Zobrazeno 1 - 10
of 278
pro vyhledávání: '"Blaschko, Matthew"'
Autor:
Liu, Enshu, Zhu, Junyi, Lin, Zinan, Ning, Xuefei, Blaschko, Matthew B., Yan, Shengen, Dai, Guohao, Yang, Huazhong, Wang, Yu
The rapid advancement of large language models (LLMs) has led to architectures with billions to trillions of parameters, posing significant deployment challenges due to their substantial demands on memory, processing power, and energy consumption. Sp
Externí odkaz:
http://arxiv.org/abs/2407.00945
Autor:
Zhu, Junyi, Liu, Shuochen, Yu, Yu, Tang, Bo, Yan, Yibo, Li, Zhiyu, Xiong, Feiyu, Xu, Tong, Blaschko, Matthew B.
Large language models (LLMs) excel in generating coherent text, but they often struggle with context awareness, leading to inaccuracies in tasks requiring faithful adherence to provided information. We introduce FastMem, a novel method designed to en
Externí odkaz:
http://arxiv.org/abs/2406.16069
Autor:
Ning, Xuefei, Wang, Zifu, Li, Shiyao, Lin, Zinan, Yao, Peiran, Fu, Tianyu, Blaschko, Matthew B., Dai, Guohao, Yang, Huazhong, Wang, Yu
Teaching to improve student models (e.g., knowledge distillation) is an extensively studied methodology in LLMs. However, for humans, teaching not only improves students but also improves teachers. We ask: Can LLMs also learn by teaching (LbT)? If ye
Externí odkaz:
http://arxiv.org/abs/2406.14629
Autor:
Tian, Chang, Blaschko, Matthew B., Yin, Wenpeng, Xing, Mingzhe, Yue, Yinliang, Moens, Marie-Francine
Fine-grained category discovery using only coarse-grained supervision is a cost-effective yet challenging task. Previous training methods focus on aligning query samples with positive samples and distancing them from negatives. They often neglect int
Externí odkaz:
http://arxiv.org/abs/2406.13103
Autor:
Van Landeghem, Jordy, Maity, Subhajit, Banerjee, Ayan, Blaschko, Matthew, Moens, Marie-Francine, Lladós, Josep, Biswas, Sanket
This work explores knowledge distillation (KD) for visually-rich document (VRD) applications such as document layout analysis (DLA) and document image classification (DIC). While VRD research is dependent on increasingly sophisticated and cumbersome
Externí odkaz:
http://arxiv.org/abs/2406.08226
Autor:
Hamed, Omar, Bakkali, Souhail, Moens, Marie-Francine, Blaschko, Matthew, Van Landeghem, Jordy
This work addresses the need for a balanced approach between performance and efficiency in scalable production environments for visually-rich document understanding (VDU) tasks. Currently, there is a reliance on large document foundation models that
Externí odkaz:
http://arxiv.org/abs/2405.12705
Learning from multiple modalities, such as audio and video, offers opportunities for leveraging complementary information, enhancing robustness, and improving contextual understanding and performance. However, combining such modalities presents chall
Externí odkaz:
http://arxiv.org/abs/2405.07930
Computed Tomography (CT) is pivotal in industrial quality control and medical diagnostics. Sparse-view CT, offering reduced ionizing radiation, faces challenges due to its under-sampled nature, leading to ill-posed reconstruction problems. Recent adv
Externí odkaz:
http://arxiv.org/abs/2405.02509
Autor:
Liu, Enshu, Zhu, Junyi, Lin, Zinan, Ning, Xuefei, Blaschko, Matthew B., Yekhanin, Sergey, Yan, Shengen, Dai, Guohao, Yang, Huazhong, Wang, Yu
Diffusion Models (DM) and Consistency Models (CM) are two types of popular generative models with good generation quality on various tasks. When training DM and CM, intermediate weight checkpoints are not fully utilized and only the last converged ch
Externí odkaz:
http://arxiv.org/abs/2404.02241
Last couple of years have witnessed a tremendous progress in self-supervised learning (SSL), the success of which can be attributed to the introduction of useful inductive biases in the learning process to learn meaningful visual representations whil
Externí odkaz:
http://arxiv.org/abs/2402.14957