mlsanityData to 7 Oct 2026

Distilling Morphology-Conditioned Hypernetworks for Efficient Universal Morphology Control

arXiv 2402.06570 ↗First posted 9 Feb 2024cs.LG cs.ROcited by 11

Not on the map. Why?

Abstract (arXiv)

Learning a universal policy across different robot morphologies can significantly improve learning efficiency and enable zero-shot generalization to unseen morphologies. However, learning a highly performant universal policy requires sophisticated architectures like transformers (TF) that have larger memory and computational cost than simpler multi-layer perceptrons (MLP). To achieve both good performance like TF and high efficiency like MLP at inference time, we propose HyperDistill, which consists of: (1) A morphology-conditioned hypernetwork (HN) that generates robot-wise MLP policies, and (2) A policy distillation approach that is essential for successful training. We show that on UNIMAL, a benchmark with hundreds of diverse morphologies, HyperDistill performs as well as a universal TF teacher policy on both training and unseen test robots, but reduces model size by 6-14 times, and computational cost by 67-160 times in different environments. Our analysis attributes the efficiency advantage of HyperDistill at inference time to knowledge decoupling, i.e., the ability to decouple inter-task and intra-task knowledge, a general principle that could also be applied to improve inference efficiency in other domains.

Key facts. Cited by 11 papers in the map (data through 7 Oct 2026). Builds on 46 of its 58 references that are in the map. First posted on arXiv 9 Feb 2024.

Builds on

46 of 58 references.

References · 46

  1. 2017Attention Is All You Needcited by 59,932
  2. 2023GPT-4 Technical Reportcited by 18,132
  3. 2015Distilling the Knowledge in a Neural Networkcited by 10,076
  4. 2023RT-2: Vision-Language-Action Models Transfer Web Knowledge to Robotic Controlcited by 2,993
  5. 2019A Comprehensive Survey on Graph Neural Networkscited by 2,906
  6. 2022RT-1: Robotics Transformer for Real-World Control at Scalecited by 2,007
  7. 2017Domain Randomization for Transferring Deep Neural Networks from Simulation to the Real Worldcited by 1,401
  8. 2016HyperNetworkscited by 1,003
  9. 2023Open X-Embodiment: Robotic Learning Datasets and RT-X Modelscited by 849
  10. 2015End-to-End Training of Deep Visuomotor Policiescited by 847
  11. 2021Scaling Vision with Sparse Mixture of Expertscited by 740
  12. 2018QT-Opt: Scalable Deep Reinforcement Learning for Vision-Based Robotic Manipulationcited by 730
34 more
  1. 2018Learning Dexterous In-Hand Manipulationcited by 643
  2. 2017Sim-to-Real Transfer of Robotic Control with Dynamics Randomizationcited by 616
  3. 2019Measuring and Relieving the Over-smoothing Problem for Graph Neural Networks from the Topological Viewcited by 544
  4. 2020Learning Quadrupedal Locomotion over Challenging Terraincited by 494
  5. 2015Policy Distillationcited by 414
  6. 2020Curriculum Learning for Reinforcement Learning Domains: A Framework and Surveycited by 350
  7. 2018Quantifying Generalization in Reinforcement Learningcited by 345
  8. 2018Learning to Adapt in Dynamic, Real-World Environments Through Meta-Reinforcement Learningcited by 270
  9. 2020Emergent Complexity and Zero-shot Transfer via Unsupervised Environment Designcited by 260
  10. 2015Contextual Markov Decision Processescited by 213
  11. 2015Actor-Mimic: Deep Multitask and Transfer Reinforcement Learningcited by 176
  12. 2021A System for General In-Hand Object Re-Orientationcited by 155
  13. 2021Embodied Intelligence via Learning and Evolutioncited by 154
  14. 2020One Policy to Control Them All: Shared Modular Policies for Agent-Agnostic Controlcited by 114
  15. 2023UniDexGrasp++: Improving Dexterous Grasping Policy Learning via Geometry-aware Curriculum and Iterative Generalist-Specialist Learningcited by 101
  16. 2023Scaling Vision-Language Models with Sparse Mixture of Expertscited by 88
  17. 2019Distilling Policy Distillationcited by 81
  18. 2022MetaMorph: Learning Universal Controllers with Transformerscited by 78
  19. 2023One-for-All: Bridge the Gap Between Heterogeneous Architectures in Knowledge Distillationcited by 77
  20. 2020Transient Non-Stationarity and Generalisation in Deep Reinforcement Learningcited by 75
  21. 2019Learning to Control Self-Assembling Morphologies: A Study of Generalization via Modularitycited by 68
  22. 2020My Body is a Cage: the Role of Morphology in Graph-Based Incompatible Controlcited by 68
  23. 2022GenLoco: Generalized Locomotion Controllers for Quadrupedal Robotscited by 36
  24. 2021Recomposing the Reinforcement Learning Building Blocks with Hypernetworkscited by 33
  25. 2022AnyMorph: Learning Transferable Polices By Inferring Agent Morphologycited by 25
  26. 2021Bayesian Meta-Learning for Few-Shot Policy Adaptation Across Robotic Platformscited by 22
  27. 2023Universal Morphology Control via Contextual Modulationcited by 21
  28. 2023SARA-RT: Scaling up Robotics Transformers with Self-Adaptive Robust Attentioncited by 20
  29. 2022Hypernetworks for Zero-shot Transfer in Reinforcement Learningcited by 18
  30. 2022Low-Rank Modular Reinforcement Learning via Muscle Synergycited by 16
  31. 2023Recurrent Hypernetworks are Surprisingly Strong in Meta-RLcited by 15
  32. 2022A System for Morphology-Task Generalization via Unified Representation and Behavior Distillationcited by 12
  33. 2023Subequivariant Graph Reinforcement Learning in 3D Environmentscited by 10
  34. 2021Linear Representation Meta-Reinforcement Learning for Instant Adaptationcited by 9

Built on by

Cited by year (2026 to 7 Oct)

Citing papers per year: 2025: 4; 2026: 7.

11 papers.

  1. 2025HyperVLA: Efficient Inference in Vision-Language-Action Models via Hypernetworkscited by 6
  2. 2025Shared Modular Recurrence in Contextual MDPs for Universal Morphology Controlcited by 3
  3. 2026Learning Diffusion Policy from Primitive Skills for Robot Manipulationcited by 3
  4. 2025Knowledge Diversion for Efficient Morphology Control and Policy Transfercited by 1
  5. 2025Efficient Morphology-Aware Policy Transfer to New Embodimentscited by 0
  6. 2026TeNet: Text-to-Network for Compact Policy Synthesiscited by 0
  7. 2026Constraint-based Pre-training: From Structured Constraints to Scalable Model Initializationcited by 0
  8. 2026The Embodiment Gap in Robot Foundation Modelscited by 0
  9. 2026Bridging Body and Brain: Gene-Driven Morphology--Control Co-Designcited by 0
  10. 2026Beam Search as Test-Time Self-Distillation via Counterfactual Contextscited by 0
  11. 2026The Weakest Link: Distilling LLM Reasoning with Worst-Case Constrained Reinforcement Learningcited by 0