV_0087
Adversarial_Attack
[-1.1883, -0.2208, -0.2428, 0.6643, -0.9285, 1.8943, 0.7511, -0.2227, -0.4044, -0.7816, 0.4493, 0.6715, -0.7214, -1.0768, -0.2164]
0
0.9111
0.9
0.9
0.9
0.9
Tiny input perturbation (ε=0.2) caused misclassification.
Model has sensitive decision boundaries, not robust to small noise.
Adversarial training, input preprocessing, ensemble defenses, certified robustness.
Critical
{'epsilon': 0.2, 'clean_acc': 0.9111}
ADV_0088
Adversarial_Attack
[0.2175, 1.173, -2.162, -0.744, 0.9998, -0.809, 0.9041, -1.6161, -1.0656, 1.6779, 1.2326, -0.2953, 2.208, 0.5131, 0.5035]
0
0.9111
0.9
0.9
0.9
0.9
Tiny input perturbation (ε=0.2) caused misclassification.
Model has sensitive decision boundaries, not robust to small noise.
Adversarial training, input preprocessing, ensemble defenses, certified robustness.
Critical
{'epsilon': 0.2, 'clean_acc': 0.9111}
ADV_0089
Adversarial_Attack
[-0.1939, 0.1697, 0.7073, -0.5169, 3.0299, 0.1519, 0.4386, -2.2507, -1.9645, -1.6602, -1.403, 1.3957, 0.6512, 0.5212, 1.1825]
0
0.9111
0.9
0.9
0.9
0.9
Tiny input perturbation (ε=0.2) caused misclassification.
Model has sensitive decision boundaries, not robust to small noise.
Adversarial training, input preprocessing, ensemble defenses, certified robustness.
Critical
{'epsilon': 0.2, 'clean_acc': 0.9111}
ADV_0090
Adversarial_Attack
[-1.3677, -0.2632, 1.2894, 0.408, 0.0478, 1.4153, 0.1866, 0.8518, -0.4412, 0.1998, 0.3483, 1.6769, -1.3116, -1.6024, -1.0558]
0
0.9111
0.9
0.9
0.9
0.9
Tiny input perturbation (ε=0.2) caused misclassification.
Model has sensitive decision boundaries, not robust to small noise.
Adversarial training, input preprocessing, ensemble defenses, certified robustness.
Critical
{'epsilon': 0.2, 'clean_acc': 0.9111}
ADV_0091
Adversarial_Attack
[-1.6696, 1.3831, 0.7643, 0.1123, 0.6982, 1.6411, -0.6002, 0.1912, 0.0614, 0.2683, 0.61, -1.7412, 1.952, -0.522, -0.8744]
0
0.9111
0.9
0.9
0.9
0.9
Tiny input perturbation (ε=0.2) caused misclassification.
Model has sensitive decision boundaries, not robust to small noise.
Adversarial training, input preprocessing, ensemble defenses, certified robustness.
Critical
{'epsilon': 0.2, 'clean_acc': 0.9111}
ADV_0092
Adversarial_Attack
[-0.3026, 0.748, -0.2866, 0.8149, -1.4428, -1.8047, -0.1014, 1.1301, -0.462, 0.8431, 0.3391, -0.4116, -0.5087, -0.7968, 2.1396]
0
0.9111
0.9
0.9
0.9
0.9
Tiny input perturbation (ε=0.2) caused misclassification.
Model has sensitive decision boundaries, not robust to small noise.
Adversarial training, input preprocessing, ensemble defenses, certified robustness.
Critical
{'epsilon': 0.2, 'clean_acc': 0.9111}
ADV_0093
Adversarial_Attack
[0.2775, 1.1152, -0.6108, -0.0465, -0.131, -0.4678, 1.3477, -0.1697, 0.1307, -0.9476, 0.8274, -0.829, 1.9666, 0.9551, -1.0072]
0
0.9111
0.9
0.9
0.9
0.9
Tiny input perturbation (ε=0.2) caused misclassification.
Model has sensitive decision boundaries, not robust to small noise.
Adversarial training, input preprocessing, ensemble defenses, certified robustness.
Critical
{'epsilon': 0.2, 'clean_acc': 0.9111}
ADV_0094
Adversarial_Attack
[1.0749, -0.4892, 1.2301, 1.434, -0.0244, -1.0298, 0.0292, -0.6322, -0.1525, -0.0795, -0.5729, -0.6613, -0.3933, -1.2665, -0.0823]
0
0.9111
0.9
0.9
0.9
0.9
Tiny input perturbation (ε=0.2) caused misclassification.
Model has sensitive decision boundaries, not robust to small noise.
Adversarial training, input preprocessing, ensemble defenses, certified robustness.
Critical
{'epsilon': 0.2, 'clean_acc': 0.9111}
ADV_0095
Adversarial_Attack
[-0.0806, -0.5567, 0.0949, -1.4464, -1.1124, 1.1086, -1.8757, 1.0728, 1.5537, -0.3726, 0.2676, -0.1951, -1.1046, 0.1841, -1.2247]
1
0.9111
0.9
0.9
0.9
0.9
Tiny input perturbation (ε=0.2) caused misclassification.
Model has sensitive decision boundaries, not robust to small noise.
Adversarial training, input preprocessing, ensemble defenses, certified robustness.
Critical
{'epsilon': 0.2, 'clean_acc': 0.9111}
ADV_0096
Adversarial_Attack
[0.4934, -0.1295, -0.4348, -0.4052, 0.3092, -0.0347, 0.5755, -0.7079, -0.6762, -0.4811, 0.7785, 0.3829, 0.4743, -0.3304, 2.2193]
1
0.9111
0.9
0.9
0.9
0.9
Tiny input perturbation (ε=0.2) caused misclassification.
Model has sensitive decision boundaries, not robust to small noise.
Adversarial training, input preprocessing, ensemble defenses, certified robustness.
Critical
{'epsilon': 0.2, 'clean_acc': 0.9111}
ADV_0097
Adversarial_Attack
[0.3915, -1.356, 1.7571, -1.4202, -1.1359, 0.0703, -0.7594, 1.1278, 1.0368, -0.0917, -0.7943, 1.323, -1.6473, 0.1189, -1.3321]
1
0.9111
0.9
0.9
0.9
0.9
Tiny input perturbation (ε=0.2) caused misclassification.
Model has sensitive decision boundaries, not robust to small noise.
Adversarial training, input preprocessing, ensemble defenses, certified robustness.
Critical
{'epsilon': 0.2, 'clean_acc': 0.9111}
ADV_0098
Adversarial_Attack
[-0.5843, 0.7378, 0.9378, -0.3633, -0.5435, 0.5889, -1.9421, 0.0217, 0.6799, -0.0974, -0.1499, -0.6725, -0.0819, 0.2387, 2.5841]
0
0.9111
0.9
0.9
0.9
0.9
Tiny input perturbation (ε=0.2) caused misclassification.
Model has sensitive decision boundaries, not robust to small noise.
Adversarial training, input preprocessing, ensemble defenses, certified robustness.
Critical
{'epsilon': 0.2, 'clean_acc': 0.9111}
ADV_0099
Adversarial_Attack
[-0.197, 1.5189, 1.2478, 0.7976, -0.8843, -0.5534, 0.1652, 0.551, 0.5848, -1.2656, 0.5357, -0.5504, 0.7953, 0.7792, -1.9652]
0
0.9111
0.9
0.9
0.9
0.9
Tiny input perturbation (ε=0.2) caused misclassification.
Model has sensitive decision boundaries, not robust to small noise.
Adversarial training, input preprocessing, ensemble defenses, certified robustness.
Critical
{'epsilon': 0.2, 'clean_acc': 0.9111}