mlsanityData to 14 Oct 2026

Riemannian Motion Generation: A Unified Framework for Human Motion Representation and Generation via Riemannian Flow Matching

arXiv 2603.15016 ↗First posted 16 Mar 2026cs.CV stat.MLcited by 2

On the map: Diffusion models → Flow matching, time series and science

Abstract (arXiv)

Human motion generation is often learned in Euclidean spaces, although valid motions follow structured non-Euclidean geometry. We present Riemannian Motion Generation (RMG), a unified framework that represents motion on a product manifold and learns dynamics via Riemannian flow matching. RMG factorizes motion into several manifold factors, yielding a scale-free representation with intrinsic normalization, and uses geodesic interpolation, tangent-space supervision, and manifold-preserving ODE integration for training and sampling. On HumanML3D, RMG achieves state-of-the-art FID in the HumanML3D format (0.043) and ranks first on all reported metrics under the MotionStreamer format. On MotionMillion, it also surpasses strong baselines (FID 5.6, R@1 0.86). Ablations show that the compact mathscrT+mathscrR (translation + rotations) representation is the most stable and effective, highlighting geometry-aware modeling as a practical and scalable route to high-fidelity motion generation.

Key facts. Cited by 2 papers in the map (data through 14 Oct 2026). Placed in Flow matching, time series and science, one of 14 directions in Diffusion models. Builds on 37 of its 55 references that are in the map. First posted on arXiv 16 Mar 2026.

Builds on

37 of 55 references.

Closest ancestors · 4

Same direction, newest first

  1. 2024Flow Matching Guide and Code9 Dec
  2. 2023Flow Matching on General Geometries7 Feb
  3. 2022Flow Matching for Generative Modeling6 Oct
  4. 2022Flow Straight and Fast: Learning to Generate and Transfer Data with Rectified Flow7 Sep

Other references · 33

  1. 2025Qwen3 Technical Reportcited by 8,820
  2. 2022Classifier-Free Diffusion Guidancecited by 5,847
  3. 2024Scaling Rectified Flow Transformers for High-Resolution Image Synthesiscited by 4,181
  4. 2015Variational Inference with Normalizing Flowscited by 2,170
  5. 2025Qwen3 Embedding: Advancing Text Embedding and Reranking Through Foundation Modelscited by 1,261
  6. 2019AMASS: Archive of Motion Capture as Surface Shapescited by 1,236
  7. 2018On the Continuity of Rotation Representations in Neural Networkscited by 1,214
  8. 2022Human Motion Diffusion Modelcited by 1,080
  9. 2023Stochastic Interpolants: A Unifying Framework for Flows and Diffusionscited by 939
  10. 2022MotionDiffuse: Text-Driven Human Motion Generation with Diffusion Modelcited by 780
  11. 2023T2M-GPT: Generating Human Motion from Textual Descriptions with Discrete Representationscited by 588
  12. 2022Executing your Commands via Motion Diffusion in Latent Spacecited by 575
21 more
  1. 2023MotionGPT: Human Motion as a Foreign Languagecited by 527
  2. 2021Action-Conditioned 3D Human Motion Synthesis with Transformer VAEcited by 501
  3. 2022TEMOS: Generating diverse human motions from textual descriptionscited by 492
  4. 2021AI Choreographer: Music Conditioned 3D Dance Generation with AIST++cited by 485
  5. 2020Action2Motion: Conditioned Generation of 3D Human Motionscited by 441
  6. 2023MoMask: Generative Masked Modeling of 3D Human Motionscited by 400
  7. 2022EDGE: Editable Dance Generation From Musiccited by 287
  8. 2025Z-Image: An Efficient Image Generation Foundation Model with Single-Stream Diffusion Transformercited by 277
  9. 2022MoFusion: A Framework for Denoising-Diffusion-based Motion Synthesiscited by 218
  10. 2022FLAME: Free-form Language-based Motion Synthesis & Editingcited by 217
  11. 2024MotionLCM: Real-time Controllable Motion Generation via Latent Consistency Modelcited by 168
  12. 2023Human Motion Generation: A Surveycited by 96
  13. 2025MotionStreamer: Streaming Motion Generation via Diffusion-based Autoregressive Model in Causal Latent Spacecited by 86
  14. 2024Rethinking Diffusion for Text-Driven Human Motion Generation: Redundant Representations, Evaluation, and Masked Autoregressioncited by 85
  15. 2020Dance Revolution: Long-Term Dance Generation with Music via Curriculum Learningcited by 81
  16. 2025Go to Zero: Towards Zero-shot Motion Generation with Million-scale Datacited by 68
  17. 2024ScaMo: Exploring the Scaling Law in Autoregressive Motion Generation Modelcited by 53
  18. 2025MotionLab: Unified Human Motion Generation and Editing via the Motion-Condition-Motion Paradigmcited by 30
  19. 2020Structure-Aware Human-Action Generationcited by 28
  20. 2021Generative Adversarial Graph Convolutional Networks for Human Action Synthesiscited by 23
  21. 2024Pay Attention and Move Better: Harnessing Attention for Interactive Motion Generation and Training-free Editingcited by 18

Built on by

Cited by year (2026 to 14 Oct)

Citing papers per year: 2026: 2.

2 papers.

  1. 2026Spatial Temporal Synergy: Balancing Change and Invariance in Text Driven 3D Human Motion Editingcited by 1
  2. 2026Skeletons in Flow: Graph Structured Flow Matching for Human Motion Predictioncited by 0

Similar, not cited

Close in wording, no citation either way

  1. 2026Riemannian MeanFlow for One-Step Generation on Manifolds
  2. 2026PRISM: Streaming Human Motion Generation with Per-Joint Latent Decomposition
  3. 2025Riemannian Consistency Model
  4. 2024Motion Manifold Flow Primitives for Task-Conditioned Trajectory Generation under Complex Task-Motion Dependencies
  5. 2024Riemannian Flow Matching Policy for Robot Motion Learning
  6. 2026Self-Intersection-Aware 3D Human Motion Generation Using an Efficient Human Sphere Proxy
  7. 2026Manifold-Stable Flow Matching
  8. 2026Learning from Demonstrations over Riemannian Manifolds using Neural ODEs: An Extended Abstract ↗