Pith. sign in

REVIEW 3 major objections 5 minor 1 cited by

ScaleMoGen claims that text-driven human motion generation is best framed as coarse-to-fine next-scale prediction over a hierarchy of discrete skeletal-temporal token maps, rather than as next-token prediction along time, and that this fram

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 ↗

ScaleMoGen applies next-scale autoregressive prediction to human motion generation with multi-scale skeletal-temporal bitwise token maps, reporting SOTA FID on HumanML3D.

T0 review reviewed 2026-08-02 challenge →

load-bearing objection Strong results, but the main configuration doesn't actually use the recursive skeletal hierarchy the paper is built around—temporal multi-scale plus fixed atomic segments may be the real story. the 3 major comments →

arxiv 2605.11704 v2 pith:TIH647K3 submitted 2026-05-12 cs.CV

ScaleMoGen: Autoregressive Next-Scale Prediction for Human Motion Generation

Inwoo Hwang , Hojun Jang , Bing Zhou , Jian Wang , Young Min Kim , Chuan Guo This is my paper
classification cs.CV
keywords multi-scale discrete tokenizationnext-scale predictionautoregressive motion generationtext-to-motionbitwise quantizationskeletal hierarchyzero-shot motion editingcoarse-to-fine generation
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 reading

ScaleMoGen sets out to show that text-driven human motion generation is best approached not as next-token prediction in time, but as coarse-to-fine next-scale prediction over a hierarchy of discrete token maps. The paper builds a multi-scale skeletal-temporal representation: motion is quantized into residual token maps of increasing temporal and skeletal granularity, using bitwise quantization that preserves the kinematic hierarchy, and a transformer predicts each finer map conditioned on all coarser ones and the text. On two standard benchmarks, this yields the best reported fidelity and text-alignment scores. The same representation makes training-free text-guided motion editing possible by selectively preserving or resampling tokens at chosen scales and body parts. If correct, this reframes a generation problem that was temporal into one that is structural.

Core claim

The paper's central claim is that human motion is naturally a coarse-to-fine structure, and that a generative model should mirror this: start with a token that encodes the whole body's global motion, then progressively add finer temporal resolution and finer skeletal decompositions. To carry this, ScaleMoGen defines a hierarchy of token maps where each scale has its own temporal resolution and skeletal partition (satisfying completeness and recursive refinement), with the finest scale at 7 atomic segments—root, spine, head, two arms, two legs. The motion encoder produces a continuous latent grid, and residual bitwise quantization turns it into binary token maps at every scale; topology-aware

What carries the argument

The load-bearing object is the multi-scale skeletal-temporal token map: a tuple of discrete grids (time and skeletal segments) at scales v=0..V, produced by residual bitwise quantization where each latent dimension becomes a binary label, so the effective vocabulary is 2^d. The operators that make it work are the topology-aware downsampling and upsampling functions, which move between the full-resolution latent grid (n x j x d, with j=7 atomic segments) and each scale's coarser grid while respecting the kinematic tree; the recursive-refinement property ensures coarser segments are unions of finer ones. This representation converts generation into next-scale prediction: a transformer, conditi

Load-bearing premise

The load-bearing premise is that the fixed, hand-defined skeletal hierarchy (whole body, then upper/lower body, then seven atomic segments: root, spine, head, arms, legs) together with its temporal upsampling schedule is sufficient to express the semantic structure of every motion that text can describe; if some motion requires a finer or differently grouped joint partition, the token maps cannot represent it and the generation claim fails.

What would settle it

Prompt a released ScaleMoGen to generate a motion requiring independent finger counting (e.g., 'counts to three with fingers raised one by one'). If the output cannot show distinct finger configurations—because fingers are not atomic segments in the 7-part hierarchy—then the skeletal hierarchy does not cover all human motions, falsifying the central claim that this multi-scale representation is sufficient for text-to-motion.

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

If this is right

  • If the next-scale formulation is correct, autoregressive motion generation no longer needs long temporal token sequences: a constant number of scale steps (7 in the paper) produces full motions, cutting inference cost relative to diffusion samplers that require many denoising steps.
  • Text-guided editing becomes a token-preservation operation: users can keep coarse tokens for global structure, or mask specific body parts and time intervals, and the model resamples only the masked tokens conditioned on the target text—no extra training.
  • Bitwise quantization gives an exponentially large discrete vocabulary (2^d) that the paper argues preserves fine motion detail better than fixed codebooks, which could make discrete token approaches more competitive with continuous diffusion on fidelity.
  • Scaling the predictor to 2.2B parameters did not improve alignment and slightly hurt the fidelity metric on current datasets, implying that for text-to-motion, data quantity—not model capacity—is the current limiting factor.

Where Pith is reading between the lines

These are editorial extensions of the paper, not claims the author makes directly.

  • The coarse-to-fine token hierarchy suggests a generic recipe for other structured temporal signals (audio, gestures, graphs): predict a hierarchical latent map from coarse to fine via residual bitwise quantization. A testable extension would apply the same architecture to speech or music generation and compare next-scale prediction against next-token baselines.
  • The fixed 7-segment skeletal partition may be too coarse for actions that live in hands or face; a learnable hierarchy—discovering segments from data—could extend the method to fine-grained control. The paper's own ablation shows dataset-specific sensitivity to hierarchy design, so a data-driven partition is a natural next step.
  • The zero-shot mask could be used beyond editing: e.g., motion interpolation between two text prompts by progressively blending token maps at different scales, or controllable generation where the user specifies which body part follows which text. These are not explored in the paper.
  • The scale-adaptive guidance schedule (stronger text guidance at finer scales) implies fine details are more text-dependent than global structure; a similar scheduling principle might improve other autoregressive or diffusion systems, though this is an editorial extrapolation.
Share X Bluesky LinkedIn Reddit HN

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

3 major / 5 minor

Summary. ScaleMoGen proposes an autoregressive next-scale prediction framework for text-driven 3D human motion generation. The motion is encoded into a skeletal-temporal latent grid and progressively quantized into multi-scale bitwise token maps; a transformer then predicts the token map at each scale conditioned on accumulated coarser scales, starting from a 1x1 map. The authors claim state-of-the-art results on HumanML3D and SnapMoGen, and additionally propose a training-free text-guided editing mechanism based on preserving/masking tokens at chosen scales, skeletal parts, or semantic-confidence thresholds. The paper includes ablations on skeletal topology, code size, model size, VQ design, CFG scheduling, and sampling efficiency, plus a small user study for editing.

Significance. If the central claims hold, ScaleMoGen would be a meaningful step in discrete motion generation: shifting from temporal next-token prediction to scale-wise coarse-to-fine prediction is a natural and underexplored idea, and the bitwise multi-scale VQ design is a clean way to avoid categorical codebook collapse. The reported HumanML3D FID (0.030) and SnapMoGen CLIP Score (0.693) are competitive with recent strong baselines, and the zero-shot editing concept is attractive. The paper also provides useful ablations (hierarchy strategy, code size, model size, CFG schedule) and reports sampling efficiency. However, the manuscript's central claim that the method exploits a recursive skeletal coarse-to-fine hierarchy is not actually supported by the configuration used for the main results, and the 'state-of-the-art' claim is metric-dependent, so the significance of the contribution as stated is uncertain.

major comments (3)
  1. [B.1 + Table 7] The main configuration (Strategy 5) is described as a '7-scale full hierarchy, representing independent anatomical joint groups across all scales.' This directly contradicts Section 3.1, which defines S(0)={J}, S(V)=atomic segments, and recursive refinement with varying m_v. If m_v=7 at every scale, the spatial partition is fixed atomic and the skeletal axis never coarsens; only temporal resolution changes. Then Eq. (2) and the central 'skeletal-temporal multi-scale' claim are not what Tables 1-2 exercise. Table 7 also uses 32-bit codes for the HumanML3D column, so those comparisons are not aligned with the 24-bit main results. Please state the explicit (h_v, m_v) sequence for Strategy 5, and either run the main experiments with a true skeletal hierarchy or revise the paper's title/abstract/claims to reflect a fixed skeletal grouping with multi-scale temporal refinement.
  2. [Abstract + Tables 1-2] The abstract's 'state-of-the-art' claim is metric-dependent. On SnapMoGen, ScaleMoGen's FID (16.35) is worse than MoMask++ (15.06), while only CLIP Score is better (0.693 vs 0.685). On HumanML3D, SALAD has higher Top-1 R-Precision (0.581 vs 0.577). The abstract cherry-picks FID for HumanML3D and CLIP Score for SnapMoGen. Please qualify the claim per metric and per dataset, and discuss the FID/CLIP tradeoff on SnapMoGen explicitly.
  3. [Section 4.3 + Table 5] The editing user study is based on only 10 examples, and Table 5 reports means and standard deviations without any significance test, inter-rater reliability, or breakdown by participant/example. The statement that ScaleMoGen 'achieves the highest scores across all criteria' is not statistically supported. Since zero-shot editing is a headline contribution, either enlarge the study and add statistical analysis, or substantially soften the claim.
minor comments (5)
  1. [Abstract] Typo: 'skeletal-emporal' should be 'skeletal-temporal'.
  2. [Section 3.2] 'The world-level text embeddings c' should be 'word-level text embeddings'.
  3. [B.1] The descriptions of Strategies 1-5 are verbal and hard to parse; please include an explicit table listing h_v, m_v, and the skeletal partition for each scale for each strategy. Figure 5 shows pooling to the 7-joint atomic topology but not the scale schedule.
  4. [Table 3] The row ordering for the 32-bit code is confusing: the text says 'increasing to 32 bits yields the opposite trend' but the FID improves (15.98) and CLIP degrades (0.683); this is opposite to 16-bit but the row order is 16, 32, 24 in the text while the table lists Base, A, B, C. Consider reordering for readability.
  5. [Section 4.2] The ablation in Table 3(A) removes skeletal topology entirely (binary condition). It does not vary the skeletal partition, so it cannot validate the specific choice of the 7 atomic segments or the recursive refinement hierarchy. Please discuss this limitation.

Circularity Check

0 steps flagged

No significant circularity: the next-scale prediction is a learned model output, not a fitted constant, and the self-citations are non-load-bearing external benchmarks.

full rationale

The paper's derivation chain is self-contained in the relevant sense. The multi-scale token maps are defined by residual quantization (Eq. 2) and reconstruction (Eq. 3); the transformer is trained with the next-scale likelihood objective (Eq. 7) and sampled at inference (Eq. 8). The generated token maps are outputs of a learned conditional distribution, not fitted parameters renamed as predictions. Evaluation metrics are computed on held-out test sets of HumanML3D and SnapMoGen, so the FID/CLIP comparisons are empirical rather than forced by construction. The paper cites MoMask++ [5], including overlapping authors, and uses the SnapMoGen dataset introduced in that prior work, but these are external data/benchmarks and are not used as the logical premise of the method; the comparison is empirically falsifiable and does not constitute load-bearing self-citation. I also considered the supplementary wording for Strategy 5 in B.1: '7-scale full hierarchy, representing independent anatomical joint groups across all scales' could be read as contradicting the recursive coarse-to-fine skeletal partition S(0)={J}, ..., S(V)=atomic described in Section 3.1. That is an internal-consistency/correctness risk about whether the main configuration actually exercises the claimed spatial hierarchy, but it is not circularity: even if the hierarchy were not implemented as advertised, the reported numbers would be empirical results, not an equation reducing to its own input. No equation in the paper reduces to a fitted parameter or to a self-citation chain, so no circular step is established.

Axiom & Free-Parameter Ledger

6 free parameters · 4 axioms · 0 invented entities

The central claim rests on the fixed 7-segment skeletal partition, the SALAD-derived encoder, the bitwise quantization with its training behavior, and the T5 text encoder. These are all either borrowed from prior work or hand-chosen hyperparameters, but none are new postulated entities. The free parameters (code size, number of scales, CFG scale, perturbation rate, editing thresholds) are all manually selected and only partially ablated.

free parameters (6)
  • Code dimension d = 24 (base; 16/32 ablated)
    Chosen as a trade-off between reconstruction fidelity and prediction difficulty; Table 3(B) shows 16-bit improves CLIP but worsens FID, 32-bit the opposite.
  • Number of scales (V+1) = 7 (Strategy 5); 10-11 in other strategies
    The hierarchy depth is a design choice; Table 7 compares 5 strategies with 7–11 scales and selects based on validation metrics.
  • CFG guidance scale S_base = 5 (increasing schedule)
    Table 8 ablates S_base ∈ {3,5,7} and guidance schedules; 5 with increasing schedule is selected.
  • Bit perturbation probability r = 0.3
    Chosen for training robustness; no ablation reported.
  • Global structure mask threshold γ = 3
    Retain tokens up to scale 3 for editing preservation; chosen without ablation.
  • Semantic-aware mask threshold τ = 0.1
    Confidence threshold for token resampling in editing; chosen without ablation.
axioms (4)
  • domain assumption The 7 atomic skeletal segments (root, spine, head, four extremities) and the recursive partition tree are a faithful abstraction of human motion semantics across all datasets.
    Section 3.1 defines the partition; all multi-scale token maps are built on this fixed 7-segment topology. If a motion's semantics rely on finer joint distinctions not captured by this grouping, the representation would be lossy.
  • domain assumption The continuous latent f = E(m) from the SALAD-style encoder preserves sufficient information for generation.
    Section 3.1 uses a skeletal-temporal encoder; the reconstruction quality of the VQ (Table 4) is good but not perfect, so generation inherits any latent information loss.
  • domain assumption The bitwise quantization Q_b with a straight-through estimator yields trainable gradients and a sufficiently expressive discrete space.
    Eq. (1) is central; training stability is reported but not analyzed. The method relies on the empirical behavior of binary spherical quantization.
  • domain assumption T5-base text embeddings provide adequate text conditioning; the text-to-motion alignment gains are not due to a stronger text encoder.
    The paper uses T5-base; a stronger text encoder could potentially shift results, but this is constant across comparisons.

reviewed 2026-08-02 · how reviews work

0 comments
Cite this review

Pith. "Pith review of ScaleMoGen: Autoregressive Next-Scale Prediction for Human Motion Generation." pith.science (2026). https://pith.science/paper/TIH647K3

@misc{pith2026260511704,
  author       = {Pith},
  title        = {Pith review of: ScaleMoGen: Autoregressive Next-Scale Prediction for Human Motion Generation},
  year         = {2026},
  howpublished = {\url{https://pith.science/paper/TIH647K3}},
  note         = {Machine review of arXiv:2605.11704}
}
Share X Bluesky LinkedIn Reddit HN
read the original abstract

We present ScaleMoGen, a scale-wise autoregressive framework for text-driven human motion generation. Unlike conventional autoregressive approaches that rely on standard next-token prediction, ScaleMoGen frames motion generation as a coarse-to-fine process. We quantize 3D motions into compositional discrete tokens across multiple skeletal-emporal scales of increasing granularity, learning to generate motion by autoregressively predicting next-scale token maps. To maintain structural integrity, our motion tokenizers and quantizers are explicitly designed so that discrete tokens at every scale strictly preserve the skeletal hierarchy. Additionally, we employ bitwise quantization and prediction, which efficiently scale up the tokenizer vocabulary to preserve motion details and stabilize optimization. Extensive experiments demonstrate that ScaleMoGen achieves state-of-the-art performance, establishing an FID of 0.030 (vs. 0.045 for MoMask) on HumanML3D and a CLIP Score of 0.693 (vs. 0.685 for MoMask++) on the SnapMoGen dataset. Furthermore, we demonstrate that our skeletal-temporal multi-scale representation naturally facilitates training-free, text-guided motion editing.

Figures

Figures reproduced from arXiv: 2605.11704 by Bing Zhou, Chuan Guo, Hojun Jang, Inwoo Hwang, Jian Wang, Young Min Kim.

Figure 1
Figure 1. Figure 1: Overview of our skeletal-temporal multi-scale motion quantization pipeline. Given an input motion sequence m, the encoder E maps it to a continuous skeletal￾temporal latent grid f. The latent is decomposed into a hierarchy of residual compo￾nents {q v } V v=0 via binary multi-scale residual quantization, where each scale has its own temporal resolution and skeletal partition. The quantized residuals are th… view at source ↗
Figure 2
Figure 2. Figure 2: (a) Text-to-Motion Generation: Given a prompt cs, we autoregressively predict the next-scale token maps {q v s } V v=0 conditioned on all coarser-scale token maps. (b) Text-Driven Motion Editing: With additional target prompt ct with a source-token preservation mask {Mv } V v=0, we predict edited tokens q ′(v) t conditioned on the re￾maining source motion context and ct. The target token maps {q v t } V v=… view at source ↗
Figure 3
Figure 3. Figure 3: Qualitative text-to-motion generation results of ScaleMoGen. Given highly descriptive, long-form text prompts, ScaleMoGen accurately synthesizes complex se￾quences of actions (top-left), fine-grained body-part articulations (top-right), and timely executed motions with precise spatial constraints (bottom). 4.1 Text-to-Motion Generation Tables 1 and 2 show the results on HumanML3D and SnapMoGen. On Hu￾manML… view at source ↗
Figure 4
Figure 4. Figure 4: Qualitative results of text-driven motion editing. Given a source motion and a new target description, ScaleMoGen accurately synthesizes the desired semantic changes, while preserving the identity and unrelated behaviors of the original source motion. baselines: MDM [35] and SALAD [16]. For each method, we first generate a source motion from the source text using its own generation pipeline, then apply the… view at source ↗
Figure 5
Figure 5. Figure 5: The full-body skeleton is spatially downsampled by merging adjacent joints’ in￾formation into coarser anatomical groups, as indicated by the color-coded regions. The initial full-resolution skeleton accommodates both the 22-joint HumanML3D format and the 24-joint SnapMoGen format. Through successive pooling stages, the model effectively captures skeletal complexity and represents the human body in an atomi… view at source ↗
Figure 6
Figure 6. Figure 6: Visualization of the intermediate accumulated token map motion reconstruc￾tion. Coarse token map captures global motions, which progressively disentangle into independent, fine-grained joint movements at finer scales, restoring full motion realism [PITH_FULL_IMAGE:figures/full_fig_p022_6.png] view at source ↗

Review history (2 revisions) →

discussion (0)

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

Forward citations

Cited by 1 Pith paper

Reviewed papers in the Pith corpus that reference this work. Sorted by Pith novelty score.

  1. DigitCode: Symbolic Tokenization of Hand Motion by Anatomical Units

    cs.RO 2026-08 conditional novelty 7.0

    DigitCode tokenizes hand motion by anatomical units, showing the token span (bone/finger/hand) matters more than the quantizer family, and reduces symbolic reconstruction error by about three quarters.

Reference graph

Works this paper leans on

45 extracted references · 8 linked inside Pith · cited by 1 Pith paper

  1. [1]

    In: SIGGRAPH Asia 2024 Conference Pa- pers (2024) 4

    Athanasiou, N., Ceske, A., Diomataris, M., Black, M.J., Varol, G.: MotionFix: Text-driven 3d human motion editing. In: SIGGRAPH Asia 2024 Conference Pa- pers (2024) 4

    2024

  2. [2]

    In: Proceedings of the IEEE/CVF International Conference on Computer Vision (ICCV)

    Bae, J., Hwang, I., Lee, Y.Y., Guo, Z., Liu, J., Ben-Shabat, Y., Kim, Y.M., Kapa- dia, M.: Less is more: Improving motion diffusion models with sparse keyframes. In: Proceedings of the IEEE/CVF International Conference on Computer Vision (ICCV). pp. 11069–11078 (October 2025) 3

    2025

  3. [3]

    In: Proceedings of the IEEE/CVF Conference on Computer Vision and Pattern Recognition

    Chen, X., Jiang, B., Liu, W., Huang, Z., Fu, B., Chen, T., Yu, G.: Executing your commands via motion diffusion in latent space. In: Proceedings of the IEEE/CVF Conference on Computer Vision and Pattern Recognition. pp. 18000–18010 (2023) 1, 3, 10, 12

    2023

  4. [4]

    In: ACM SIGGRAPH (2025) 2, 3

    Ghosh, A., Zhou, B., Dabral, R., Wang, J., Golyanik, V., Theobalt, C., Slusallek, P., Guo, C.: Duetgen: Music driven two-person dance generation via hierarchical masked modeling. In: ACM SIGGRAPH (2025) 2, 3

    2025

  5. [5]

    In: The Thirty-ninth Annual Conference on Neural Informa- tion Processing Systems (2025),https://openreview.net/forum?id=pdE9onSn2h 2, 3, 10, 11, 12, 18, 20

    Guo, C., Hwang, I., Wang, J., Zhou, B.: Snapmogen: Human motion generation from expressive texts. In: The Thirty-ninth Annual Conference on Neural Informa- tion Processing Systems (2025),https://openreview.net/forum?id=pdE9onSn2h 2, 3, 10, 11, 12, 18, 20

    2025

  6. [6]

    In: Proceedings of the IEEE/CVF Conference on Computer Vision and Pattern Recognition

    Guo, C., Mu, Y., Javed, M.G., Wang, S., Cheng, L.: Momask: Generative masked modeling of 3d human motions. In: Proceedings of the IEEE/CVF Conference on Computer Vision and Pattern Recognition. pp. 1900–1910 (2024) 2, 3, 10, 12

    1900

  7. [7]

    In: Proceedings of the IEEE/CVF Con- ference on Computer Vision and Pattern Recognition

    Guo, C., Zou, S., Zuo, X., Wang, S., Ji, W., Li, X., Cheng, L.: Generating diverse and natural 3d human motions from text. In: Proceedings of the IEEE/CVF Con- ference on Computer Vision and Pattern Recognition. pp. 5152–5161 (2022) 1, 2, 3, 10, 12, 20

    2022

  8. [8]

    In: European Conference on Computer Vision

    Guo, C., Zuo, X., Wang, S., Cheng, L.: Tm2t: Stochastic and tokenized model- ing for the reciprocal generation of 3d human motions and texts. In: European Conference on Computer Vision. pp. 580–597. Springer (2022) 2, 3, 10, 12

    2022

  9. [9]

    In: Proceedings of the 28th ACM International Conference on Multimedia

    Guo, C., Zuo, X., Wang, S., Zou, S., Sun, Q., Deng, A., Gong, M., Cheng, L.: Action2motion: Conditioned generation of 3d human motions. In: Proceedings of the 28th ACM International Conference on Multimedia. pp. 2021–2029 (2020) 3, 10

    2021

  10. [10]

    In: Proceedings of the IEEE/CVF Conference on Computer Vision and Pattern Recognition (CVPR)

    Han, J., Liu, J., Jiang, Y., Yan, B., Zhang, Y., Yuan, Z., Peng, B., Liu, X.: In- finity: Scaling bitwise autoregressive modeling for high-resolution image synthesis. In: Proceedings of the IEEE/CVF Conference on Computer Vision and Pattern Recognition (CVPR). pp. 15733–15744 (June 2025) 4 16 I. Hwang et al

    2025

  11. [11]

    In: IEEE International Conference on Acoustics, Speech and Signal Processing (ICASSP) (2025) 2, 4

    Han, S.H.K., et al.: Bad: Bidirectional auto-regressive diffusion for text-to-motion generation. In: IEEE International Conference on Acoustics, Speech and Signal Processing (ICASSP) (2025) 2, 4

    2025

  12. [12]

    arXiv preprint arXiv:2010.14701 (2020) 13

    Henighan, T., Kaplan, J., Katz, M., Chen, M., Hesse, C., Jackson, J., Jun, H., Brown, T.B., Dhariwal, P., Gray, S., et al.: Scaling laws for autoregressive genera- tive modeling. arXiv preprint arXiv:2010.14701 (2020) 13

    Pith/arXiv arXiv 2010

  13. [13]

    In: European Conference on Computer Vision (ECCV) (2024) 8

    Heo, B., Park, S., Han, D., Yun, S.: Rotary position embedding for vision trans- former. In: European Conference on Computer Vision (ECCV) (2024) 8

    2024

  14. [14]

    Hertz, A., Mokady, R., Tenenbaum, J., Aberman, K., Pritch, Y., Cohen-Or, D.: Prompt-to-prompt image editing with cross attention control (2022) 4

    2022

  15. [15]

    Ho, J., Salimans, T.: Classifier-free diffusion guidance (2022),https://arxiv.org/ abs/2207.1259820

    Pith/arXiv arXiv 2022

  16. [16]

    In: Proceedings of the Computer Vision and Pattern Recognition Conference (CVPR)

    Hong, S., Kim, C., Yoon, S., Nam, J., Cha, S., Noh, J.: Salad: Skeleton-aware latent diffusion for text-driven motion generation and editing. In: Proceedings of the Computer Vision and Pattern Recognition Conference (CVPR). pp. 7158–7168 (June 2025) 3, 4, 10, 11, 12, 13, 14, 18

    2025

  17. [17]

    In: Proceedings of the 32nd ACM International Conference on Multimedia

    Huang, Y., Yang, H., Luo, C., Wang, Y., Xu, S., Zhang, Z., Zhang, M., Peng, J.: Stablemofusion: Towards robust and efficient diffusion-based motion genera- tion framework. In: Proceedings of the 32nd ACM International Conference on Multimedia. pp. 224–232 (2024) 10, 12

    2024

  18. [18]

    In: Proceedings of the Computer Vision and Pattern Recognition Conference (CVPR) Workshops

    Hwang, I., Bae, J., Lim, D., Kim, Y.M.: Goal-driven human motion synthesis in diverse task. In: Proceedings of the Computer Vision and Pattern Recognition Conference (CVPR) Workshops. pp. 2920–2930 (June 2025) 3

    2025

  19. [19]

    In: Proceedings of the IEEE/CVF International Conference on Computer Vision (ICCV)

    Hwang, I., Bae, J., Lim, D., Kim, Y.M.: Motion synthesis with sparse and flexible keyjoint control. In: Proceedings of the IEEE/CVF International Conference on Computer Vision (ICCV). pp. 13203–13213 (October 2025) 3

    2025

  20. [20]

    In: Proceedings of the IEEE/CVF Interna- tional Conference on Computer Vision (ICCV)

    Hwang, I., Zhou, B., Kim, Y.M., Wang, J., Guo, C.: Scenemi: Motion in-betweening for modeling human-scene interaction. In: Proceedings of the IEEE/CVF Interna- tional Conference on Computer Vision (ICCV). pp. 6034–6045 (October 2025) 3

    2025

  21. [21]

    Advances in Neural Information Processing Systems36, 20067–20079 (2023) 2, 3

    Jiang, B., Chen, X., Liu, W., Yu, J., Yu, G., Chen, T.: Motiongpt: Human motion as a foreign language. Advances in Neural Information Processing Systems36, 20067–20079 (2023) 2, 3

    2023

  22. [22]

    In: Proceedings of the AAAI Conference on Artificial Intelligence

    Kim, J., Kim, J., Choi, S.: Flame: Free-form language-based motion synthesis & editing. In: Proceedings of the AAAI Conference on Artificial Intelligence. vol. 37, pp. 8255–8263 (2023) 4

    2023

  23. [23]

    ACM Transactions on Graphics (TOG)41(4), 1–12 (2022) 2

    Li, P., Aberman, K., Zhang, Z., Hanocka, R., Sorkine-Hornung, O.: Ganimator: Neural motion synthesis from a single sequence. ACM Transactions on Graphics (TOG)41(4), 1–12 (2022) 2

    2022

  24. [24]

    Li, Z., Cheng, K., Ghosh, A., Bhattacharya, U., Gui, L., Bera, A.: Simmotionedit: Text-based human motion editing with motion similarity prediction (2025) 4

    2025

  25. [25]

    In: Proceedings of the Computer Vision and Pattern Recognition Conference

    Lu, S., Wang, J., Lu, Z., Chen, L.H., Dai, W., Dong, J., Dou, Z., Dai, B., Zhang, R.: Scamo: Exploring the scaling law in autoregressive motion generation model. In: Proceedings of the Computer Vision and Pattern Recognition Conference. pp. 27872–27882 (2025) 1, 13

    2025

  26. [26]

    In: Proceedings of the IEEE/CVF international conference on computer vision

    Mahmood, N., Ghorbani, N., Troje, N.F., Pons-Moll, G., Black, M.J.: Amass: Archive of motion capture as surface shapes. In: Proceedings of the IEEE/CVF international conference on computer vision. pp. 5442–5451 (2019) 10

    2019

  27. [27]

    In: International Conference on Learning Representations (2022) 4 ScaleMoGen 17

    Meng, C., He, Y., Song, Y., Song, J., Wu, J., Zhu, J.Y., Ermon, S.: SDEdit: Guided image synthesis and editing with stochastic differential equations. In: International Conference on Learning Representations (2022) 4 ScaleMoGen 17

    2022

  28. [28]

    arXiv preprint arXiv:2411.16575 (2024) 1, 3, 10, 12

    Meng, Z., Xie, Y., Peng, X., Han, Z., Jiang, H.: Rethinking diffusion for text-driven human motion generation. arXiv preprint arXiv:2411.16575 (2024) 1, 3, 10, 12

    Pith/arXiv arXiv 2024

  29. [29]

    In: European Conference on Computer Vision

    Petrovich, M., Black, M.J., Varol, G.: Temos: Generating diverse human motions from textual descriptions. In: European Conference on Computer Vision. pp. 480–

  30. [30]

    In: Proceedings of the IEEE/CVF Interna- tional Conference on Computer Vision

    Petrovich, M., Black, M.J., Varol, G.: Tmr: Text-to-motion retrieval using con- trastive 3d human motion synthesis. In: Proceedings of the IEEE/CVF Interna- tional Conference on Computer Vision. pp. 9488–9497 (2023) 10

    2023

  31. [31]

    In: Computer Vision – ECCV 2024 (2024) 2, 4

    Pinyoanuntapong, E., Saleem, M.U., Wang, P., Lee, M., Das, S., Chen, C.: Bamm: Bidirectional autoregressive motion model. In: Computer Vision – ECCV 2024 (2024) 2, 4

    2024

  32. [32]

    In: Proceedings of the IEEE/CVF Conference on Computer Vision and Pattern Recognition

    Pinyoanuntapong, E., Wang, P., Lee, M., Chen, C.: Mmm: Generative masked motion model. In: Proceedings of the IEEE/CVF Conference on Computer Vision and Pattern Recognition. pp. 1546–1555 (2024) 2, 3, 10, 12

    2024

  33. [33]

    Journal of Machine Learning Research21(140), 1–67 (2020), http://jmlr.org/papers/v21/20-074.html7

    Raffel, C., Shazeer, N., Roberts, A., Lee, K., Narang, S., Matena, M., Zhou, Y., Li, W., Liu, P.J.: Exploring the limits of transfer learning with a unified text- to-text transformer. Journal of Machine Learning Research21(140), 1–67 (2020), http://jmlr.org/papers/v21/20-074.html7

    2020

  34. [34]

    Sui, K., Ghosh, A., Hwang, I., Wang, J., Guo, C.: A survey on human interaction motion generation (2025),https://arxiv.org/abs/2503.127631

    arXiv 2025

  35. [35]

    arXiv preprint arXiv:2209.14916 (2022) 1, 3, 10, 12, 13, 14, 18

    Tevet, G., Raab, S., Gordon, B., Shafir, Y., Cohen-Or, D., Bermano, A.H.: Human motion diffusion model. arXiv preprint arXiv:2209.14916 (2022) 1, 3, 10, 12, 13, 14, 18

    Pith/arXiv arXiv 2022

  36. [36]

    In: The Thirty-eighth Annual Conference on Neural Information Processing Systems (2024),https: //openreview.net/forum?id=gojL67CfS84, 20

    Tian, K., Jiang, Y., Yuan, Z., PENG, B., Wang, L.: Visual autoregressive model- ing: Scalable image generation via next-scale prediction. In: The Thirty-eighth Annual Conference on Neural Information Processing Systems (2024),https: //openreview.net/forum?id=gojL67CfS84, 20

    2024

  37. [37]

    Advances in neural information processing systems30(2017) 1, 3

    Van Den Oord, A., Vinyals, O., et al.: Neural discrete representation learning. Advances in neural information processing systems30(2017) 1, 3

    2017

  38. [38]

    In: Proceedings of the IEEE/CVF International Conference on Computer Vision (ICCV)

    Wang, Y., Guo, L., Li, Z., Huang, J., Wang, P., Wen, B., Wang, J.: Training- free text-guided image editing with visual autoregressive model. In: Proceedings of the IEEE/CVF International Conference on Computer Vision (ICCV). pp. 17577– 17586 (October 2025) 4, 10

    2025

  39. [39]

    Neural In- formation Processing Systems (NeurIPS) (2024) 2, 3, 10, 12, 13, 18

    Yuan, W., Shen, W., HE, Y., Dong, Y., Gu, X., Dong, Z., Bo, L., Huang, Q.: Mogents: Motion generation based on spatial-temporal joint modeling. Neural In- formation Processing Systems (NeurIPS) (2024) 2, 3, 10, 12, 13, 18

    2024

  40. [40]

    arXiv preprint arXiv:2301.06052 (2023) 2, 3, 10, 12, 18

    Zhang, J., Zhang, Y., Cun, X., Huang, S., Zhang, Y., Zhao, H., Lu, H., Shen, X.: T2m-gpt: Generating human motion from textual descriptions with discrete representations. arXiv preprint arXiv:2301.06052 (2023) 2, 3, 10, 12, 18

    Pith/arXiv arXiv 2023

  41. [41]

    arXiv preprint arXiv:2208.15001 (2022) 1, 3, 10, 12

    Zhang, M., Cai, Z., Pan, L., Hong, F., Guo, X., Yang, L., Liu, Z.: Motiondif- fuse: Text-driven human motion generation with diffusion model. arXiv preprint arXiv:2208.15001 (2022) 1, 3, 10, 12

    Pith/arXiv arXiv 2022

  42. [42]

    Zhang, M., Guo, X., Pan, L., Cai, Z., Hong, F., Li, H., Yang, L., Liu, Z.: Remodif- fuse:Retrieval-augmentedmotiondiffusionmodel.arXivpreprintarXiv:2304.01116 (2023) 1, 3

    Pith/arXiv arXiv 2023

  43. [43]

    Advances in Neural Information Processing Systems36, 13981–13992 (2023) 2, 4

    Zhang, M., Li, H., Cai, Z., Ren, J., Yang, L., Liu, Z.: Finemogen: Fine-grained spatio-temporal motion generation and editing. Advances in Neural Information Processing Systems36, 13981–13992 (2023) 2, 4

    2023

  44. [44]

    inner-to- outer

    Zhao, Y., Xiong, Y., Krähenbühl, P.: Image and video tokenization with binary spherical quantization. arXiv preprint arXiv:2406.07548 (2024) 2, 6 18 I. Hwang et al. In the supplementary materials, we evaluate sampling efficiency (Section A), and provide implementation details on the model architecture and algorithm (Section B). We further analyze the prop...

    Pith/arXiv arXiv 2024

  45. [497]

    Springer (2022) 1, 3

    2022

This paper was first reviewed by deepseek-v4-flash on August 2, 2026.