Pith. sign in

REVIEW 2 major objections 7 minor 78 references

Real-time motion generation obeys text and spatial goals simultaneously

Reviewed by Pith at T0; open to challenge. T0 means a machine referee read the full paper against a public rubric. the ladder, T0–T4 →

Challenge this review Re-run · record.json Download PDF Read on arXiv ↗

T0 review · glm-5.2

2026-07-10 01:57 UTC pith:7MXJ2M6I

load-bearing objection ARDY: Autoregressive Diffusion with Hybrid Representation for Interactive Human Motion Generation the 2 major comments →

arxiv 2607.08741 v1 pith:7MXJ2M6I submitted 2026-07-09 cs.GR cs.CVcs.LGcs.RO

ARDY: Autoregressive Diffusion with Hybrid Representation for Interactive Human Motion Generation

Kaifeng Zhao , Mathis Petrovich , Haotian Zhang , Tingwu Wang , Siyu Tang , Davis Rempe This is my paper
classification cs.GR cs.CVcs.LGcs.RO
keywords ardyconstraintsgenerationinteractivemotiontextcontrolkinematic
verification ladder T0 review T1 audit T2 compute T3 formal T4 reserved

The pith

A machine-rendered reading of the paper's core claim, the machinery that carries it, and where it could break.

The paper claims that interactive, real-time 3D human motion generation need not trade away controllability. By decomposing each pose into an explicit global root trajectory (which can be directly overwritten with user-specified waypoints or paths) and a compact latent body embedding (which a learned tokenizer compresses for efficient diffusion-based generation), the method achieves precise spatial control and high-fidelity body motion in a single streaming model. A two-stage denoising transformer first predicts the clean root, then predicts the latent body conditioned on that root, iterating within each diffusion step so root and body mutually constrain each other. The model accepts variable-length history (up to 8 seconds) and future constraint horizons (up to 10 seconds), enabling it to handle complex multi-step text prompts and long-horizon goals that prior autoregressive methods with short context windows cannot. Because constraints are sampled from ground-truth motion during training and injected as masked conditioning tokens, the model learns to follow them natively, without test-time optimization or reinforcement learning policies. The authors validate on both a public benchmark and a large-scale proprietary dataset, showing competitive motion quality and substantially lower constraint errors than prior offline and autoregressive baselines, including a 4-step diffusion variant achieving 33 ms generation latency.

Core claim

The central mechanism is the hybrid representation: explicit global root features concatenated with learned latent body tokens, processed by an interleaved two-stage diffusion denoiser. This decomposition lets the model overwrite root features directly for precise trajectory control while keeping the body representation compact enough for efficient few-step diffusion generation. Combined with variable-length history and out-of-window future constraint conditioning, this architecture natively supports online text prompting and flexible kinematic constraints in a single streaming framework, a combination no prior method achieved.

What carries the argument

Hybrid motion tokens (explicit global root + FSQ-quantized latent body), interleaved two-stage transformer denoiser (root-first, body-second), masked constraint injection with root overwriting, variable-length history and future-goal conditioning, latency-aware replanning buffer

Load-bearing premise

The paper assumes that decomposing generation into a root-prediction stage followed by a body-prediction stage is necessary for jointly achieving controllability and motion fidelity. The ablation supporting this shows the one-stage baseline is competitive on some metrics, so the gap is quantitative rather than a fundamental qualitative barrier.

What would settle it

A monolithic one-stage denoiser that jointly predicts root and body from the same hybrid representation, trained and evaluated under identical conditions, would match or exceed the two-stage architecture on both motion quality (FID, foot skating) and constraint adherence (joint error, trajectory error). If such a baseline existed, the central architectural claim would collapse.

Watch this falsifier — get emailed when new claim-graph text bears on it.

Editorial analysis

A structured set of objections, weighed in public.

Desk editor's note, referee report, simulated authors' rebuttal, and a circularity audit.

Referee Report

2 major / 7 minor

Summary. The paper introduces ARDY, an autoregressive diffusion model for interactive 3D human motion generation that supports online text prompting and flexible, long-horizon kinematic constraints (root waypoints/trajectories, full-body keyframes, end-effector positions/rotations) without test-time optimization or RL policies. The core technical contributions are: (1) a hybrid motion representation combining explicit global root features with a learned latent body embedding, (2) a two-stage interleaved transformer denoiser that first predicts clean root motion and then predicts latent body motion conditioned on that root, and (3) a variable-length history context and out-of-window future constraint conditioning mechanism. The method is evaluated on the public HumanML3D benchmark against offline (MaskControl) and online (DiP) baselines, and on the proprietary Bones Rigplay dataset via ablation studies. An interactive demo demonstrates real-time control via text, mouse, and keyboard inputs at 33–63 ms latency.

Significance. The paper addresses a genuine and important gap in the motion generation literature: existing offline methods offer rich controllability but are too slow for interactive use, while existing online methods are fast but sacrifice either text conditioning, kinematic control, or long-horizon planning. The combination of capabilities in Table 1—real-time generation, online text prompting, diverse spatial control types, native control without optimization or RL, and long history/future context—is unmatched by prior work. The two-stage denoiser and hybrid representation are well-motivated and validated by ablation (Table 2). The release of code and models, along with the interactive demo, strengthens reproducibility and practical impact. The evaluation on HumanML3D against both offline and autoregressive baselines is appropriate and the results are competitive or superior. The work is a solid contribution to the interactive motion generation field.

major comments (2)
  1. [§5.1, Constraint Sampling and Table 2/Table 3 evaluation] The central claim that ARDY 'natively learns controllable generation' (Sec. 1, Sec. 3.3) is primarily validated on the proprietary Bones Rigplay dataset where spatial constraints are 'sampled directly from the ground-truth test set' (Sec. 5.1). This means every constraint is guaranteed to be physically plausible and consistent with natural human motion. The only robustness test mentioned—adding 'slight random perturbations to the global translation and heading of a subset of sampled constraints' (Sec. 5.1)—reports no quantitative results for perturbed vs. unperturbed settings, and body-level constraints (joint positions/rotations, full-body keyframes) are never perturbed. On the public HumanML3D benchmark (Table 4), the raw (non-optimized) joint position error is 4.15 cm, substantially worse than the ~2.5 cm on Bones Rigplay (Table 2). This gap raises the question of whether the strong '
  2. [§3.4, Two-Stage Denoiser and Table 2 ablation] The two-stage architecture is presented as a key design decision, and the ablation in Table 2 shows improvements over the one-stage baseline. However, the one-stage baseline still achieves competitive foot skating (0.264 vs. 0.264 m/s text-only; 0.248 vs. 0.250 m/s constraints) and R-precision (65.84 vs. 65.47 text-only). The performance gap is concentrated in constraint accuracy (e.g., waypoint error 0.164 m vs. 0.024 m). The paper should more precisely characterize when the two-stage design matters most (e.g., for which constraint types) and whether the one-stage baseline could be improved with better constraint conditioning, to clarify whether the two-stage decomposition is fundamentally necessary or simply a useful heuristic that happens to help with certain constraint types.
minor comments (7)
  1. Table 2: The one-stage baseline achieves identical foot skating (0.264 m/s) and slightly higher R-precision (65.84 vs. 65.47) in text-only generation. The paper should acknowledge this more directly rather than focusing only on the improvements in constraint-conditioned metrics.
  2. §3.5, Eq. (11): The loss combines four terms with equal weight (1:1:1:1). No justification is given for this weighting. A brief ablation or discussion of sensitivity to loss weights would help reproducibility.
  3. §5.1: The Bones Rigplay dataset is proprietary and not publicly available. While the HumanML3D evaluation partially addresses this, the main ablation analysis (Tables 2–3) cannot be independently reproduced. The paper should note this limitation more explicitly.
  4. §6.2, Table 4: The paper states that using the original HumanML3D evaluator models disadvantages ARDY on FID and R-precision. However, the R-precision gap between ARDY (0.729) and MaskControl* (0.760) is non-trivial. A brief discussion of why the retargeting difference affects R-precision specifically would strengthen the comparison.
  5. Fig. 3: The architecture diagram is informative but dense. A clearer separation of the root transformer and body transformer data flows, perhaps with color-coded arrows, would improve readability for readers unfamiliar with the two-stage diffusion paradigm.
  6. §4.1: The replan buffer mechanism is described as optional, with the 4-step model using no buffer and the 10-step model using a single buffer frame. It would be helpful to report the actual end-to-end latency experienced by the user (including buffer wait time) for both configurations, not just the model inference latency.
  7. Table 3, Diffusion steps section: The 4-step model achieves FID 0.034 and the 10-step model achieves 0.027. The paper claims 'performance is still acceptable for most applications when going as low as four steps' (Sec. 3.5). Quantifying what 'acceptable' means in terms of perceptual quality would be helpful, especially given the interactive demo uses the 4-step model.

Circularity Check

0 steps flagged

No significant circularity: the architecture and losses are defined independently; the only self-citation is a non-load-bearing architectural motif borrowed from prior work.

full rationale

The paper's central claims are architectural and empirical, not derivational. The hybrid representation (Eq. 2–3) is defined by construction: a tokenizer compresses body motion into latent tokens, and the diffusion model learns to denoise them. No equation reduces to its own input by definition. The two-stage denoiser (Sec. 3.4) is a design choice validated by ablation (Table 2), not a circular derivation. The training losses (Eqs. 7–11) are standard supervised objectives comparing predictions to ground truth; none are self-referential. The claim that ARDY 'natively learns controllable generation' (Sec. 1, Sec. 3.3) is an empirical claim supported by training on ground-truth-sampled constraints and evaluating on held-out test data — it is not a prediction forced by construction. The one self-citation to Rempe et al. 2026 for the interleaved two-stage design concept (Sec. 3.4: 'Our autoregressive transformer denoiser employs an interleaved, two-stage design [Rempe et al. 2026]') is architectural borrowing, not a load-bearing mathematical premise: the paper independently defines its own two-stage architecture, trains it from scratch, and validates it via ablation against a one-stage baseline. The evaluation on the proprietary Bones Rigplay dataset (Sec. 5) is a data-availability concern, not circularity, and is independently grounded by the public HumanML3D benchmark results (Sec. 6, Tables 4–5). No fitted parameter is renamed as a prediction; no uniqueness theorem is invoked; no ansatz is smuggled through self-citation. The derivation chain is self-contained.

Axiom & Free-Parameter Ledger

6 free parameters · 4 axioms · 2 invented entities

The model relies on standard DDPM and transformer assumptions. Key hyperparameters (patch size, window size, diffusion steps) are tuned empirically. The core architectural innovations (hybrid representation, two-stage denoiser) are supported by ablation studies.

free parameters (6)
  • Patch size P = 4
    Chosen hyperparameter for the motion tokenizer (Sec. 3.5).
  • Generation window G = 40
    Chosen frame window size for generation (Sec. 3.5).
  • Diffusion steps = 10 (default), 4 (efficient)
    Number of denoising steps used during training and inference (Sec. 3.5).
  • FSQ quantization levels = 64
    Discrete quantization levels for the latent space (Sec. 3.5).
  • FSQ latent dimensions = 128
    Dimensionality of the FSQ latent space (Sec. 3.5).
  • Loss weights = 1 (hybrid, dec, goal), 0.01 (skate)
    Weights for the combined loss function L (Sec. 3.5, Eq. 11).
axioms (4)
  • standard math DDPM framework validity
    The denoiser is trained using the DDPM framework (Sec. 3.5).
  • domain assumption Transformer architecture suitability
    Assumes transformers are effective for autoregressive motion denoising (Sec. 3.4).
  • domain assumption Explicit root overwriting accuracy
    Assumes that overwriting root features with constraints during denoising leads to accurate trajectory control (Sec. 3.4).
  • ad hoc to paper Local root representation reduces foot skating
    Assumes converting global root to local representation in the tokenizer decoder mitigates foot skating (Sec. 3.2, Table 2).
invented entities (2)
  • Hybrid motion representation independent evidence
    purpose: Combine explicit root features with latent body embeddings
    Ablation in Table 2 shows it outperforms purely explicit representation.
  • Two-stage interleaved denoiser independent evidence
    purpose: Decompose root and body prediction for better control
    Ablation in Table 2 shows it outperforms a one-stage baseline.

pith-pipeline@v1.1.0-glm · 28098 in / 2354 out tokens · 388357 ms · 2026-07-10T01:57:59.129349+00:00 · methodology

0 comments
read the original abstract

Generating realistic 3D human motions in real-time within interactive applications is key for animation, simulation, and humanoid robotics. While recent offline motion generation approaches offer precise control via text and kinematic constraints, they lack the inference speed required for interactive settings. Conversely, existing online methods enable real-time synthesis but often sacrifice controllability or struggle with complex text semantics and long-horizon goals due to limited context windows. In this work, we introduce ARDY, a streaming generation framework that bridges this gap by enabling high-fidelity motion generation controllable via online text prompts and flexible kinematic constraints. ARDY employs a hybrid representation that combines explicit root features with a latent body embedding, balancing precise trajectory control with efficient generative learning. We propose a two-stage autoregressive transformer denoiser that features variable history context and supports conditioning on flexible, long-horizon kinematic constraints. By training on a large-scale motion capture dataset and being directly conditioned on text labels and kinematic constraints sampled from ground truth poses, ARDY natively learns controllable generation that supports online prompting and flexible long-horizon goals. Extensive evaluations on the HumanML3D benchmark and the large-scale, high-fidelity Bones Rigplay dataset demonstrate ARDY's high motion quality and constraint adherence, validating the efficacy of our key architectural decisions. Finally, we demonstrate the method's practical versatility through an interactive demo featuring dynamic text control, diverse keyframe pose constraints, path following, and interactive locomotion control via mouse and keyboard. Supplementary video results, code, and model releases can be found at https://research.nvidia.com/labs/sil/projects/ardy/.

Figures

Figures reproduced from arXiv: 2607.08741 by Davis Rempe, Haotian Zhang, Kaifeng Zhao, Mathis Petrovich, Siyu Tang, Tingwu Wang.

Figure 1
Figure 1. Figure 1: We present ARDY, an autoregressive diffusion model designed for interactive human motion generation. Our approach natively supports online text prompting alongside a comprehensive suite of flexible kinematic constraints — including root waypoints and trajectories, full-body keyframes, and sparse joint positions and rotations — over long horizons. ARDY enables controllable and responsive interactive motion … view at source ↗
Figure 2
Figure 2. Figure 2: Motion Tokenizer. The encoder first embeds the patchified body motion into a latent representation. This latent body motion is concatenated with the patchified global root motion to form our hybrid representation, which is decoded back to reconstruct the body motion. feature with a latent embedding. Concretely, a single pose x of a motion using the hybrid representation is a tuple x = (mroot, xbody) (2) wh… view at source ↗
Figure 3
Figure 3. Figure 3: Autoregressive Two-Stage Transformer Denoiser. (Left) Conditioned on a variable-length history context and optional spatial goal constraints, the autoregressive denoiser predicts a sequence of 𝐶 clean motion tokens within the current generation window. Spatial goal constraints can be arbitrarily sparse and may be located within or beyond the current motion generation window. (Right) The two-stage denoiser … view at source ↗
Figure 4
Figure 4. Figure 4: Interactive Demo Interface. This web interface allows generating motion and interacting with ARDY in real time. The control panel at the top right allows dynamically changing the text prompt or input constraints. Input constraints are visualized in red within the 3D scene as the model generates motion to follow them. The timeline tracks on the bottom of the interface intuitively show upcoming text prompts … view at source ↗
Figure 5
Figure 5. Figure 5: Latency-Aware Replanning. We utilize a non-blocking strategy where a buffer of 𝐵 frames is simultaneously played back and fed into the generation thread as history context. This buffer effectively hides the inference latency of slower models, ensuring that the transition to the newly generated sequence remains smooth and continuous. windows are configurable in our interactive demo, up to a maximum of 8 sec… view at source ↗
Figure 6
Figure 6. Figure 6: Motion Generation with Kinematic Constraints. Qualitative results for motion generation conditioned on text prompts and diverse kinematic constraints, including dense root trajectories, sparse root waypoints (visualized as red rings), full-body keyframes (visualized as red skeletons), sparse joint positions (visualized as white skeletons with constrained joints highlighted as red spheres), and joint rotati… view at source ↗

discussion (0)

Sign in with ORCID, Apple, or X to comment. Anyone can read and Pith papers without signing in.

Reference graph

Works this paper leans on

78 extracted references · 78 canonical work pages · 9 internal anchors

  1. [1]

    The Eleventh International Conference on Learning Representations , year=

    Human Motion Diffusion Model , author=. The Eleventh International Conference on Learning Representations , year=

    work page

  2. [2]

    Proceedings of the IEEE/CVF International Conference on Computer Vision (ICCV) , pages =

    MaskControl: Spatio-Temporal Control for Masked Motion Synthesis , author =. Proceedings of the IEEE/CVF International Conference on Computer Vision (ICCV) , pages =

    work page

  3. [3]

    Proceedings of the IEEE/CVF Conference on Computer Vision and Pattern Recognition , pages=

    Momask: Generative masked modeling of 3d human motions , author=. Proceedings of the IEEE/CVF Conference on Computer Vision and Pattern Recognition , pages=

    work page

  4. [4]

    ECCV , pages=

    Motionlcm: Real-time controllable motion generation via latent consistency model , author=. ECCV , pages=

    work page

  5. [5]

    2025 , url=

    Guy Tevet and Sigal Raab and Setareh Cohan and Daniele Reda and Zhengyi Luo and Xue Bin Peng and Amit Haim Bermano and Michiel van de Panne , booktitle=. 2025 , url=

    work page 2025

  6. [6]

    Zhao, Kaifeng and Li, Gen and Tang, Siyu , booktitle =

    work page

  7. [7]

    ACM Trans

    Shi, Yi and Wang, Jingbo and Jiang, Xuekun and Lin, Bingkun and Dai, Bo and Peng, Xue Bin , title =. ACM Trans. Graph. , month =. 2024 , issue_date =

    work page 2024

  8. [8]

    2024 , publisher =

    Taming Diffusion Probabilistic Models for Character Control , author =. 2024 , publisher =. doi:10.1145/3641519.3657440 , booktitle =

    work page doi:10.1145/3641519.3657440 2024

  9. [9]

    2024 , journal=

    Tessler, Chen and Guo, Yunrong and Nabati, Ofir and Chechik, Gal and Peng, Xue Bin , title =. 2024 , journal=

    work page 2024

  10. [10]

    Finite Scalar Quantization: VQ-VAE Made Simple

    Finite scalar quantization: Vq-vae made simple , author=. arXiv preprint arXiv:2309.15505 , year=

    work page internal anchor Pith review Pith/arXiv arXiv

  11. [11]

    European Conference on Computer Vision , pages=

    Tlcontrol: Trajectory and language control for human motion synthesis , author=. European Conference on Computer Vision , pages=. 2024 , organization=

    work page 2024

  12. [12]

    The Twelfth International Conference on Learning Representations , year=

    OmniControl: Control Any Joint at Any Time for Human Motion Generation , author=. The Twelfth International Conference on Learning Representations , year=

    work page

  13. [13]

    Proceedings of the IEEE/CVF Conference on Computer Vision and Pattern Recognition , pages=

    Executing your Commands via Motion Diffusion in Latent Space , author=. Proceedings of the IEEE/CVF Conference on Computer Vision and Pattern Recognition , pages=

    work page

  14. [14]

    Go to Zero: Towards Zero-shot Motion Generation with Million-scale Data

    Go to Zero: Towards Zero-shot Motion Generation with Million-scale Data , author=. Proceedings of the IEEE/CVF International Conference on Computer Vision (ICCV) , year=. 2507.07095 , archivePrefix=

    work page internal anchor Pith review Pith/arXiv arXiv

  15. [15]

    Proceedings of the Computer Vision and Pattern Recognition Conference , pages=

    Scamo: Exploring the scaling law in autoregressive motion generation model , author=. Proceedings of the Computer Vision and Pattern Recognition Conference , pages=

    work page

  16. [16]

    Proceedings of the Computer Vision and Pattern Recognition Conference , pages=

    Rethinking Diffusion for Text-Driven Human Motion Generation: Redundant Representations, Evaluation, and Masked Autoregression , author=. Proceedings of the Computer Vision and Pattern Recognition Conference , pages=

    work page

  17. [17]

    European Conference on Computer Vision , pages=

    Emdm: Efficient motion diffusion model for fast and high-quality motion generation , author=. European Conference on Computer Vision , pages=. 2024 , organization=

    work page 2024

  18. [18]

    UniPhys: Unified Planner and Controller with Diffusion for Flexible Physics-Based Character Control

    UniPhys: Unified Planner and Controller with Diffusion for Flexible Physics-Based Character Control , author=. arXiv preprint arXiv:2504.12540 , year=

    work page internal anchor Pith review Pith/arXiv arXiv

  19. [19]

    ACM Transactions on Graphics (TOG) , volume=

    Diffuse-cloc: Guided diffusion for physics-based character look-ahead control , author=. ACM Transactions on Graphics (TOG) , volume=. 2025 , publisher=

    work page 2025

  20. [20]

    Proceedings of the IEEE/CVF Conference on Computer Vision and Pattern Recognition , pages=

    Optimizing diffusion noise can serve as universal motion priors , author=. Proceedings of the IEEE/CVF Conference on Computer Vision and Pattern Recognition , pages=

    work page

  21. [21]

    Proceedings of the IEEE/CVF Conference on Computer Vision and Pattern Recognition , year=

    Seamless Human Motion Composition with Blended Positional Encodings , author=. Proceedings of the IEEE/CVF Conference on Computer Vision and Pattern Recognition , year=

    work page

  22. [22]

    Proceedings of the IEEE/CVF Conference on Computer Vision and Pattern Recognition , year=

    From Sparse Signal to Smooth Motion: Real-Time Motion Generation with Rolling Prediction Models , author=. Proceedings of the IEEE/CVF Conference on Computer Vision and Pattern Recognition , year=

    work page

  23. [23]

    CVPR Workshop on Human Motion Generation , year =

    Multi-Track Timeline Control for Text-Driven 3D Human Motion Generation , author =. CVPR Workshop on Human Motion Generation , year =

    work page

  24. [24]

    Proceedings of the IEEE/CVF Conference on Computer Vision and Pattern Recognition , pages=

    Mmm: Generative masked motion model , author=. Proceedings of the IEEE/CVF Conference on Computer Vision and Pattern Recognition , pages=

    work page

  25. [25]

    ACM Transactions on Graphics (TOG) , volume=

    Character controllers using motion vaes , author=. ACM Transactions on Graphics (TOG) , volume=. 2020 , publisher=

    work page 2020

  26. [26]

    ACM Transactions on Graphics (ToG) , volume=

    Deepphase: Periodic autoencoders for learning motion phase manifolds , author=. ACM Transactions on Graphics (ToG) , volume=. 2022 , publisher=

    work page 2022

  27. [27]

    ACM Transactions on Graphics , volume=

    Neural state machine for character-scene interactions , author=. ACM Transactions on Graphics , volume=. 2019 , publisher=

    work page 2019

  28. [28]

    Proceedings of the IEEE/CVF International Conference on Computer Vision , pages=

    Stochastic scene-aware motion prediction , author=. Proceedings of the IEEE/CVF International Conference on Computer Vision , pages=

    work page

  29. [29]

    Universal Humanoid Motion Representations for Physics-Based Control

    Universal humanoid motion representations for physics-based control , author=. arXiv preprint arXiv:2310.04582 , year=

    work page internal anchor Pith review Pith/arXiv arXiv

  30. [30]

    ACM Transactions on Graphics (TOG) , volume=

    Phase-functioned neural networks for character control , author=. ACM Transactions on Graphics (TOG) , volume=. 2017 , publisher=

    work page 2017

  31. [31]

    SIGGRAPH Asia 2024 Conference Papers , pages=

    Autonomous character-scene interaction synthesis from text instruction , author=. SIGGRAPH Asia 2024 Conference Papers , pages=

    work page 2024

  32. [32]

    Proceedings of the IEEE/CVF Conference on Computer Vision and Pattern Recognition , pages=

    Trace and pace: Controllable pedestrian animation via guided trajectory diffusion , author=. Proceedings of the IEEE/CVF Conference on Computer Vision and Pattern Recognition , pages=

    work page

  33. [33]

    Proceedings of the IEEE/CVF international conference on computer vision , pages=

    Humor: 3d human motion model for robust pose estimation , author=. Proceedings of the IEEE/CVF international conference on computer vision , pages=

    work page

  34. [34]

    ACM SIGGRAPH 2024 Conference Papers , pages=

    Tedi: Temporally-entangled diffusion for long-term motion synthesis , author=. ACM SIGGRAPH 2024 Conference Papers , pages=

    work page 2024

  35. [35]

    and Varol, G

    Petrovich, Mathis and Black, Michael J. and Varol, G. European Conference on Computer Vision (

    work page

  36. [36]

    IEEE transactions on pattern analysis and machine intelligence , volume=

    Motiondiffuse: Text-driven human motion generation with diffusion model , author=. IEEE transactions on pattern analysis and machine intelligence , volume=. 2024 , publisher=

    work page 2024

  37. [37]

    MotionStreamer: Streaming Motion Generation via Diffusion-based Autoregressive Model in Causal Latent Space

    MotionStreamer: Streaming Motion Generation via Diffusion-based Autoregressive Model in Causal Latent Space , author=. arXiv preprint arXiv:2503.15451 , year=

    work page internal anchor Pith review Pith/arXiv arXiv

  38. [38]

    Proceedings of the IEEE/CVF International Conference on Computer Vision , pages=

    Towards immersive human-x interaction: A real-time framework for physically plausible motion synthesis , author=. Proceedings of the IEEE/CVF International Conference on Computer Vision , pages=

    work page

  39. [39]

    Proceedings of the IEEE/CVF Conference on Computer Vision and Pattern Recognition (CVPR) , year=

    T2M-GPT: Generating Human Motion from Textual Descriptions with Discrete Representations , author=. Proceedings of the IEEE/CVF Conference on Computer Vision and Pattern Recognition (CVPR) , year=

    work page

  40. [40]

    Advances in Neural Information Processing Systems , volume=

    Motiongpt: Human motion as a foreign language , author=. Advances in Neural Information Processing Systems , volume=

    work page

  41. [41]

    ICLR , year=

    Ready-to-React: Online Reaction Policy for Two-Character Interaction Generation , author=. ICLR , year=

    work page

  42. [42]

    2025 International Conference on 3D Vision (3DV) , pages=

    Unimotion: Unifying 3d human motion synthesis and understanding , author=. 2025 International Conference on 3D Vision (3DV) , pages=. 2025 , organization=

    work page 2025

  43. [43]

    Proceedings of the IEEE/CVF International Conference on Computer Vision , pages=

    Guided motion diffusion for controllable human motion synthesis , author=. Proceedings of the IEEE/CVF International Conference on Computer Vision , pages=

    work page

  44. [44]

    BeyondMimic: From Motion Tracking to Versatile Humanoid Control via Guided Diffusion

    Beyondmimic: From motion tracking to versatile humanoid control via guided diffusion , author=. arXiv preprint arXiv:2508.08241 , year=

    work page internal anchor Pith review Pith/arXiv arXiv

  45. [45]

    ASAP: Aligning Simulation and Real-World Physics for Learning Agile Humanoid Whole-Body Skills

    Asap: Aligning simulation and real-world physics for learning agile humanoid whole-body skills , author=. arXiv preprint arXiv:2502.01143 , year=

    work page internal anchor Pith review Pith/arXiv arXiv

  46. [46]

    arXiv preprint arXiv:2510.05070 , year=

    ResMimic: From General Motion Tracking to Humanoid Whole-body Loco-Manipulation via Residual Learning , author=. arXiv preprint arXiv:2510.05070 , year=

    work page arXiv

  47. [47]

    ACM SIGGRAPH 2024 Conference Proceedings , location =

    Setareh, Cohan and Tevet, Guy and Reda, Daniele and Peng, Xue Bin and van de Panne, Michiel , title =. ACM SIGGRAPH 2024 Conference Proceedings , location =. 2024 , publisher =

    work page 2024

  48. [48]

    Proceedings of the IEEE/CVF Conference on Computer Vision and Pattern Recognition (CVPR) , month =

    Guo, Chuan and Zou, Shihao and Zuo, Xinxin and Wang, Sen and Ji, Wei and Li, Xingyu and Cheng, Li , title =. Proceedings of the IEEE/CVF Conference on Computer Vision and Pattern Recognition (CVPR) , month =. 2022 , pages =

    work page 2022

  49. [49]

    , title =

    Zhang, Yan and Feng, Yao and Cseke, Alpár and Saini, Nitin and Bajandas, Nathan and Heron, Nicolas and Black, Michael J. , title =. Proceedings of the IEEE/CVF International Conference on Computer Vision (ICCV) , month = oct, year =

    work page

  50. [50]

    NeurIPS , year=

    InsActor: Instruction-driven Physics-based Characters , author=. NeurIPS , year=

    work page

  51. [51]

    IEEE transactions on pattern analysis and machine intelligence , volume=

    Gaussian process dynamical models for human motion , author=. IEEE transactions on pattern analysis and machine intelligence , volume=. 2007 , publisher=

    work page 2007

  52. [52]

    Proceedings of the IEEE international conference on computer vision , pages=

    Recurrent network models for human dynamics , author=. Proceedings of the IEEE international conference on computer vision , pages=

    work page

  53. [53]

    Advances in neural information processing systems , volume=

    Modeling human motion using binary latent variables , author=. Advances in neural information processing systems , volume=

    work page

  54. [54]

    European Conference on Computer Vision , pages=

    Bamm: Bidirectional autoregressive motion model , author=. European Conference on Computer Vision , pages=. 2024 , organization=

    work page 2024

  55. [55]

    Big data , volume=

    The kit motion-language dataset , author=. Big data , volume=. 2016 , publisher=

    work page 2016

  56. [56]

    2018 , publisher=

    Improving language understanding by generative pre-training , author=. 2018 , publisher=

    work page 2018

  57. [57]

    Proceedings of the IEEE/CVF international conference on computer vision , pages=

    Synthesizing diverse human motions in 3d indoor scenes , author=. Proceedings of the IEEE/CVF international conference on computer vision , pages=

    work page

  58. [58]

    Proceedings of the IEEE/CVF Conference on Computer Vision and Pattern Recognition , pages=

    The wanderings of odysseus in 3d scenes , author=. Proceedings of the IEEE/CVF Conference on Computer Vision and Pattern Recognition , pages=

    work page

  59. [59]

    ACM Transactions On Graphics (TOG) , volume=

    Ase: Large-scale reusable adversarial skill embeddings for physically simulated characters , author=. ACM Transactions On Graphics (TOG) , volume=. 2022 , publisher=

    work page 2022

  60. [60]

    Proceedings of the IEEE/CVF conference on computer vision and pattern recognition , pages=

    On the continuity of rotation representations in neural networks , author=. Proceedings of the IEEE/CVF conference on computer vision and pattern recognition , pages=

    work page

  61. [61]

    2024 , url=

    Parishad BehnamGhader and Vaibhav Adlakha and Marius Mosbach and Dzmitry Bahdanau and Nicolas Chapados and Siva Reddy , booktitle=. 2024 , url=

    work page 2024

  62. [62]

    Proceedings of the IEEE/CVF international conference on computer vision , pages=

    AMASS: Archive of motion capture as surface shapes , author=. Proceedings of the IEEE/CVF international conference on computer vision , pages=

    work page

  63. [63]

    Proceedings of the 28th ACM international conference on multimedia , pages=

    Action2motion: Conditioned generation of 3d human motions , author=. Proceedings of the 28th ACM international conference on multimedia , pages=

    work page

  64. [64]

    International Conference on Learning Representations , year=

    Auto-Encoding Variational Bayes , author=. International Conference on Learning Representations , year=

    work page

  65. [65]

    Girshick, Ross , booktitle=

    work page

  66. [66]

    NeurIPS 2021 Workshop on Deep Generative Models and Downstream Applications , year=

    Classifier-Free Diffusion Guidance , author=. NeurIPS 2021 Workshop on Deep Generative Models and Downstream Applications , year=

    work page 2021

  67. [67]

    and Varol, G

    Petrovich, Mathis and Black, Michael J. and Varol, G. International Conference on Computer Vision (

    work page

  68. [68]

    Viser: Imperative, Web-based 3D Visualization in Python

    Viser: Imperative, web-based 3d visualization in python , author=. arXiv preprint arXiv:2507.22885 , year=

    work page internal anchor Pith review Pith/arXiv arXiv

  69. [69]

    The Thirty-ninth Annual Conference on Neural Information Processing Systems , year=

    Mean Flows for One-step Generative Modeling , author=. The Thirty-ninth Annual Conference on Neural Information Processing Systems , year=

    work page

  70. [70]

    The Thirteenth International Conference on Learning Representations , year=

    Simplifying, Stabilizing and Scaling Continuous-time Consistency Models , author=. The Thirteenth International Conference on Learning Representations , year=

    work page

  71. [71]

    , journal =

    Loper, Matthew and Mahmood, Naureen and Romero, Javier and Pons-Moll, Gerard and Black, Michael J. , journal =

    work page

  72. [72]

    Advances in neural information processing systems , volume=

    Denoising diffusion probabilistic models , author=. Advances in neural information processing systems , volume=

    work page

  73. [73]

    2024 , url =

    Llama 3 Model Card , author=. 2024 , url =

    work page 2024

  74. [74]

    International Conference on Machine Learning , year=

    Scaling exponents across parameterizations and optimizers , author=. International Conference on Machine Learning , year=

    work page

  75. [75]

    ACM Transactions on Graphics (ToG) , year=

    Learned motion matching , author=. ACM Transactions on Graphics (ToG) , year=

    work page

  76. [76]

    ACM Transactions on Graphics (TOG) , year=

    Control Operators for Interactive Character Animation , author=. ACM Transactions on Graphics (TOG) , year=

    work page

  77. [77]

    arXiv:2603.15546 , year=

    Kimodo: Scaling Controllable Human Motion Generation , author=. arXiv:2603.15546 , year=

    work page arXiv

  78. [78]

    SONIC: Supersizing Motion Tracking for Natural Humanoid Whole-Body Control

    SONIC: Supersizing Motion Tracking for Natural Humanoid Whole-Body Control , author=. arXiv preprint arXiv:2511.07820 , year=

    work page internal anchor Pith review Pith/arXiv arXiv