When Better Teachers Don't Make Better Students: Revisiting Knowledge Distillation for CLIP Models in VQA

• URL: http://arxiv.org/abs/2511.17886v1
• Date: Sat, 22 Nov 2025 02:30:18 GMT
• Title: When Better Teachers Don't Make Better Students: Revisiting Knowledge Distillation for CLIP Models in VQA
• Authors: Pume Tuchinda, Parinthapat Pengpun, Romrawin Chumpu, Sarana Nutanong, Peerat Limkonchotiwat,
• Abstract summary: Vision-language models (VLMs) have achieved remarkable success across multimodal tasks, yet their substantial computational demands hinder efficient deployment.<n> Knowledge distillation (KD) has emerged as a powerful approach for building lightweight but competitive models.<n>We present the first systematic study of distillation across a range of CLIP-style teacher models, ranging from standard baselines to large-scale state-of-the-art models.
• Score: 13.795585671688213
• License: http://arxiv.org/licenses/nonexclusive-distrib/1.0/
• Abstract: Vision-language models (VLMs) have achieved remarkable success across multimodal tasks, yet their substantial computational demands hinder efficient deployment. Knowledge distillation (KD) has emerged as a powerful approach for building lightweight but competitive models, with strong evidence from both language and vision domains. However, its application to VLMs, particularly CLIP-style models, remains limited, often constrained to small-scale teachers and narrow evaluation tasks such as classification or retrieval. In this work, we present the first systematic study of distillation across a range of CLIP-style teacher models, ranging from standard baselines to large-scale state-of-the-art models. Contrary to trends observed in NLP and vision, we find that stronger teachers do not consistently yield better students; in fact, existing distillation frameworks often fail to scale, leading to degraded performance in downstream multimodal tasks such as visual question answering. Our findings challenge prevailing assumptions in KD and point toward new directions for designing parameter-efficient multimodal models.

Related papers

• CustomKD: Customizing Large Vision Foundation for Edge Model Improvement via Knowledge Distillation [57.91828170220308]
  We propose a knowledge distillation approach, CustomKD, that effectively leverages large vision foundation models (LVFMs) to enhance the performance of edge models.<n>Our simple yet effective CustomKD customizes the well-generalized features inherent in LVFMs to a given student model in order to reduce model discrepancies.
  arXiv  Detail & Related papers  (2025-03-23T23:53:08Z)
• Boosting the Generalization and Reasoning of Vision Language Models with Curriculum Reinforcement Learning [12.728451197053321]
  We propose Curriculum Reinforcement Finetuning (Curr-ReFT), a novel post-training paradigm specifically designed for small-scale vision-language models (VLMs)<n>Curr-ReFT comprises two sequential stages: Curriculum Reinforcement Learning and Rejected Sampling-based Self-improvement.<n>Our experiments demonstrate that models trained with Curr-ReFT paradigm achieve state-of-the-art performance across various visual tasks.
  arXiv  Detail & Related papers  (2025-03-10T08:48:50Z)
• RADIOv2.5: Improved Baselines for Agglomerative Vision Foundation Models [60.596005921295806]
  Agglomerative models have emerged as a powerful approach to training vision foundation models.<n>We identify critical challenges including resolution mode shifts, teacher imbalance, idiosyncratic teacher artifacts, and an excessive number of output tokens.<n>We propose several novel solutions: multi-resolution training, mosaic augmentation, and improved balancing of teacher loss functions.
  arXiv  Detail & Related papers  (2024-12-10T17:06:41Z)
• Active Data Curation Effectively Distills Large-Scale Multimodal Models [66.23057263509027]
  Knowledge distillation (KD) is the de facto standard for compressing large-scale models into smaller ones.<n>In this work we explore an alternative, yet simple approach -- active data curation as effective distillation for contrastive multimodal pretraining.<n>Our simple online batch selection method, ACID, outperforms strong KD baselines across various model-, data- and compute-configurations.
  arXiv  Detail & Related papers  (2024-11-27T18:50:15Z)
• Linear Projections of Teacher Embeddings for Few-Class Distillation [14.99228980898161]
  Knowledge Distillation (KD) has emerged as a promising approach for transferring knowledge from a larger, more complex teacher model to a smaller student model. We introduce a novel method for distilling knowledge from the teacher's model representations, which we term Learning Embedding Linear Projections (LELP) Our experimental evaluation on large-scale NLP benchmarks like Amazon Reviews and Sentiment140 demonstrate the LELP is consistently competitive with, and typically superior to, existing state-of-the-art distillation algorithms for binary and few-class problems.
  arXiv  Detail & Related papers  (2024-09-30T16:07:34Z)
• Exploring and Enhancing the Transfer of Distribution in Knowledge Distillation for Autoregressive Language Models [62.5501109475725]
  Knowledge distillation (KD) is a technique that compresses large teacher models by training smaller student models to mimic them. This paper introduces Online Knowledge Distillation (OKD), where the teacher network integrates small online modules to concurrently train with the student model. OKD achieves or exceeds the performance of leading methods in various model architectures and sizes, reducing training time by up to fourfold.
  arXiv  Detail & Related papers  (2024-09-19T07:05:26Z)
• LLAVADI: What Matters For Multimodal Large Language Models Distillation [77.73964744238519]
  In this work, we do not propose a new efficient model structure or train small-scale MLLMs from scratch. Our studies involve training strategies, model choices, and distillation algorithms in the knowledge distillation process. By evaluating different benchmarks and proper strategy, even a 2.7B small-scale model can perform on par with larger models with 7B or 13B parameters.
  arXiv  Detail & Related papers  (2024-07-28T06:10:47Z)
• Multimodal CLIP Inference for Meta-Few-Shot Image Classification [0.0]
  Multimodal foundation models like CLIP learn a joint (image, text) embedding. This study demonstrates that combining modalities from CLIP's text and image encoders outperforms state-of-the-art meta-few-shot learners on widely adopted benchmarks.
  arXiv  Detail & Related papers  (2024-03-26T17:47:54Z)
• Continual Learners are Incremental Model Generalizers [70.34479702177988]
  This paper extensively studies the impact of Continual Learning (CL) models as pre-trainers. We find that the transfer quality of the representation often increases gradually without noticeable degradation in fine-tuning performance. We propose a new fine-tuning scheme, GLobal Attention Discretization (GLAD), that preserves rich task-generic representation during solving downstream tasks.
  arXiv  Detail & Related papers  (2023-06-21T05:26:28Z)

This list is automatically generated from the titles and abstracts of the papers in this site.

This site does not guarantee the quality of this site (including all information) and is not responsible for any consequences.