# Sparse Fusion for Multimodal Transformers

arXiv: https://arxiv.org/abs/2111.11992 · HTML: https://mlsanity.com/paper/2111.11992/

## Key facts

- Cited by 4 papers in the map (data through 14 Oct 2026).
- Builds on 28 of its 37 references that are in the map.
- First posted on arXiv 23 Nov 2021.

## Builds on (most cited first, 28 of 28)

- [Attention Is All You Need](https://mlsanity.com/paper/1706.03762/) (2017)
- [An Image is Worth 16x16 Words: Transformers for Image Recognition at Scale](https://mlsanity.com/paper/2010.11929/) (2020)
- [mixup: Beyond Empirical Risk Minimization](https://mlsanity.com/paper/1710.09412/) (2017)
- [Training data-efficient image transformers & distillation through attention](https://mlsanity.com/paper/2012.12877/) (2020)
- [Quo Vadis, Action Recognition? A New Model and the Kinetics Dataset](https://mlsanity.com/paper/1705.07750/) (2017)
- [BEiT: BERT Pre-Training of Image Transformers](https://mlsanity.com/paper/2106.08254/) (2021)
- [Big Bird: Transformers for Longer Sequences](https://mlsanity.com/paper/2007.14062/) (2020)
- [Reformer: The Efficient Transformer](https://mlsanity.com/paper/2001.04451/) (2020)
- [Generating Long Sequences with Sparse Transformers](https://mlsanity.com/paper/1904.10509/) (2019)
- [MixMatch: A Holistic Approach to Semi-Supervised Learning](https://mlsanity.com/paper/1905.02249/) (2019)
- [Perceiver: General Perception with Iterative Attention](https://mlsanity.com/paper/2103.03206/) (2021)
- [Multimodal Machine Learning: A Survey and Taxonomy](https://mlsanity.com/paper/1705.09406/) (2017)
- [Compressive Transformers for Long-Range Sequence Modelling](https://mlsanity.com/paper/1911.05507/) (2019)
- [VGGSound: A Large-scale Audio-Visual Dataset](https://mlsanity.com/paper/2004.14368/) (2020)
- [Multimodal Transformer for Unaligned Multimodal Language Sequences](https://mlsanity.com/paper/1906.00295/) (2019)
- [Convolutional Two-Stream Network Fusion for Video Action Recognition](https://mlsanity.com/paper/1604.06573/) (2016)
- [Attention Bottlenecks for Multimodal Fusion](https://mlsanity.com/paper/2107.00135/) (2021)
- [ZeRO-Offload: Democratizing Billion-Scale Model Training](https://mlsanity.com/paper/2101.06840/) (2021)
- [Memory Fusion Network for Multi-view Sequential Learning](https://mlsanity.com/paper/1802.00927/) (2018)
- [Not All Images are Worth 16x16 Words: Dynamic Transformers for Efficient Image Recognition](https://mlsanity.com/paper/2105.15075/) (2021)
- [TransformerFusion: Monocular RGB Scene Reconstruction using Transformers](https://mlsanity.com/paper/2107.02191/) (2021)
- [SiT: Self-supervised vIsion Transformer](https://mlsanity.com/paper/2104.03602/) (2021)
- [Multi-passage BERT: A Globally Normalized BERT Model for Open-domain Question Answering](https://mlsanity.com/paper/1908.08167/) (2019)
- [Accelerating Training of Transformer-Based Language Models with Progressive Layer Dropping](https://mlsanity.com/paper/2010.13369/) (2020)
- [Words Can Shift: Dynamically Adjusting Word Representations Using Nonverbal Behaviors](https://mlsanity.com/paper/1811.09362/) (2018)
- [Augmentation Strategies for Learning with Noisy Labels](https://mlsanity.com/paper/2103.02130/) (2021)
- [Scalable Vision Transformers with Hierarchical Pooling](https://mlsanity.com/paper/2103.10619/) (2021)
- [BP-Transformer: Modelling Long-Range Context via Binary Partitioning](https://mlsanity.com/paper/1911.04070/) (2019)

## Built on by (most cited first, 4 of 4)

- [Diffusion with Forward Models: Solving Stochastic Inverse Problems Without Direct Supervision](https://mlsanity.com/paper/2306.11719/) (2023)
- [Multimodal Alignment and Fusion: A Survey](https://mlsanity.com/paper/2411.17040/) (2024)
- [A Transformer-Based Model With Self-Distillation for Multimodal Emotion Recognition in Conversations](https://mlsanity.com/paper/2310.20494/) (2023)
- [Transformer-based Self-supervised Multimodal Representation Learning for Wearable Emotion Recognition](https://mlsanity.com/paper/2303.17611/) (2023)

## Abstract (arXiv)

Multimodal classification is a core task in human-centric machine learning. We observe that information is highly complementary across modalities, thus unimodal information can be drastically sparsified prior to multimodal fusion without loss of accuracy. To this end, we present Sparse Fusion Transformers (SFT), a novel multimodal fusion method for transformers that performs comparably to existing state-of-the-art methods while having greatly reduced memory footprint and computation cost. Key to our idea is a sparse-pooling block that reduces unimodal token sets prior to cross-modality modeling. Evaluations are conducted on multiple multimodal benchmark datasets for a wide range of classification tasks. State-of-the-art performance is obtained on multiple benchmarks under similar experiment conditions, while reporting up to six-fold reduction in computational cost and memory requirements. Extensive ablation studies showcase our benefits of combining sparsification and multimodal learning over naive approaches. This paves the way for enabling multimodal learning on low-resource devices.