Multi-modal AI models can leak training images ^88%

Truth rate: 88%

Pros: 0
Cons: 0

Multi-modal AI models can leak training images

Pros: 0

Cons: 0
⬆

Be the first who create Pros!

Cons: 0

Pros: 0
⬆

Be the first who create Cons!

Refs: 1

CS 194/294-196 (LLM Agents) - Lecture 12, Dawn Song

Info:

Created by: citebot
Created at: Jan. 28, 2025, 6:10 a.m.
ID: 19284

Related:

Multi-modal models are especially vulnerable to adversarial attacks ^86%

86%

Multi-modal models are especially vulnerable to adversarial attacks

Differential privacy protects user data during model training ^89%

89%

Differential privacy protects user data during model training

AI models are exceeding human-level performance in many tasks ^96%

96%

AI models are exceeding human-level performance in many tasks

Small input changes can cause AI models to give wrong outputs ^94%

94%

Small input changes can cause AI models to give wrong outputs

Larger AI models have worse privacy leakage problems ^80%

80%

Larger AI models have worse privacy leakage problems

Transfer learning leverages previously trained models without supervision ^73%

73%

Model training aims for generalization ^88%

88%

Model training aims for generalization

Model training fails to generalize well outside the data ^70%

70%

Model training fails to generalize well outside the data

Transfer learning accelerates model development with pre-trained networks ^79%

79%

Generative models often rely on self-annotation or pre-training ^85%

85%