mlsanityData to 7 Oct 2026

Tunable Soft Equivariance with Guarantees

arXiv 2603.26657 ↗First posted 27 Mar 2026cs.CV cs.LGcited by 0

Not on the map. Why?

Abstract (arXiv)

Equivariance is a fundamental property in computer vision models, yet strict equivariance is rarely satisfied in real-world data, which can limit a model's performance. Controlling the degree of equivariance is therefore desirable. We propose a general framework for constructing soft equivariant models by projecting the model weights into a designed subspace. The method applies to any pre-trained architecture and provides theoretical bounds on the induced equivariance error. Empirically, we demonstrate the effectiveness of our method on multiple pre-trained backbones, including ViT and ResNet, across image classification, semantic segmentation, and human-trajectory prediction tasks. Notably, our approach improves the performance while simultaneously reducing equivariance error on the competitive ImageNet benchmark.

Key facts. No citing papers in the map yet (data through 7 Oct 2026). Builds on 59 of its 76 references that are in the map. First posted on arXiv 27 Mar 2026.

Builds on

59 of 76 references.

References · 59

  1. 2015Deep Residual Learning for Image Recognitioncited by 51,809
  2. 2020An Image is Worth 16x16 Words: Transformers for Image Recognition at Scalecited by 29,915
  3. 2014Very Deep Convolutional Networks for Large-Scale Image Recognitioncited by 15,267
  4. 2016Semi-Supervised Classification with Graph Convolutional Networkscited by 11,282
  5. 2023DINOv2: Learning Robust Visual Features without Supervisioncited by 7,764
  6. 2021Masked Autoencoders Are Scalable Vision Learnerscited by 7,339
  7. 2016PointNet: Deep Learning on Point Sets for 3D Classification and Segmentationcited by 5,223
  8. 2014Going Deeper with Convolutionscited by 4,231
  9. 2021SegFormer: Simple and Efficient Design for Semantic Segmentation with Transformerscited by 2,923
  10. 2017Deep Setscited by 1,394
  11. 2016Group Equivariant Convolutional Networkscited by 1,154
  12. 2016Semantic Understanding of Scenes through the ADE20K Datasetcited by 1,063
38 more
  1. 2018Tensor field networks: Rotation- and translation-equivariant neural networks for 3D point cloudscited by 735
  2. 2019Making Convolutional Networks Shift-Invariant Againcited by 390
  3. 2018Invariant and Equivariant Graph Networkscited by 339
  4. 2021Vector Neurons: A General Framework for SO(3)-Equivariant Networkscited by 304
  5. 2016Steerable CNNscited by 298
  6. 2022Equiformer: Equivariant Graph Attention Transformer for 3D Atomistic Graphscited by 258
  7. 2022Structure-based Drug Design with Equivariant Diffusion Modelscited by 213
  8. 2021A Practical Method for Constructing Equivariant Multilayer Perceptrons for Arbitrary Matrix Groupscited by 185
  9. 2020Transformer Networks for Trajectory Forecastingcited by 180
  10. 2020MDP Homomorphic Networks: Group Symmetries in Reinforcement Learningcited by 133
  11. 2019Deep Scale-spaces: Equivariance Over Scalecited by 113
  12. 2019Scale-Equivariant Steerable Networkscited by 110
  13. 2022Equivariance with Learned Canonicalization Functionscited by 106
  14. 2022Approximately Equivariant Networks for Imperfectly Symmetric Dynamicscited by 95
  15. 2018Roto-Translation Covariant Convolutional Networks for Medical Image Analysiscited by 91
  16. 2021Residual Pathway Priors for Soft Equivariance Constraintscited by 74
  17. 2018Deep Models of Interactions Across Setscited by 68
  18. 2024Pretraining Codomain Attention Neural Operators for Solving Multiphysics PDEscited by 58
  19. 2020Truly shift-invariant convolutional neural networkscited by 53
  20. 2019PIC: Permutation Invariant Critic for Multi-Agent Deep Reinforcement Learningcited by 46
  21. 2023A new perspective on building efficient and expressive 3D equivariant graph neural networkscited by 46
  22. 2022Relaxing Equivariance Constraints with Non-stationary Continuous Filterscited by 41
  23. 2023Equivariant Adaptation of Large Pretrained Modelscited by 40
  24. 2022SpeqNets: Sparsity-aware Permutation-equivariant Graph Networkscited by 37
  25. 2015Discrete Signal Processing on Graphs: Sampling Theorycited by 33
  26. 2019Chirality Nets for Human Pose Regressioncited by 31
  27. 2023Learning Layer-wise Equivariances Automatically using Gradientscited by 23
  28. 2022Equi-Tuning: Group Equivariant Fine-Tuning of Pretrained Modelscited by 22
  29. 2022Equivariance Discovery by Learned Parameter-Sharingcited by 18
  30. 2022Learnable Polyphase Sampling for Shift Invariant and Equivariant Convolutional Networkscited by 18
  31. 2023A General Theory of Correct, Incorrect, and Extrinsic Equivariancecited by 18
  32. 2023Making Vision Transformers Truly Shift-Equivariantcited by 16
  33. 2023Regularizing Towards Soft Equivariance Under Mixed Symmetriescited by 16
  34. 2023Truly Scale-Equivariant Deep Nets with Fourier Layerscited by 16
  35. 2025Alias-Free Latent Diffusion Models: Improving Fractional Shift Equivariance of Diffusion Latent Spacecited by 14
  36. 2024Relaxed Equivariance via Multitask Learningcited by 13
  37. 2023Almost Equivariance via Lie Algebra Convolutionscited by 10
  38. 2025Learning (Approximately) Equivariant Networks via Constrained Optimizationcited by 10

50 most cited of 59 All 59 (JSON)

Built on by

Not cited yet.