memorization
In the context of AI, particularly neural networks, memorization refers to the model's ability to recall specific examples from the training data rather than generalizing or learning underlying patterns, often resulting in overfitting.
- A Black-Box Debiasing Framework for Conditional Sampling
- A Closer Look at Model Collapse: From a Generalization-to-Memorization Perspective
- Adjusting Initial Noise to Mitigate Memorization in Text-to-Image Diffusion Models
- Data Mixing Can Induce Phase Transitions in Knowledge Acquisition
- Feedback Guidance of Diffusion Models
- Flatness is Necessary, Neural Collapse is Not: Rethinking Generalization via Grokking
- For Better or for Worse, Transformers Seek Patterns for Memorization
- Gradient Variance Reveals Failure Modes in Flow-Based Generative Models
- Impact of Layer Norm on Memorization and Generalization in Transformers
- Keeping an Eye on LLM Unlearning: The Hidden Risk and Remedy
- LOMIA: Label-Only Membership Inference Attacks against Pre-trained Large Vision-Language Models
- On the Edge of Memorization in Diffusion Models
- Quantifying Cross-Modality Memorization in Vision-Language Models
- Rethinking the Role of Verbatim Memorization in LLM Privacy
- Tighter CMI-Based Generalization Bounds via Stochastic Projection and Quantization
- TokenSwap: A Lightweight Method to Disrupt Memorized Sequences in LLMs
- Why Diffusion Models Don’t Memorize: The Role of Implicit Dynamical Regularization in Training
- Why Diffusion Models Don’t Memorize: The Role of Implicit Dynamical Regularization in Training