mlsanityData to 14 Oct 2026

A Framework for Controllable Multi-objective Learning with Annealed Stein Variational Hypernetworks

arXiv 2506.06715 ↗First posted 7 Jun 2025cs.LG stat.MLcited by 1

Not on the map. Why?

Abstract (arXiv)

Pareto Set Learning (PSL) is popular as an efficient approach to obtaining the complete optimal solution in Multi-objective Learning (MOL). A set of optimal solutions approximates the Pareto set, and its mapping is a set of dense points in the Pareto front in objective space. However, some current methods face a challenge: how to make the Pareto solution is diverse while maximizing the hypervolume value. In this paper, we propose a novel method to address this challenge, which employs Stein Variational Gradient Descent (SVGD) to approximate the entire Pareto set. SVGD pushes a set of particles towards the Pareto set by applying a form of functional gradient descent, which helps to converge and diversify optimal solutions. Additionally, we employ diverse gradient direction strategies to thoroughly investigate a unified framework for SVGD in multi-objective optimization and adapt this framework with an annealing schedule to promote stability. We introduce our method, SVH-MOL, and validate its effectiveness through extensive experiments on multi-objective problems and multi-task learning, demonstrating its superior performance.

Key facts. Cited by 1 paper in the map (data through 14 Oct 2026). Builds on 29 of its 51 references that are in the map. First posted on arXiv 7 Jun 2025.

Builds on

29 of 51 references.

References · 29

  1. 2017Multi-Task Learning Using Uncertainty to Weigh Losses for Scene Geometry and Semanticscited by 1,358
  2. 2020Gradient Surgery for Multi-Task Learningcited by 1,256
  3. 2016HyperNetworkscited by 1,027
  4. 2017GradNorm: Gradient Normalization for Adaptive Loss Balancing in Deep Multitask Networkscited by 834
  5. 2018Multi-Task Learning as Multi-Objective Optimizationcited by 821
  6. 2018End-to-End Multi-Task Learning with Attentioncited by 534
  7. 2016Stein Variational Gradient Descent: A General Purpose Bayesian Inference Algorithmcited by 507
  8. 2023Fine-Grained Human Feedback Gives Better Rewards for Language Model Trainingcited by 401
  9. 2020Multi-Task Learning with Deep Neural Networks: A Surveycited by 375
  10. 2019Pareto Multi-Task Learningcited by 224
  11. 2020Learning the Pareto Front with Hypernetworkscited by 125
  12. 2023Beyond One-Preference-Fits-All Alignment: Multi-Objective Direct Preference Optimizationcited by 124
17 more
  1. 2019On the geometry of Stein variational gradient descentcited by 82
  2. 2020An Easy-to-use Real-world Multi-objective Optimization Problem Suitecited by 71
  3. 2024Smooth Tchebycheff Scalarization for Multi-Objective Optimizationcited by 54
  4. 2022Pareto Set Learning for Expensive Multi-Objective Optimizationcited by 51
  5. 2017Message Passing Stein Variational Gradient Descentcited by 47
  6. 2023MFTCoder: Boosting Code LLMs with Multitask Fine-Tuningcited by 35
  7. 2021Annealed Stein Variational Gradient Descentcited by 26
  8. 2019Multi-Gradient Descent for Multi-Objective Recommender Systemscited by 20
  9. 2025Pareto Set Learning for Multi-Objective Reinforcement Learningcited by 20
  10. 2025Gradient-Based Multi-Objective Deep Learning: Algorithms, Theories, Applications, and Beyondcited by 18
  11. 2021Multi-Objective Learning to Predict Pareto Fronts Using Hypervolume Maximizationcited by 16
  12. 2022Improving Pareto Front Learning via Multi-Sample Hypernetworkscited by 16
  13. 2023A Framework for Controllable Pareto Front Learning with Completed Scalarization Functions and its Applicationscited by 13
  14. 2021Stein Variational Gradient Descent with Multiple Kernelcited by 4
  15. 2024Improving Pareto Set Learning for Expensive Multi-objective Optimization via Stein Variational Hypernetworkscited by 4
  16. 2024Preference-Optimized Pareto Set Learning for Blackbox Optimizationcited by 2
  17. 2024High-Dimensional Bayesian Optimization via Random Projection of Manifold Subspacescited by 2

Built on by

Cited by year (2026 to 14 Oct)

Citing papers per year: 2026: 1.

1 paper.

  1. 2026Universal Hypernetworks for Arbitrary Modelscited by 0