# Cross-Modal Knowledge Distillation without Paired Data: Theoretical Foundation and Algorithm

arXiv: https://arxiv.org/abs/2606.10504 · HTML: https://mlsanity.com/paper/2606.10504/

## Key facts

- No citing papers in the map yet (data through 14 Oct 2026).
- Placed in Multimodal distillation, one of 12 directions in Knowledge distillation.
- Builds on 26 of its 41 references that are in the map.
- First posted on arXiv 9 Jun 2026.

## Placed in

- Knowledge distillation / [Multimodal distillation](https://mlsanity.com/distill/multimodal-distillation/)

## Builds on (most cited first, 26 of 26)

- [Deep Residual Learning for Image Recognition](https://mlsanity.com/paper/1512.03385/) (2015)
- [Distilling the Knowledge in a Neural Network](https://mlsanity.com/paper/1503.02531/) (2015)
- [Model-Agnostic Meta-Learning for Fast Adaptation of Deep Networks](https://mlsanity.com/paper/1703.03400/) (2017)
- [FitNets: Hints for Thin Deep Nets](https://mlsanity.com/paper/1412.6550/) (2014)
- [Relational Knowledge Distillation](https://mlsanity.com/paper/1904.05068/) (2019)
- [Computational Optimal Transport](https://mlsanity.com/paper/1803.00567/) (2018)
- [Contrastive Representation Distillation](https://mlsanity.com/paper/1910.10699/) (2019)
- [VGGSound: A Large-scale Audio-Visual Dataset](https://mlsanity.com/paper/2004.14368/) (2020)
- [Distilling Knowledge via Knowledge Review](https://mlsanity.com/paper/2104.09044/) (2021)
- [Balanced Multimodal Learning via On-the-fly Gradient Modulation](https://mlsanity.com/paper/2203.15332/) (2022)
- [DeepJDOT: Deep Joint Distribution Optimal Transport for Unsupervised Domain Adaptation](https://mlsanity.com/paper/1803.10081/) (2018)
- [Knowledge Distillation from A Stronger Teacher](https://mlsanity.com/paper/2205.10536/) (2022)
- [LEEP: A New Measure to Evaluate Transferability of Learned Representations](https://mlsanity.com/paper/2002.12462/) (2020)
- [Cross Modal Distillation for Supervision Transfer](https://mlsanity.com/paper/1507.00448/) (2015)
- [The Modality Focusing Hypothesis: Towards Understanding Crossmodal Knowledge Distillation](https://mlsanity.com/paper/2206.06487/) (2022)
- [NORM: Knowledge Distillation via N-to-One Representation Matching](https://mlsanity.com/paper/2305.13803/) (2023)
- [Leveraging Vision-Language Models for Improving Domain Generalization in Image Classification](https://mlsanity.com/paper/2310.08255/) (2023)
- [XKD: Cross-modal Knowledge Distillation with Domain Alignment for Video Representation Learning](https://mlsanity.com/paper/2211.13929/) (2022)
- [Multimodal Knowledge Expansion](https://mlsanity.com/paper/2103.14431/) (2021)
- [Wasserstein Distance Rivals Kullback-Leibler Divergence for Knowledge Distillation](https://mlsanity.com/paper/2412.08139/) (2024)
- [COSMOS: Cross-Modality Self-Distillation for Vision Language Pre-training](https://mlsanity.com/paper/2412.01814/) (2024)
- [Multi-Label Knowledge Distillation](https://mlsanity.com/paper/2308.06453/) (2023)
- [Cross-Modality Distillation: A case for Conditional Generative Adversarial Networks](https://mlsanity.com/paper/1807.07682/) (2018)
- [Knowledge Distillation with Refined Logits](https://mlsanity.com/paper/2408.07703/) (2024)
- [A Cross Modal Knowledge Distillation & Data Augmentation Recipe for Improving Transcriptomics Representations through Morphological Features](https://mlsanity.com/paper/2505.21317/) (2025)
- [Rethinking Cross-Modal Fine-Tuning: Optimizing the Interaction Between Feature Alignment and Target Fitting](https://mlsanity.com/paper/2601.18231/) (2026)

## Abstract (arXiv)

Cross-modal knowledge distillation (CMKD) studies how a (large) teacher model trained on one type of data (e.g., images) can guide a (smaller) student model building on another type of data (e.g., text/audio). Existing CMKD methods often require paired multi-modal data with aligned semantics, but obtaining such paired data are often costly and impractical. To mitigate this limitation, we develop a new CMKD framework for the more challenging setting where paired data are unavailable. In particular, we establish a cross-modal distributional relationship between teacher and student models, which reveals two fundamental quantities governing effective distillation: feature alignment and label alignment. These quantities characterize semantic discrepancy between modalities at the levels of representation and prediction distributions, respectively. Motivated by this insight, we propose a principled framework, with theoretical guarantees, that enables effective cross-modal knowledge distillation by aligning distributions rather than individual samples. Extensive experiments across a wide range of multimodal benchmarks show that our framework is highly effective in both unpaired and paired data settings, improving significantly over prior work.