mlsanityData to 7 Oct 2026

Adapting Auxiliary Losses Using Gradient Similarity

arXiv 1812.02224 ↗First posted 5 Dec 2018stat.ML cs.LGcited by 98

Not on the map. Why?

Abstract (arXiv)

One approach to deal with the statistical inefficiency of neural networks is to rely on auxiliary losses that help to build useful representations. However, it is not always trivial to know if an auxiliary task will be helpful for the main task and when it could start hurting. We propose to use the cosine similarity between gradients of tasks as an adaptive weight to detect when an auxiliary loss is helpful to the main loss. We show that our approach is guaranteed to converge to critical points of the main task and demonstrate the practical usefulness of the proposed algorithm in a few domains: multi-task supervised learning on subsets of ImageNet, reinforcement learning on gridworld, and reinforcement learning on Atari games.

Key facts. Cited by 98 papers in the map (data through 7 Oct 2026). First posted on arXiv 5 Dec 2018.

Builds on

No reference list yet.

Built on by

Cited by year (2026 to 7 Oct)

Citing papers per year: 2020: 1; 2021: 15; 2022: 16; 2023: 20; 2024: 19; 2025: 10; 2026: 17.

98 papers.

Sort:

By area:

  1. 2021ImageNet-21K Pretraining for the Massescited by 607
  2. 2022Multi-Task Learning as a Bargaining Gamecited by 209
  3. 2021Fishr: Invariant Gradient Variances for Out-of-Distribution Generalizationcited by 186
  4. 2021Multi-Task Reinforcement Learning with Context-based Representationscited by 169
  5. 2021Continual World: A Robotic Benchmark For Continual Reinforcement Learningcited by 112
  6. 2022The CLRS Algorithmic Reasoning Benchmarkcited by 103
  7. 2021Evading the Simplicity Bias: Training a Diverse Set of Models Discovers Solutions with Superior OOD Generalizationcited by 83
  8. 2022Auto-Lambda: Disentangling Dynamic Task Relationshipscited by 73
  9. 2021On The Effect of Auxiliary Tasks on Representation Dynamicscited by 56
  10. 2024ConFIG: Towards Conflict-free Training of Physics Informed Neural Networkscited by 43
  11. 2024Make Continual Learning Stronger via C-Flatcited by 35
  12. 2022MetaBalance: Improving Multi-Task Recommendations via Adapting Gradient Magnitudes of Auxiliary Taskscited by 34
38 more
  1. 2021Gradient Imitation Reinforcement Learning for Low Resource Relation Extractioncited by 33
  2. 2021Adaptive Transfer Learning on Graph Neural Networkscited by 31
  3. 2023ForkMerge: Mitigating Negative Transfer in Auxiliary-Task Learningcited by 28
  4. 2023Fast-Slow Test-Time Adaptation for Online Vision-and-Language Navigationcited by 25
  5. 2021Avoiding Catastrophe: Active Dendrites Enable Multi-Task Learning in Dynamic Environmentscited by 24
  6. 2022It's All In the Teacher: Zero-Shot Quantization Brought Closer to the Teachercited by 20
  7. 2023Scalarization for Multi-Task and Multi-Domain Learning at Scalecited by 19
  8. 2022Deep AutoAugmentcited by 18
  9. 2023Exploring the Relationship between Alignment and Cross-lingual Transfer in Multilingual Transformerscited by 18
  10. 2021Auxiliary Task Update Decomposition: The Good, The Bad and The Neutralcited by 17
  11. 2025A Survey of State Representation Learning for Deep Reinforcement Learningcited by 17
  12. 2020Lipschitzness Is All You Need To Tame Off-policy Generative Adversarial Imitation Learningcited by 15
  13. 2024Mixture-of-Skills: Learning to Optimize Data Usage for Fine-Tuning Large Language Modelscited by 15
  14. 2022Multi-Agent Policy Transfer via Task Relationship Modelingcited by 14
  15. 2024AMAGO-2: Breaking the Multi-Task Barrier in Meta-Reinforcement Learning with Transformerscited by 14
  16. 2023Improving Few-Shot Generalization by Exploring and Exploiting Auxiliary Datacited by 13
  17. 2023Auxiliary Learning as an Asymmetric Bargaining Gamecited by 12
  18. 2022Fix-A-Step: Semi-supervised Learning from Uncurated Unlabeled Datacited by 11
  19. 2024A Survey on Self-Supervised Graph Foundation Models: Knowledge-Based Perspectivecited by 9
  20. 2025Forget Forgetting: Continual Learning in a World of Abundant Memorycited by 9
  21. 2022Recent Advances in Embedding Methods for Multi-Object Tracking: A Surveycited by 8
  22. 2025OASIS: Online Sample Selection for Continual Visual Instruction Tuningcited by 8
  23. 2023Gradient Shaping for Multi-Constraint Safe Reinforcement Learningcited by 7
  24. 2024A Layer Selection Approach to Test Time Adaptationcited by 7
  25. 2024Budgeted Online Continual Learning by Adaptive Layer Freezing and Frequency-based Samplingcited by 7
  26. 2023GDOD: Effective Gradient Descent using Orthogonal Decomposition for Multi-Task Learningcited by 6
  27. 2023Sample-Level Weighting for Multi-Task Learning with Auxiliary Taskscited by 6
  28. 2023When Multi-Task Learning Meets Partial Supervision: A Computer Vision Reviewcited by 6
  29. 2023Measuring and Mitigating Interference in Reinforcement Learningcited by 5
  30. 2025CSTA: Spatial-Temporal Causal Adaptive Learning for Exemplar-Free Video Class-Incremental Learningcited by 5
  31. 2023Overcoming Distribution Mismatch in Quantizing Image Super-Resolution Networkscited by 4
  32. 2024Careful with that Scalpel: Improving Gradient Surgery with an EMAcited by 4
  33. 2024Using dynamic loss weighting to boost improvements in forecast stabilitycited by 4
  34. 2025Understanding Knowledge Transferability for Transfer Learning: A Surveycited by 4
  35. 2021Learning Multi-Objective Curricula for Robotic Policy Learningcited by 3
  36. 2022Adaptive Meta-learner via Gradient Similarity for Few-shot Text Classificationcited by 3
  37. 2023Investigating Navigation Strategies in the Morris Water Maze through Deep Reinforcement Learningcited by 3
  38. 2021Multiple Instance Learning with Auxiliary Task Weighting for Multiple Myeloma Classificationcited by 2

50 most cited of 98 All 98 (JSON)