Long Span
"Long span" research addresses the limitations of current models in processing and generating lengthy sequences of data, whether text, audio, or video. Current efforts focus on improving large language models (LLMs) and other deep learning architectures like transformers (including Longformer and variations) and LSTMs to handle longer contexts effectively, often employing techniques like coreference resolution, hierarchical attention, and efficient attention mechanisms. This research is crucial for advancing natural language processing, improving video and audio analysis, and enabling more sophisticated applications in diverse fields such as medical diagnosis, legal document processing, and personalized search.
Papers
Manta: Enhancing Mamba for Few-Shot Action Recognition of Long Sub-Sequence
Wenbo Huang, Jinghui Zhang, Guang Li, Lei Zhang, Shuoyuan Wang, Fang Dong, Jiahui Jin, Takahiro Ogawa, Miki Haseyama
Breaking the Stage Barrier: A Novel Single-Stage Approach to Long Context Extension for Large Language Models
Haoran Lian, Junmin Chen, Wei Huang, Yizhe Xiong, Wenping Hu, Guiguang Ding, Hui Chen, Jianwei Niu, Zijia Lin, Fuzheng Zhang, Di Zhang