Abstract:Knowledge distillation (KD) is a machine learning framework that transfers knowledge from a teacher model to a student model. The vanilla KD proposed by Hinton et al. has been the dominant approach in logit-based distillation and demonstrates compelling performance. However, it only performs sample-wise probability alignment between teacher and student's predictions, lacking an mechanism for class-wise comparison. Besides, vanilla KD imposes no structural constraint on the probability space. In this work, we propose a simple yet effective methodology, bilateral contrastive knowledge distillation (BicKD). This approach introduces a novel bilateral contrastive loss, which intensifies the orthogonality among different class generalization spaces while preserving consistency within the same class. The bilateral formulation enables explicit comparison of both sample-wise and class-wise prediction patterns between teacher and student. By emphasizing probabilistic orthogonality, BicKD further regularizes the geometric structure of the predictive distribution. Extensive experiments show that our BicKD method enhances knowledge transfer, and consistently outperforms state-of-the-art knowledge distillation techniques across various model architectures and benchmarks.
| Comments: | Accepted to the 2026 IEEE/INNS International Joint Conference on Neural Networks (IJCNN 2026) |
| Subjects: | Machine Learning (cs.LG) |
| Cite as: | arXiv:2602.01265 [cs.LG] |
| (or arXiv:2602.01265v2 [cs.LG] for this version) | |
| https://doi.org/10.48550/arXiv.2602.01265 arXiv-issued DOI via DataCite |
From: Jiangnan Zhu [view email]
[v1]
Sun, 1 Feb 2026 14:54:34 UTC (2,659 KB)
[v2]
Thu, 30 Apr 2026 13:24:30 UTC (1,372 KB)
此内容由惯性聚合(RSS阅读器)自动聚合整理，仅供阅读参考。 原文来自 — 版权归原作者所有。