mlsanityData to 7 Oct 2026

LocoFormer: Generalist Locomotion via Long-context Adaptation

arXiv 2509.23745 ↗First posted 28 Sep 2025cs.RO cs.AIcited by 14

Not on the map. Why?

Abstract (arXiv)

Modern locomotion controllers are manually tuned for specific embodiments. We present LocoFormer, a generalist omni-bodied locomotion model that can control previously unseen legged and wheeled robots, even without precise knowledge of their kinematics. LocoFormer is able to adapt to changes in morphology and dynamics at test time. We find that two key choices enable adaptation. First, we train massive scale RL on procedurally generated robots with aggressive domain randomization. Second, in contrast to previous policies that are myopic with short context lengths, we extend context by orders of magnitude to span episode boundaries. We deploy the same LocoFormer to varied robots and show robust control even with large disturbances such as weight change and motor failures. In extreme scenarios, we see emergent adaptation across episodes, LocoFormer learns from falls in early episodes to improve control strategies in later ones. We believe that this simple, yet general recipe can be used to train foundation models for other robotic skills in the future. Videos at generalist-locomotion.github.io.

Key facts. Cited by 14 papers in the map (data through 7 Oct 2026). Builds on 39 of its 44 references that are in the map. First posted on arXiv 28 Sep 2025.

Builds on

39 of 44 references.

References · 39

  1. 2017Attention Is All You Needcited by 59,932
  2. 2020Language Models are Few-Shot Learnerscited by 32,199
  3. 2017Proximal Policy Optimization Algorithmscited by 14,441
  4. 2017Model-Agnostic Meta-Learning for Fast Adaptation of Deep Networkscited by 4,861
  5. 2019Transformer-XL: Attentive Language Models Beyond a Fixed-Length Contextcited by 1,830
  6. 2022A Generalist Agentcited by 802
  7. 2022Efficiently Scaling Transformer Inferencecited by 529
  8. 2020Learning Quadrupedal Locomotion over Challenging Terraincited by 494
  9. 2016RL2: Fast Reinforcement Learning via Slow Reinforcement Learningcited by 490
  10. 2021RMA: Rapid Motor Adaptation for Legged Robotscited by 430
  11. 2022Learning robust perceptive locomotion for quadrupedal robots in the wildcited by 299
  12. 2021Embodied Intelligence via Learning and Evolutioncited by 154
27 more
  1. 2023Human-Timescale Adaptation in an Open-Ended Task Spacecited by 136
  2. 2023Real-World Humanoid Locomotion with Reinforcement Learningcited by 132
  3. 2022Legged Locomotion in Challenging Terrains using Egocentric Visioncited by 131
  4. 2023Extreme Parkour with Legged Robotscited by 120
  5. 2024Expressive Whole-Body Control for Humanoid Robotscited by 117
  6. 2023Robot Parkour Learningcited by 115
  7. 2020One Policy to Control Them All: Shared Modular Policies for Agent-Agnostic Controlcited by 114
  8. 2022Walk These Ways: Tuning Robot Control for Generalization with Multiplicity of Behaviorcited by 113
  9. 2024Scaling Cross-Embodied Learning: One Policy for Manipulation, Navigation, Locomotion and Aviationcited by 108
  10. 2022A Walk in the Park: Learning to Walk in 20 Minutes With Model-Free Reinforcement Learningcited by 76
  11. 2024HOVER: Versatile Neural Whole-Body Controller for Humanoid Robotscited by 76
  12. 2020Sim-to-Real Learning of All Common Bipedal Gaits via Periodic Reward Compositioncited by 71
  13. 2020Rapidly Adaptable Legged Robots via Evolutionary Meta-Learningcited by 42
  14. 2023Neural Volumetric Memory for Visual Locomotion Controlcited by 40
  15. 2023DreamWaQ: Learning Robust Quadrupedal Locomotion With Implicit Terrain Imagination via Deep Reinforcement Learningcited by 37
  16. 2022GenLoco: Generalized Locomotion Controllers for Quadrupedal Robotscited by 36
  17. 2024One Policy to Run Them All: an End-to-end Learning Approach to Multi-Embodiment Locomotioncited by 34
  18. 2023Not Only Rewards But Also Constraints: Applications on Legged Robot Locomotioncited by 33
  19. 2025BeamDojo: Learning Agile Humanoid Locomotion on Sparse Footholdscited by 32
  20. 2023Learning Vision-Based Bipedal Locomotion for Challenging Terraincited by 30
  21. 2024Body Transformer: Leveraging Robot Embodiment for Policy Learningcited by 26
  22. 2024Learning Smooth Humanoid Locomotion through Lipschitz-Constrained Policiescited by 26
  23. 2023Learning and Adapting Agile Locomotion Skills by Transferring Experiencecited by 25
  24. 2024SAPG: Split and Aggregate Policy Gradientscited by 17
  25. 2024PIE: Parkour with Implicit-Explicit Learning Framework for Legged Robotscited by 17
  26. 2025Learning Perceptive Humanoid Locomotion over Challenging Terraincited by 8
  27. 2025Symmetry-Guided Memory Augmentation for Efficient Locomotion Learningcited by 3

Built on by

Cited by year (2026 to 7 Oct)

Citing papers per year: 2026: 14.

14 papers.

Sort:

By area:

  1. 2026Gated Memory Policy: In-Context Memorization and Adaptationcited by 13
  2. 2026VLANeXt: Recipes for Building Strong VLA Modelscited by 12
  3. 2026SimToolReal: An Object-Centric Policy for Zero-Shot Dexterous Tool Manipulationcited by 10
  4. 2026An Unexpected Robot Policy: Early Evaluations of GPT-6 Astra on RoboDojo and Beyondcited by 8
  5. 2026FAR: Failure-Aware Retry for Test-Time Recovery and Continual Policy Improvementcited by 4
  6. 2026Embodiment-Aware Generalist Specialist Distillation for Unified Humanoid Whole-Body Controlcited by 2
  7. 2026Planning for Change: Reinforcement Learning Combined with Bounded Extremum Seeking for Robotic Control under Distribution Shiftcited by 2
  8. 2026DADP: Domain Adaptive Diffusion Policycited by 1
  9. 2026Shape Your Body: Value Gradients for Multi-Embodiment Robot Designcited by 1
  10. 2026StellaVLA: In-Context Structured Demonstration for Generalizable Vision-Language-Action Modelscited by 1
  11. 2026Any2Any: Efficient Cross-Embodiment Transfer for Humanoid Whole-Body Trackingcited by 0
  12. 2026TAM: Torque Adaptation Module for Robust Motion Transfer in Manipulationcited by 0
2 more
  1. 2026Amortized Low-Rank Adaptation for Model-Based Reinforcement Learningcited by 0
  2. 2026Intelligence Across Embodimentscited by 0