Simple yet Effective Semi-supervised Knowledge Distillation from Vision-Language Models via Dual-Head Optimization

Fuente: arXiv
Salvato in:
Dettagli Bibliografici
Autori principali: Kang, Seongjae, Lee, Dong Bok, Jang, Hyungjoon, Hwang, Sung Ju
Natura: Preprint
Pubblicazione: 2025
Soggetti:
Accesso online:
Tags: Aggiungi Tag
Nessun Tag, puoi essere il primo ad aggiungerne!!
_version_ 1866911184708435968
author Kang, Seongjae
Lee, Dong Bok
Jang, Hyungjoon
Hwang, Sung Ju
author_facet Kang, Seongjae
Lee, Dong Bok
Jang, Hyungjoon
Hwang, Sung Ju
contents Semi-supervised learning (SSL) has emerged as a practical solution for addressing data scarcity challenges by leveraging unlabeled data. Recently, vision-language models (VLMs), pre-trained on massive image-text pairs, have demonstrated remarkable zero-/few-shot performance that often surpasses SSL approaches due to their exceptional generalization capabilities. This gap motivates us to question: how can we effectively harness the powerful generalization capabilities of VLMs into task-specific models? Knowledge distillation (KD) offers a natural framework for transferring VLM capabilities, but we identify that it suffers from gradient conflicts between supervised and distillation losses. To address this challenge, we propose Dual-Head Optimization (DHO), which introduces dual prediction heads for each distinct signal. We observe that DHO resolves gradient conflicts, enabling improved feature learning compared to single-head KD baselines, with practical benefits of minimal computational overhead and test-time hyperparameter tuning without retraining. Extensive experiments across 15 datasets show that DHO consistently outperforms KD baselines, often outperforming teacher models with smaller student models. DHO also achieves new state-of-the-art performance on both in-distribution ImageNet semi-supervised learning and out-of-distribution generalization across ImageNet variants. We publicly release our code and model checkpoints to facilitate future research at https://github.com/erjui/DHO.
format Preprint
id arxiv_https___arxiv_org_abs_2505_07675
institution arXiv
publishDate 2025
record_format arxiv
spellingShingle Simple yet Effective Semi-supervised Knowledge Distillation from Vision-Language Models via Dual-Head Optimization
Kang, Seongjae
Lee, Dong Bok
Jang, Hyungjoon
Hwang, Sung Ju
Machine Learning
Artificial Intelligence
Computer Vision and Pattern Recognition
Semi-supervised learning (SSL) has emerged as a practical solution for addressing data scarcity challenges by leveraging unlabeled data. Recently, vision-language models (VLMs), pre-trained on massive image-text pairs, have demonstrated remarkable zero-/few-shot performance that often surpasses SSL approaches due to their exceptional generalization capabilities. This gap motivates us to question: how can we effectively harness the powerful generalization capabilities of VLMs into task-specific models? Knowledge distillation (KD) offers a natural framework for transferring VLM capabilities, but we identify that it suffers from gradient conflicts between supervised and distillation losses. To address this challenge, we propose Dual-Head Optimization (DHO), which introduces dual prediction heads for each distinct signal. We observe that DHO resolves gradient conflicts, enabling improved feature learning compared to single-head KD baselines, with practical benefits of minimal computational overhead and test-time hyperparameter tuning without retraining. Extensive experiments across 15 datasets show that DHO consistently outperforms KD baselines, often outperforming teacher models with smaller student models. DHO also achieves new state-of-the-art performance on both in-distribution ImageNet semi-supervised learning and out-of-distribution generalization across ImageNet variants. We publicly release our code and model checkpoints to facilitate future research at https://github.com/erjui/DHO.
title Simple yet Effective Semi-supervised Knowledge Distillation from Vision-Language Models via Dual-Head Optimization
topic Machine Learning
Artificial Intelligence
Computer Vision and Pattern Recognition
url https://arxiv.org/abs/2505.07675