Page cover image

📚LoRACSE

LoRACSE: Contrastive Learning of Sentence Embedding using LoRA

  • Proposed LoRACSE, a novel contrastive learning method for sentence embedding using LoRA. With 0.067% parameters of the SotA model, it remains comparable performance. Enhanced training efficiency while reducing computational resources.

  • Evaluated our method on different downstream tasks, including sentiment analysis and text classification with BERT and other contrastive learning models, showcasing comparable performance to SotA models.

https://github.com/tic-top/LoraCSE

Last updated