Low rank adaptation

Technique used in machine learning, particularly in the context of fine-tuning large language models. It involves adapting a pre-trained model to a specific task by introducing low-rank matrices into the model's architecture, reducing the number of parameters that need to be updated during fine-tuning. This approach is beneficial for efficiently adapting very large models without retraining the entire model, making it suitable for resource-constrained environments.

Low rank adaptation was discovered on April 15th 2025 and it currently has a search volume of 3.6K with a growth of +2600%.