pre-trained model
A model that has undergone initial training on a large dataset and can be fine-tuned on a smaller task-specific dataset, benefiting from the learned features and knowledge to expedite training and improve performance.
- Buffer layers for Test-Time Adaptation
- Hybrid Re-matching for Continual Learning with Parameter-Efficient Tuning
- Low-Rank Head Avatar Personalization with Registers
- On Minimax Estimation of Parameters in Softmax-Contaminated Mixture of Experts
- PAID: Pairwise Angular-Invariant Decomposition for Continual Test-Time Adaptation
- Synthetic Series-Symbol Data Generation for Time Series Foundation Models
- Token Bottleneck: One Token to Remember Dynamics
- Understanding the Gain from Data Filtering in Multimodal Contrastive Learning