Unverified50% confidenceFactExact time
The transferability of adversarial attacks refers to the ability of adversarial examples generated for one model to mislead another model with a different structure or different training data with high probability, arising from similarities in the decision boundaries of different models.
1
Sources
50%
Confidence
Long-term
Relevance
10/27/2026
First Seen
Sources
Related Entities
Related Claims
UnverifiedSensitive information from training data can be extracted by reverse-engineering model weights, which is known as a model inversion attack.78% similarUnverified预训练模型迁移复用的假设是模型在源域学到的特征具有跨域泛化能力，但高层语义特征可能存在显著的域偏移73% similarVerifiedTransfer learning refers to migrating model parameters pre-trained on large-scale datasets to a specific small-sample task for fine-tuning.73% similarUnverifiedAdversarial examples refer to inputs that attackers deliberately construct to deceive model classification.72% similarUnverifiedModel poisoning involves implanting backdoors or biases during the training phase of an AI model, causing the model to produce erroneous judgments under specific conditions.71% similar
Cite This Claim
Stable URI
https://kongchang.com/claim/1102182API
curl https://kongchang.com/api/v1/knowledge/claims/1102182MCP
get_claim(id=1102182)