File size: 2,435 Bytes
95a9764 | 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 | {
"term": "AdversarialRobustness",
"role": "concept",
"parent_concepts": [
"Helping"
],
"layer": 3,
"domain": "SafetyAndSecurity",
"definition": "The ability of models to maintain correct behavior when faced with adversarial inputs or attacks",
"definition_source": "SUMO",
"aliases": [
"RobustML",
"AdversarialDefense",
"AttackResistance"
],
"wordnet": {
"synsets": [],
"canonical_synset": "",
"lemmas": [],
"pos": "noun"
},
"relationships": {
"related": [
"AdversarialImage",
"AdversarialTraining",
"ModelSecurity"
],
"antonyms": [],
"has_part": [],
"part_of": []
},
"safety_tags": {
"risk_level": "low",
"impacts": [],
"treaty_relevant": false,
"harness_relevant": true
},
"training_hints": {
"positive_examples": [
"Adversarial training improves robustness by including adversarial examples during training.",
"The model achieved certified robustness guarantees against perturbations within an epsilon ball.",
"Adversarial robustness often trades off against standard accuracy.",
"Randomized smoothing provides provable robustness certificates for classifiers.",
"Adversarial robustness testing is essential for safety-critical computer vision deployments.",
"The robust model maintained accuracy even under projected gradient descent attacks.",
"Adversarial robustness requires defending against attacks that haven't been seen yet.",
"The model's adversarial robustness was validated through comprehensive red-teaming.",
"Adversarial training with PGD attacks significantly improved model robustness.",
"Certified robustness provides mathematical guarantees about model behavior under attack."
],
"negative_examples": [
"The model is accurate.",
"It works well.",
"The system is reliable.",
"The model performs well.",
"The system is stable.",
"It is robust.",
"The model is strong.",
"The system is secure.",
"It works correctly.",
"The model is good."
],
"disambiguation": "Specifically resistance to adversarial attacks, not general robustness or reliability"
},
"is_category_lens": true,
"child_count": 0,
"meld_source": {
"meld_id": "org.hatcat/multimodal-safety@0.2.0",
"applied_at": "2025-12-10T20:54:17.844067Z",
"pack_version": "6.0.0"
}
} |