paper-with-me

홈 › Papers

Switch-KD: Visual-Switch Knowledge Distillation for Vision-Language Models

2026-04-16 · Haoyi Sun, Xiaoxiao Wang, Ning Mao, Qian Wang, Lifu Mu, Wen Zheng, Tao Wei, Wei Chen arxiv

Vision-Language Models (VLMs) have shown remarkable capabilities in joint vision-language understanding, but their large scale poses significant challenges for deployment in resource-constrained scenarios. Knowledge Distillation (KD) offers a viable way to improve model capabilities without increasing model size or data requirements, making deployment more efficient. However, applying KD to VLMs is challenged by modality-specific supervision: although multimodal knowledge in VLMs is fused within the language space, current methods supervise each modality separately without explicitly addressing multimodal alignment, leading to inconsistent multimodal knowledge transfer. To address this, we propose Switch-KD, a visual-switch distillation framework that unifies vision-language knowledge transfer within a shared text-probability space. Switch-KD comprises two key components: (1) Visual-Switch Distillation, which switches the student's visual outputs into the teacher's language pathway to construct cross-modal probabilistic references for implicit visual knowledge transfer; and (2) Dynamic Bi-directional Logits Difference (DBiLD) loss, which adaptively aligns informative probability regions while preserving the distributional structures of teacher and student through bidirectional supervision. Guided by Switch-KD, a 0.5B TinyLLaVA effectively distills rich multimodal knowledge from its 3B teacher, yielding an average improvement of 3.6 points across 10 multimodal benchmarks without any architectural modification.

📄 PDF Abstract BibTeX arXiv:2604.14629

Code (0)

등록된 구현이 없습니다.

Tasks

Knowledge Distillation

Similar Papers 제목 키워드 기반

AdaSwitch: Balancing Exploration and Guidance in Knowledge Distillation via Adaptive Switching

2025-10-09 · Jingyu Peng, Maolin Wang, Hengyi Cai, Yuchen Li 외 arxiv

Small language models (SLMs) are crucial for applications with strict latency and computational constraints, yet achieving high performance remains challenging. Knowledge distillation (KD) can transfer capabilities from …

Knowledge Distillation

Switchable Online Knowledge Distillation

2022-09-12 · Biao Qian, Yang Wang, Hongzhi Yin, Richang Hong 외

Online Knowledge Distillation (OKD) improves the involved models by reciprocally exploiting the difference between teacher and student. Several crucial bottlenecks over the gap between them -- e.g., Why and when does a l…

Knowledge Distillation

Knowledge Distillation During Mid-Training Favors Reasoning over Factual Recall

2026-09-01 · Jacqueline He, Howard Yen, Shuyue Stella Li, Margaret Li 외 hf

Logit-based knowledge distillation (KD) is used to train smaller language models (LMs) via supervision from stronger teachers, but whether its benefits are consistent across training stages remains unclear. Through contr…

Self-Supervised LearningKnowledge Distillation

Not All Prompts Are Secure: A Switchable Backdoor Attack Against Pre-trained Vision Transfomers

2024-01-01 · CVPR 2024 1 · Sheng Yang, Jiawang Bai, Kuofeng Gao, Yong Yang 외

Given the power of vision transformers a new learning paradigm pre-training and then prompting makes it more efficient and effective to address downstream visual recognition tasks. In this paper we identify a novel s…

AllBackdoor Attack

Not All Prompts Are Secure: A Switchable Backdoor Attack Against Pre-trained Vision Transformers

2024-05-17 · Sheng Yang, Jiawang Bai, Kuofeng Gao, Yong Yang 외

Given the power of vision transformers, a new learning paradigm, pre-training and then prompting, makes it more efficient and effective to address downstream visual recognition tasks. In this paper, we identify a novel s…

AllBackdoor Attack