中文
相关论文

相关论文: Trust the uncertain teacher: distilling dark knowl…

200 篇论文

Pretrained models have become a commodity and offer strong results on a broad range of tasks. In this work, we focus on classification and seek to learn a unique encoder able to take from several complementary pretrained models. We aim at…

计算机视觉与模式识别 · 计算机科学 2024-08-12 Mert Bulent Sariyildiz , Philippe Weinzaepfel , Thomas Lucas , Diane Larlus , Yannis Kalantidis

Knowledge distillation aims at transferring knowledge acquired in one model (a teacher) to another model (a student) that is typically smaller. Previous approaches can be expressed as a form of training the student to mimic output…

计算机视觉与模式识别 · 计算机科学 2019-05-02 Wonpyo Park , Dongju Kim , Yan Lu , Minsu Cho

Knowledge distillation is a key technique for transferring the capabilities of large language models (LLMs) into smaller, more efficient student models. Existing distillation approaches often overlook two critical factors: the learning…

机器学习 · 计算机科学 2026-05-13 Jincheng Cao , Fanzhi Zeng , Leqi Liu , Aryan Mokhtari

Calibrated estimates of uncertainty are critical for many real-world computer vision applications of deep learning. While there are several widely-used uncertainty estimation methods, dropout inference stands out for its simplicity and…

计算机视觉与模式识别 · 计算机科学 2020-11-09 Yichen Shen , Zhilu Zhang , Mert R. Sabuncu , Lin Sun

Reliable uncertainty quantification remains a major obstacle to the deployment of deep learning models under distributional shift. Existing post-hoc approaches that retrofit pretrained models either inherit misplaced confidence or merely…

机器学习 · 计算机科学 2025-09-30 Charmaine Barker , Daniel Bethell , Simos Gerasimou

Knowledge distillation constitutes a simple yet effective way to improve the performance of a compact student network by exploiting the knowledge of a more powerful teacher. Nevertheless, the knowledge distillation literature remains…

计算机视觉与模式识别 · 计算机科学 2022-02-11 Shuxuan Guo , Jose M. Alvarez , Mathieu Salzmann

Traditional knowledge distillation (KD) relies on a proficient teacher trained on the target task, which is not always available. In this setting, cross-task distillation can be used, enabling the use of any teacher model trained on a…

计算机视觉与模式识别 · 计算机科学 2024-11-28 Dylan Auty , Roy Miles , Benedikt Kolbeinsson , Krystian Mikolajczyk

Knowledge distillation field delicately designs various types of knowledge to shrink the performance gap between compact student and large-scale teacher. These existing distillation approaches simply focus on the improvement of…

计算机视觉与模式识别 · 计算机科学 2021-09-28 Xuanyang Zhang , Xiangyu Zhang , Jian Sun

Benefiting from well-trained deep neural networks (DNNs), model compression have captured special attention for computing resource limited equipment, especially edge devices. Knowledge distillation (KD) is one of the widely used compression…

机器学习 · 计算机科学 2024-06-06 Jinyin Chen , Xiaoming Zhao , Haibin Zheng , Xiao Li , Sheng Xiang , Haifeng Guo

Causal language models have demonstrated remarkable capabilities, but their size poses significant challenges for deployment in resource-constrained environments. Knowledge distillation, a widely-used technique for transferring knowledge…

机器学习 · 计算机科学 2025-03-03 Makoto Shing , Kou Misaki , Han Bao , Sho Yokoi , Takuya Akiba

Knowledge distillation involves transferring knowledge from large, cumbersome teacher models to more compact student models. The standard approach minimizes the Kullback-Leibler (KL) divergence between the probabilistic outputs of a teacher…

计算机视觉与模式识别 · 计算机科学 2025-05-14 Nikolaos Giakoumoglou , Tania Stathaki

Knowledge distillation has attracted a great deal of interest recently to compress pre-trained language models. However, existing knowledge distillation methods suffer from two limitations. First, the student model simply imitates the…

计算与语言 · 计算机科学 2023-05-18 Siyue Wu , Hongzhan Chen , Xiaojun Quan , Qifan Wang , Rui Wang

The outpouring of various pre-trained models empowers knowledge distillation by providing abundant teacher resources, but there lacks a developed mechanism to utilize these teachers adequately. With a massive model repository composed of…

机器学习 · 计算机科学 2022-09-29 Su Lu , Han-Jia Ye , De-Chuan Zhan

Recently, deep learning-based models have been widely studied for click-through rate (CTR) prediction and lead to improved prediction accuracy in many industrial applications. However, current research focuses primarily on building complex…

机器学习 · 计算机科学 2023-07-06 Jieming Zhu , Jinyang Liu , Weiqi Li , Jincai Lai , Xiuqiang He , Liang Chen , Zibin Zheng

Knowledge distillation from large language models (LLMs) assumes that the teacher's output distribution is a high-quality training signal. On reasoning tasks, this assumption is frequently violated. A model's intermediate representations…

计算与语言 · 计算机科学 2026-03-16 Ryan Brown , Chris Russell

Knowledge distillation has become a cornerstone in modern machine learning systems, celebrated for its ability to transfer knowledge from a large, complex teacher model to a more efficient student model. Traditionally, this process is…

密码学与安全 · 计算机科学 2026-01-13 Chen Wu , Qian Ma , Prasenjit Mitra , Sencun Zhu

Knowledge Distillation (KD) aims at transferring the knowledge of a well-performed neural network (the {\it teacher}) to a weaker one (the {\it student}). A peculiar phenomenon is that a more accurate model doesn't necessarily teach better,…

机器学习 · 计算机科学 2022-10-14 Xin-Chun Li , Wen-Shu Fan , Shaoming Song , Yinchuan Li , Bingshuai Li , Yunfeng Shao , De-Chuan Zhan

Knowledge distillation is a popular machine learning technique that aims to transfer knowledge from a large 'teacher' network to a smaller 'student' network and improve the student's performance by training it to emulate the teacher. In…

机器学习 · 计算机科学 2022-10-19 Sushil Thapa

Knowledge distillation leverages a teacher model to improve the training of a student model. A persistent challenge is that a better teacher does not always yield a better student, to which a common mitigation is to use additional…

机器学习 · 计算机科学 2024-10-10 Abhishek Panigrahi , Bingbin Liu , Sadhika Malladi , Andrej Risteski , Surbhi Goel

Knowledge distillation transfers knowledge from a high capacity teacher to a compact student using a mixture of hard and soft losses. On imbalanced data, a fixed weighting between hard and soft losses becomes brittle the learning process.…

机器学习 · 计算机科学 2026-05-20 Anh B. H. Nguyen , Ba Tho Phan , Viet Cuong Ta