mlsanityData to 7 Oct 2026

SMP: Reusable Score-Matching Motion Priors for Physics-Based Character Control

arXiv 2512.03028 ↗First posted 2 Dec 2025cs.GR cs.AI cs.CVcited by 14

On the map: Offline RL and imitation learning from data → Imitation learning methods

Abstract (arXiv)

Data-driven motion priors that can guide agents toward producing naturalistic behaviors play a pivotal role in creating life-like virtual characters. Adversarial imitation learning has been a highly effective method for learning motion priors from reference motion data. However, adversarial priors, with few exceptions, need to be retrained for each new controller, thereby limiting their reusability and necessitating the retention of the reference motion data when applied to downstream tasks. In this work, we present Score-Matching Motion Priors (SMP), which leverages pre-trained motion diffusion models and score distillation sampling (SDS) to create reusable task-agnostic motion priors. SMPs can be pre-trained on a motion dataset, independent of any control policy or task. Once trained, SMPs can be kept frozen and reused as general-purpose reward functions to train new policies to produce naturalistic behaviors for downstream tasks. We show that a general motion prior trained on large-scale datasets can be repurposed into a variety of style-specific priors. Furthermore, SMP can compose different styles to synthesize new styles not present in the original dataset. Our method can create reusable and modular motion priors that produce high-quality motions comparable to state-of-the-art adversarial imitation learning methods. In our experiments, we demonstrate the effectiveness of SMP across a diverse suite of control tasks with physically simulated humanoid characters. Video available at https://youtu.be/jBA2tWk6vzU

Key facts. Cited by 14 papers in the map (data through 7 Oct 2026). Placed in Imitation learning methods, one of 13 directions in Offline RL and imitation learning from data. Builds on 44 of its 88 references that are in the map. First posted on arXiv 2 Dec 2025.

Builds on

44 of 88 references.

Closest ancestors · 3

Same direction, newest first

  1. 2024Diffusing States and Matching Scores: A New Framework for Imitation Learning17 Oct
  2. 2024Diffusion Imitation from Observation7 Oct
  3. 2023DiffAIL: Diffusion Adversarial Imitation Learning11 Dec

Other references · 41

  1. 2020Denoising Diffusion Probabilistic Modelscited by 20,851
  2. 2017Proximal Policy Optimization Algorithmscited by 14,441
  3. 2020Denoising Diffusion Implicit Modelscited by 9,655
  4. 2020Score-Based Generative Modeling through Stochastic Differential Equationscited by 8,964
  5. 2021Diffusion Models Beat GANs on Image Synthesiscited by 8,185
  6. 2019Generative Modeling by Estimating Gradients of the Data Distributioncited by 4,067
  7. 2022DreamFusion: Text-to-3D using 2D Diffusioncited by 3,006
  8. 2023Diffusion Policy: Visuomotor Policy Learning via Action Diffusioncited by 2,730
  9. 2015High-Dimensional Continuous Control Using Generalized Advantage Estimationcited by 1,916
  10. 2016Generative Adversarial Imitation Learningcited by 1,576
  11. 2022Magic3D: High-Resolution Text-to-3D Content Creationcited by 1,285
  12. 2023ProlificDreamer: High-Fidelity and Diverse Text-to-3D Generation with Variational Score Distillationcited by 1,207
29 more
  1. 2018On the Continuity of Rotation Representations in Neural Networkscited by 1,186
  2. 2022Human Motion Diffusion Modelcited by 1,062
  3. 2023One-step Diffusion with Distribution Matching Distillationcited by 978
  4. 2021Isaac Gym: High Performance GPU-Based Physics Simulation For Robot Learningcited by 882
  5. 2022MotionDiffuse: Text-Driven Human Motion Generation with Diffusion Modelcited by 768
  6. 2017Sim-to-Real Transfer of Robotic Control with Dynamics Randomizationcited by 616
  7. 2020Action2Motion: Conditioned Generation of 3D Human Motionscited by 438
  8. 2023Analyzing and Improving the Training Dynamics of Diffusion Modelscited by 396
  9. 2021HuMoR: 3D Human Motion Model for Robust Pose Estimationcited by 319
  10. 2023LucidDreamer: Towards High-Fidelity Text-to-3D Generation via Interval Score Matchingcited by 290
  11. 2021Robust Motion In-betweeningcited by 223
  12. 2017Asymmetric Actor Critic for Image-Based Robot Learningcited by 202
  13. 2021Character Controllers Using Motion VAEscited by 182
  14. 2024CLoSD: Closing the Loop between Simulation and Diffusion for multi-task character controlcited by 103
  15. 2022Real-Time Style Modelling of Human Locomotion via Feature-Wise Transformations and Local Motion Phasescited by 71
  16. 2023MoConVQ: Unified Physics-Based Motion Control via Scalable Discrete Representationscited by 58
  17. 2023Interactive Character Control with Auto-Regressive Motion Diffusion Modelscited by 52
  18. 2024DiffuseLoco: Real-Time Legged Locomotion Control with Diffusion from Offline Datasetscited by 47
  19. 2025TokenHSI: Unified Synthesis of Physical Human-Scene Interactions through Task Tokenizationcited by 46
  20. 2023Neural Categorical Priors for Physics-Based Character Controlcited by 37
  21. 2024PDP: Physics-Based Character Animation via Diffusion Policycited by 30
  22. 2025Diffuse-CLoC: Guided Diffusion for Physics-based Character Look-ahead Controlcited by 27
  23. 2025UniPhys: Unified Planner and Controller with Diffusion for Flexible Physics-Based Character Controlcited by 24
  24. 2025PARC: Physics-based Augmentation with Reinforcement Learning for Character Controllerscited by 24
  25. 2023InsActor: Instruction-driven Physics-based Characterscited by 23
  26. 2022ControlVAE: Model-Based Learning of Generative Controllers for Physics-Based Characterscited by 21
  27. 2025Learning Getting-Up Policies for Real-World Humanoid Robotscited by 21
  28. 2023Animating Human Athleticscited by 16
  29. 2022Learning to Get Upcited by 11

Built on by

Cited by year (2026 to 7 Oct)

Citing papers per year: 2026: 14.

14 papers.

Sort:

By area:

  1. 2026SCRIPT: Scalable Diffusion Policy with Multi-stage Training for Language-driven Physics-Based Humanoid Controlcited by 4
  2. 2026PAC-MAN: Perception-Aware CBF-RL for Whole-Body Safety in Humanoid Dodgeballcited by 2
  3. 2026PDF-HR: Pose Distance Fields for Humanoid Robotscited by 1
  4. 2026ExpertGen: Scalable Sim-to-Real Expert Policy Learning from Imperfect Behavior Priorscited by 1
  5. 2026MIND: Multi-Scale Intent Diffusion for Text-Driven Physics-Based Humanoid Controlcited by 1
  6. 2026Sketch2Motion: Text-driven 2D Sketch to 3D Animation via Diffusion-guided Skeleton Optimizationcited by 1
  7. 2026MotionRFT: Unified Reinforcement Fine-Tuning for Text-to-Motion Generationcited by 0
  8. 2026PFM-HR: Pose Flow Matching for Humanoid Robotscited by 0
  9. 2026Progressively Learning Heterogeneous Skills in a Unified Latent Spacecited by 0
  10. 2026Flow-Matched Motion Priors: Online Optimal-Transport Rewards for Imitation Learningcited by 0
  11. 2026DSD: Learning Diverse and Reusable Motor Skills via Diffusion Skill Discoverycited by 0
  12. 2026LYRIC: Language-Driven Physics-Based Character Control for Contact-Rich Whole-Body Object Interactioncited by 0
2 more
  1. 2026OmniMimic: Dynamics-completed Motion Augmentation for Multi-style Omnidirectional Quadruped Locomotioncited by 0
  2. 2026World Motion Models: Flexible Sequence Modeling of SE(3) Trajectoriescited by 0

Similar, not cited

Close in wording, no citation either way

  1. 2021AMP: Adversarial Motion Priors for Stylized Physics-Based Character Control
  2. 2022ASE: Large-Scale Reusable Adversarial Skill Embeddings for Physically Simulated Characters
  3. 2026Task-Centric Policy Optimization from Misaligned Motion Priors
  4. 2023CALM: Conditional Adversarial Latent Models for Directable Virtual Characters
  5. 2021A GAN-Like Approach for Physics-Based Imitation Learning and Interactive Character Control
  6. 2023C·ASE: Learning Conditional Adversarial Skill Embeddings for Physics-based Characters
  7. 2023Synthesizing Physical Character-Scene Interactions
  8. 2025ModSkill: Physical Character Skill Modularization