• Home
  • Search
  • Revisiting Disentanglement in Downstream Tasks: A Study on Its Necessity for Abstract Visual Reasoning
  • Cite Icon2
  • https://doi.org/10.1609/aaai.v38i13.29354Copy DOI Icon

Revisiting Disentanglement in Downstream Tasks: A Study on Its Necessity for Abstract Visual Reasoning

  • Abstract
  • Literature Map
  • References
  • Citations
  • Similar Papers
Abstract

In representation learning, a disentangled representation is highly desirable as it encodes generative factors of data in a separable and compact pattern. Researchers have advocated leveraging disentangled representations to complete downstream tasks with encouraging empirical evidence. This paper further investigates the necessity of disentangled representation in downstream applications. Specifically, we show that dimension-wise disentangled representations are unnecessary on a fundamental downstream task, abstract visual reasoning. We provide extensive empirical evidence against the necessity of disentanglement, covering multiple datasets, representation learning methods, and downstream network architectures. Furthermore, our findings suggest that the informativeness of representations is a better indicator of downstream performance than disentanglement. Finally, the positive correlation between informativeness and disentanglement explains the claimed usefulness of disentangled representations in previous works. The source code is available at https://github.com/Richard-coder-Nai/disentanglement-lib-necessity.git

Similar Papers
  • Research Article
  • Citations29

Unsupervised Adversarially Robust Representation Learning on Graphs

  • Jun 28, 2022
  • Proceedings of the AAAI Conference on Artificial Intelligence
  • Jiarong Xu +6
  • Research Article
  • Citations13

Geographical Supervision Correction for Remote Sensing Representation Learning

  • Jan 01, 2022
  • IEEE Transactions on Geoscience and Remote Sensing
  • Wenyuan Li +2
  • Research Article
  • Citations11

Recent Advances in Representation Learning for Electronic Health Records: A Systematic Review

  • Feb 01, 2022
  • Journal of Physics: Conference Series
  • Xiaocong Liu +4
  • Research Article
  • Citations1

Representation Learning in PET Scans Enhanced by Semantic and 3D Position Specific Characteristics.

  • Sep 01, 2025
  • IEEE transactions on medical imaging
  • Theodoros P Vagenas +4
  • Conference Article
  • Citations93

MixGen: A New Multi-Modal Data Augmentation

  • Jan 01, 2023
  • Xiaoshuai Hao +6
  • Conference Article
  • Citations81

Emoji-Powered Representation Learning for Cross-Lingual Sentiment Classification

  • May 13, 2019
  • Zhenpeng Chen +5
  • Research Article

Relieving the Over-Smoothing Problem using Graph Disentanglement with Latent Bottlenecks

  • Oct 20, 2025
  • International Journal of Computational Intelligence Systems
  • Guoqiang Zhou +6
  • Research Article
  • Citations1

Self-Supervised Video Representation Learning via Capturing Semantic Changes Indicated by Saccades

  • Aug 01, 2024
  • IEEE Transactions on Circuits and Systems for Video Technology
  • Qiuxia Lai +5
  • Book Chapter
  • Citations8

Unsupervised Learning of Local Discriminative Representation for Medical Images

  • Jan 01, 2021
  • Huai Chen +7
  • Research Article
  • Citations6

STDNet: Rethinking Disentanglement Learning With Information Theory.

  • Aug 01, 2024
  • IEEE transactions on neural networks and learning systems
  • Ziwen Liu +4
  • Book Chapter
  • Citations10

Attention-Enhanced Disentangled Representation Learning for Unsupervised Domain Adaptation in Cardiac Segmentation

  • Jan 01, 2022
  • Xiaoyi Sun +5
  • Research Article
  • Citations10

Multimodal Sentiment Analysis With Mutual Information-Based Disentangled Representation Learning

  • Jul 01, 2025
  • IEEE Transactions on Affective Computing
  • Hao Sun +6
  • Research Article
  • Citations49

Structure-aware protein self-supervised learning

  • Apr 03, 2023
  • Bioinformatics
  • Can (Sam) Chen +4
  • Conference Article
  • Citations9

MAMO: Fine-Grained Vision-Language Representations Learning with Masked Multimodal Modeling

  • Jul 18, 2023
  • Zijia Zhao +5
  • PDF
  • Research Article
  • Citations12

Common Sense Knowledge Infusion for Visual Understanding and Reasoning: Approaches, Challenges, and Applications

  • Jul 01, 2022
  • IEEE Internet Computing
  • Muhammad Jaleed Khan +2
Cactus Communications logo

Copyright 2026 Cactus Communications. All rights reserved.