model updates
Model updates refer to the adjustments made to the parameters of an AI model based on new data or feedback. Regular updates are essential for maintaining the relevance and performance of models, especially in dynamic environments where data distributions can shift over time.
- $\boldsymbol{\lambda}$-Orthogonality Regularization for Compatible Representation Learning
- Exploiting Task Relationships in Continual Learning via Transferability-Aware Task Embeddings
- Layer-wise Update Aggregation with Recycling for Communication-Efficient Federated Learning
- Lifelong Test-Time Adaptation via Online Learning in Tracked Low-Dimensional Subspace
- Robust Federated Finetuning of LLMs via Alternating Optimization of LoRA