mlsanityData to 14 Oct 2026

Synchronizing Task Behavior: Aligning Multiple Tasks during Test-Time Training

arXiv 2507.07778 ↗First posted 10 Jul 2025cs.LG cs.AI cs.CVcited by 5

Not on the map. Why?

Abstract (arXiv)

Generalizing neural networks to unseen target domains is a significant challenge in real-world deployments. Test-time training (TTT) addresses this by using an auxiliary self-supervised task to reduce the domain gap caused by distribution shifts between the source and target. However, we find that when models are required to perform multiple tasks under domain shifts, conventional TTT methods suffer from unsynchronized task behavior, where the adaptation steps needed for optimal performance in one task may not align with the requirements of other tasks. To address this, we propose a novel TTT approach called Synchronizing Tasks for Test-time Training (S4T), which enables the concurrent handling of multiple tasks. The core idea behind S4T is that predicting task relations across domain shifts is key to synchronizing tasks during test time. To validate our approach, we apply S4T to conventional multi-task benchmarks, integrating it with traditional TTT protocols. Our empirical results show that S4T outperforms state-of-the-art TTT methods across various benchmarks.

Key facts. Cited by 5 papers in the map (data through 14 Oct 2026). Builds on 45 of its 63 references that are in the map. First posted on arXiv 10 Jul 2025.

Builds on

45 of 63 references.

References · 45

  1. 2015Deep Residual Learning for Image Recognitioncited by 52,096
  2. 2020Denoising Diffusion Probabilistic Modelscited by 21,076
  3. 2021Masked Autoencoders Are Scalable Vision Learnerscited by 7,403
  4. 2015Variational Inference with Normalizing Flowscited by 2,170
  5. 2017Multi-Task Learning Using Uncertainty to Weigh Losses for Scene Geometry and Semanticscited by 1,358
  6. 2020Gradient Surgery for Multi-Task Learningcited by 1,256
  7. 2021Scaling Vision with Sparse Mixture of Expertscited by 747
  8. 2018Taskonomy: Disentangling Task Transfer Learningcited by 634
  9. 2022Continual Test-Time Domain Adaptationcited by 525
  10. 2021Conflict-Averse Gradient Descent for Multi-task Learningcited by 459
  11. 2023A Comprehensive Survey on Test-Time Adaptation under Distribution Shiftscited by 359
  12. 2022Multi-Task Learning as a Bargaining Gamecited by 215
33 more
  1. 2022Test-Time Training with Masked Autoencoderscited by 212
  2. 2022Multimodal Contrastive Learning with LIMoE: the Language-Image Mixture of Expertscited by 203
  3. 2020MTI-Net: Multi-Scale Task Interaction Networks for Multi-Task Learningcited by 157
  4. 2019Attentive Single-Tasking of Multiple Taskscited by 148
  5. 2019Pattern-Affinitive Propagation across Depth, Surface Normal and Semantic Segmentationcited by 142
  6. 2018Auxiliary Tasks in Multi-task Learningcited by 108
  7. 2020Robust Learning Through Cross-Task Consistencycited by 105
  8. 2022M3ViT: Mixture-of-Experts Vision Transformer for Efficient Multi-task Learning with Model-Accelerator Co-designcited by 98
  9. 2022InvPT: Inverted Pyramid Multi-task Transformer for Dense Scene Understandingcited by 89
  10. 2023FAMO: Fast Adaptive Multitask Optimizationcited by 78
  11. 2023Independent Component Alignment for Multi-Task Learningcited by 70
  12. 2022Mixture of Attention Heads: Selecting Attention Heads Per Tokencited by 60
  13. 2024Multi-Task Dense Prediction via Mixture of Low-Rank Expertscited by 58
  14. 2023Improved Test-Time Adaptation for Domain Generalizationcited by 47
  15. 2022Revisiting Realistic Test-Time Training: Sequential Inference and Adaptation by Anchored Clusteringcited by 44
  16. 2021Learning Multiple Dense Prediction Tasks from Partially Annotated Datacited by 42
  17. 2022ActMAD: Activation Matching to Align Distributions for Test-Time-Trainingcited by 41
  18. 2023On the Robustness of Open-World Test-Time Training: Self-Training with Dynamic Prototype Expansioncited by 34
  19. 2022TTTFlow: Unsupervised Test-Time Training with Normalizing Flowcited by 33
  20. 2021Learning to Relate Depth and Semantics for Unsupervised Domain Adaptationcited by 32
  21. 2024BECoTTA: Input-dependent Online Blending of Experts for Continual Test-time Adaptationcited by 25
  22. 2022On the Road to Online Adaptation for Semantic Image Segmentationcited by 23
  23. 2021Semi-supervised Multi-task Learning for Semantics and Depthcited by 22
  24. 2023ClusT3: Information Invariant Test-Time Trainingcited by 21
  25. 2020Generalizable Model-agnostic Semantic Segmentation via Target-specific Normalizationcited by 20
  26. 2020Taskology: Utilizing Task Relations at Scalecited by 19
  27. 2024DiffusionMTL: Learning Multi-Task Denoising Diffusion Model from Partially Annotated Datacited by 18
  28. 2025Selective Task Group Updates for Multi-Task Optimizationcited by 17
  29. 2019Multi-Task Semi-Supervised Adversarial Autoencoding for Speech Emotion Recognitioncited by 15
  30. 2020Partly Supervised Multitask Learningcited by 11
  31. 2024Quantifying Task Priority for Multi-Task Optimizationcited by 11
  32. 2019Semi-Supervised Multi-Task Learning With Chest X-Ray Imagescited by 8
  33. 2024Joint-Task Regularization for Partially Labeled Multi-Task Learningcited by 6

Built on by

Cited by year (2026 to 14 Oct)

Citing papers per year: 2026: 5.

5 papers.

  1. 2026MMEarth-Bench: Global Model Adaptation via Multimodal Test-Time Trainingcited by 7
  2. 2026Label-Free Cross-Task LoRA Merging with Null-Space Compressioncited by 4
  3. 2026Preference-Aligned LoRA Merging: Preserving Subspace Coverage and Addressing Directional Anisotropycited by 2
  4. 2026EVA-0: Test-Time Model Evolution with Only Two Forward Passes per Samplecited by 1
  5. 2026Triggering Generalist Reasoning via Predictive Uncertainty for Dual-System VLAcited by 0