Xiaoying Fan

RepCali: High Efficient Fine-tuning Via Representation Calibration in Latent Space for Pre-trained Language Models

2025 · arXiv (Cornell University)

Fine-tuning pre-trained language models (PLMs) has become a dominant paradigm in applying PLMs to downstream tasks. However, with limited fine-tuning, PLMs still struggle with the discrepancies between the representation obtained from the PLMs' encoder and …