Large Pre Trained Model
Large pre-trained models (LPMs) are massive neural networks trained on enormous datasets, aiming to achieve strong generalization across diverse downstream tasks with minimal further training. Current research emphasizes efficient fine-tuning techniques, such as prompt engineering, low-rank adaptation (e.g., LoRA, SVFit), and sparse parameter updates, to reduce computational costs and improve model adaptability while addressing issues like overfitting and catastrophic forgetting. This field is significant due to LPMs' transformative impact on various applications, from natural language processing and computer vision to robotics and education, driving advancements in both theoretical understanding and practical deployment of AI systems.
Papers
Semantic Residual Prompts for Continual Learning
Martin Menabue, Emanuele Frascaroli, Matteo Boschini, Enver Sangineto, Lorenzo Bonicelli, Angelo Porrello, Simone Calderara
Learning with Noisy Foundation Models
Hao Chen, Jindong Wang, Zihan Wang, Ran Tao, Hongxin Wei, Xing Xie, Masashi Sugiyama, Bhiksha Raj
A Segmentation Foundation Model for Diverse-type Tumors
Jianhao Xie, Ziang Zhang, Guibo Luo, Yuesheng Zhu
Empirical Analysis of Efficient Fine-Tuning Methods for Large Pre-Trained Language Models
Nigel Doering, Cyril Gorlla, Trevor Tuttle, Adhvaith Vijay
Tiny Time Mixers (TTMs): Fast Pre-trained Models for Enhanced Zero/Few-Shot Forecasting of Multivariate Time Series
Vijay Ekambaram, Arindam Jati, Pankaj Dayama, Sumanta Mukherjee, Nam H. Nguyen, Wesley M. Gifford, Chandra Reddy, Jayant Kalagnanam