Pith. sign in

REVIEW 5 major objections 8 minor 70 references

PMG: Progressive Motion Generation via Sparse Anchor Postures Curriculum Learning

T0 review · 5 major / 8 minor · reviewed 2026-08-16 · deepseek-v4-flash

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

Pith's one-line read ProMoGen claims that one diffusion model can obey both a global trajectory and a handful of sparse anchor postures, and that this dual control beats single-condition motion generators.

desk verdict Plausible engineering contribution with a clean curriculum idea, but the SOTA claim rests on uncheckable baseline adaptations and a missing appendix. read the letter →

arxiv 2504.16722 v1 pith:VU4F2ZC4 submitted 2025-04-23 cs.CV cs.AI

Yingjie Xi , Jian Jun Zhang , Xiaosong Yang This is my paper · ORCID
classification cs.CVcs.AI
keywords humanmotionsynthesistrajectoryconditioningsparsekeyframecontroldiffusionmodelscurriculumlearningin-betweeningcontrollablegenerationanimation
verification ladder T0 review T1 audit T2 compute T3 formal

The pith

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

The reading

The paper argues that the most useful way to control generated human motion is to combine a global trajectory, which sets where the character goes, with sparse anchor postures, which set what the body looks like at chosen moments. It claims these two signals can be cleanly decoupled, and that a single diffusion model can be trained on both at once. To make training stable when anchors are very sparse, it introduces a curriculum that starts with many anchor frames and gradually reduces their number. On HumanML3D and CombatMotion, the system is reported to beat existing text-, trajectory-, and keyframe-based methods on pose error, fidelity, and anchor faithfulness. If correct, the contribution is a unified control interface: a user specifies a path and a few poses and gets smooth, customized motion.

What carries the argument

The load-bearing mechanism is the Sparse Anchor Posture Curriculum Learning (SAP-CL) schedule paired with a two-stream diffusion-transformer architecture. SAP-CL splits training into stages with anchor-count interval $[K^{(s)}_{\min}, K_{\max}]$, where $K_{\max}=30$; the minimum starts at 20 and linearly decays to 1, and each stage samples temporal density $f_n^k$ and interval elasticity $f_s^k$ uniformly. This progressive move from dense to sparse supervision is what avoids unstable gradients from extremely sparse signals and lets the model generalize to arbitrary anchor placements at inference. The Filtering Module implements the placement by selecting virtual points, distributing elastic interval increments $\delta_i$, and mapping them through $x_j = j+(j-1)f_s + \sum_{i=0}^{j-1}\delta_i$, which guarantees the minimum spacing $x_k-x_{k-1}\ge f_s+1$ and makes user-specified anchor positions reproducible during training.

What would settle it

Run the same trajectory-and-anchor test set through a strong text-conditioned motion model using text prompts that describe the actions at the anchor poses, and compare its MPJPE and FID with ProMoGen's; if it matches or beats them, the claimed superiority of sparse-anchor control is not established.

Watch

Extended reading notes

Core claim

The central claim is that motion synthesis is best driven by two decoupled conditions: a global trajectory $\tau\in\mathbb{R}^{N\times 3}$ and sparse anchor postures $X_s\in\mathbb{R}^{M\times 6}$, and that this pairing yields more faithful motion than any single-condition baseline. The network assigns each condition its own transformer encoder, builds a coarse motion from the trajectory alone, then refines it by fusing both feature streams. A Filtering Module samples anchor frames under user-given density and minimum-interval constraints, while SAP-CL schedules the anchor count from 20 down to 1 during training. Reported numbers include MPJPE 3.257 on HumanML3D and 3.654 on CombatMotion, with FID 0.279 and 0.412, each best among the listed baselines; the curriculum-trained variant further lowers MPJPE to 3.047 at five anchors. The paper's claim is that sparse anchors plus trajectory is not merely a harder in-betweening problem but a controllable generation task, and that the easy-to-hard curriculum is what makes learning it stable.

Load-bearing premise

The claim that ProMoGen beats state-of-the-art methods rests on the assumption that replacing the text encoders of existing models with two linear layers leaves those baselines unimpaired; if that swap handicaps them, the comparison is not meaningful.

Editorial extensions

If this is right

  • A single unified model can serve both trajectory-only and trajectory-plus-anchor control, since dual and single control share one training process.
  • The reported anchor-position error K-MPJPE is below the global MPJPE on both datasets, meaning the model is especially faithful at the poses the user actually specified.
  • Using only a linear encoder for both conditions already beats all listed baselines, implying much of the gain comes from the decoupled conditioning design rather than the elaborate modules.
  • The curriculum evaluation shows that at the same inference-time anchor count, a model trained with the progressive schedule performs better, shifting the accuracy-versus-sparsity tradeoff in the user's favor.

Reading between the lines

Editorial extensions of the paper, not claims the author makes directly.

  • The decoupling assumption is testable beyond the paper: if trajectory and local pose are truly independent, adapting the model to a new skeleton or motion style should require retraining only the anchor encoder, not the whole network.
  • The same framework could ingest anchor postures detected from images or video, since the Filtering Module already imitates arbitrary user placement by uniform sampling over valid anchor sets.
  • A fairer head-to-head would also run text-conditioned baselines prompted with descriptions of the anchor actions, because the paper's comparison strips text encoders from prior models and replaces them with linear layers.
Share X Bluesky LinkedIn Reddit HN

Signed reviews

No signed human review yet.

Editorial analysis

A structured set of objections, weighed in public.

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

Referee Report

5 major / 8 minor

Summary. The paper proposes ProMoGen, a diffusion-transformer-based framework for human motion generation conditioned jointly on a global trajectory and sparse anchor postures, together with SAP-CL, a curriculum learning strategy that progressively reduces the number of anchor poses during training. The authors report large gains over existing methods on HumanML3D and CombatMotion across MPJPE, K-MPJPE, FID, and related metrics, and they provide ablations of the main architectural components and of the curriculum strategy. The central claim is that a single model can combine trajectory control with user-specified sparse pose constraints and outperform prior trajectory-, keyframe-, and text-conditioned motion generators.

Significance. If the reported results are reproducible, ProMoGen would be a practically useful contribution: it addresses a real gap by allowing simultaneous trajectory and sparse-pose control, and the SAP-CL idea is a sensible response to the instability of training from very sparse anchors. The paper is also honest about its task formulation and evaluates on two datasets. However, the significance is currently limited by verification gaps: the state-of-the-art comparison rests on baselines that the authors re-implemented with text encoders replaced by two linear layers, and the paper does not supply code, official checkpoints, or a complete appendix. The quantitative story is therefore plausible but not yet established to the standard the abstract claims.

major comments (5)
  1. [§4 (Implement Details)] The headline claim of outperforming state-of-the-art methods is not established by Table 1 because all baselines are "reconstructed" by replacing their text encoders with two linear layers, and the reconstruction protocol is underspecified. The paper does not report the baselines' training losses, anchor-sampling distributions, curriculum or regularization hyperparameters, or whether the same Filtering Module and anchor set were used for every method. If the linear adapters are a bottleneck for methods whose architectures assume text conditioning, the comparison is biased toward ProMoGen. The description of Ours-v1 is also ambiguous: if it is ProMoGen with linear encoders rather than an adapted baseline, then Table 1 does not demonstrate that the adapted baselines were given their best possible configuration. Please clarify the experimental protocol, provide per-baseline settings, or re-run the comparison against official checkpoints/code.
  2. [§3.2.2] The description of the Initial Motion Generator is internally inconsistent. The text states that G "synthesizes coarse motion priors conditioned solely on trajectory features," but the equation defining it is M_init = G(x_t, E_k(X_s)), which uses the anchor motion encoder and gives no trajectory input. Please correct either the text or the equation; this ambiguity makes the architecture non-reproducible.
  3. [§3.2.1] The anchor-mapping equations contain indexing errors. The constraint is written as sum_{i=0}^{f_n} δ_i = R, but only f_n deltas (δ_0 to δ_{f_n-1}) are defined, and their actual sum is T_total - f_n = R. Moreover, the recurrence x_j = x_{j-1} + f_s + δ_{j-1} + 1 implies x_j = p_j + (j-1) f_s, so adjacent anchors are separated by f_s + δ_{j-1} + 1 ≥ f_s + 1; the text says f_s is the "minimum interval," which is off by one. Please verify the formulas and restate them precisely.
  4. [§5 (Conclusion) and Experiments] The conclusion promises that "More details and results are shown in the Appendix," but no appendix is present in the submitted manuscript. Important experimental details—dataset splits, baseline reconstruction settings, evaluation protocol, and the exact relation between "Regular" and "Curriculum" training in Tables 2 and 3—are therefore missing. The GitHub link is given without a commit hash or released code, so the experiments cannot be checked.
  5. [Table 3] The ablation results are not fully consistent with the narrative. Removing the physical-constraint loss gives MPJPE 3.251 and K-MPJPE 3.221, which are slightly better than the full model's 3.257 and 3.247, while the text states that "GAN and Phys Loss can further improve the precision." Also, Table 3's "w/. Reg Learning" and "w/. Cur Learning" rows do not match the corresponding rows in Table 2, so the reader cannot tell which anchor count and training schedule these numbers refer to. Please clarify these inconsistencies.
minor comments (8)
  1. [Abstract] The abstract contains a comma splice ("inherently unstable, we introduce"); please rewrite the sentence.
  2. [References / Datasets] Reference [56] is a GitHub repository rather than a peer-reviewed dataset; please provide a stable citation and describe the train/test split, retargeting procedure, and preprocessing for CombatMotion.
  3. [Table 1] The table reports "Diversity→ closer to the value of Real is better," but the real reference values are not reported; please add them.
  4. [Table 3] Several entries in Table 3 are missing the '±' separator (e.g., "3.5310.066" and "3.3260.054").
  5. [Notation] The paper uses "Dit" and "DiT" inconsistently; please choose one spelling and use it throughout.
  6. [Diffusion solver] DPM-Solver++ is mentioned with an update rule but is neither cited nor defined; please add the appropriate reference and explain how the coefficients are obtained.
  7. [Statistics] No statistical significance tests are reported; given the small standard deviations in Table 1, please provide p-values or confidence intervals for the key comparisons.
  8. [General] There are several typos, including "anchor posetures," "On the on hand," and "oberseved"; please proofread the manuscript.

Circularity Check

0 steps flagged · score 0.0 of 10

No significant circularity: the core claim is an empirical benchmark comparison against external datasets and published baselines, and no derivation step reduces to its own inputs by construction.

full rationale

The paper's derivation chain is a standard diffusion-model pipeline: a DDPM forward/reverse process, a DiT-style architecture (TFE, AME, IMG, refinement, decoder), a Filtering Module for anchor sampling, a composite loss, and SAP-CL curriculum training. None of these steps defines a predicted quantity in terms of the input conditions in a way that makes the reported result true by construction. The headline claim is an empirical SOTA comparison in Table 1 against published baselines on HumanML3D and CombatMotion, with metrics computed from held-out test motion. The K-MPJPE metric is related to the Anchor Motion Loss term, since both measure error at anchor positions, but K-MPJPE is evaluated on test samples after training, so reporting it is ordinary evaluation of a trained objective rather than a fitted input renamed as a prediction. The adaptation of text-conditioned baselines by replacing text encoders with two linear layers is a potential fairness and reproducibility concern, not a circularity concern; the same is true of the missing appendix and unversioned code link advertised in the paper. There is no load-bearing self-citation chain, no imported uniqueness theorem, and no ansatz smuggled in via a same-author citation. The verifiability of the SOTA claim depends on protocol release and baseline fairness, but the derivation itself is not circular.

Assumptions & free parameters 5 free parameters · 4 assumptions · 0 invented entities

The central empirical claim rests on standard diffusion math, on a modeling assumption that trajectory and sparse postures occupy separable subspaces, and on an evaluation assumption that author-adapted baselines are fair. The main numbers also depend on hand-selected hyperparameters such as the anchor-count cap, the curriculum schedule, and the loss weights. No new physical entities are introduced.

free parameters (5)
  • Kmax (maximum anchor count) = 30
    Section 3.3 sets Kmax=30 because the authors posit that more anchors add redundant information; this cap shapes the curriculum distribution and is not derived from data.
  • SAP-CL stage schedule = Kmin from 20 to 1 over E_stage=4 stages; E_total=100 epochs
    Algorithm 1 and Section 4 choose the stage count and anchor-count interval by hand; these numbers determine how the model is trained and are not justified by a selection procedure.
  • Loss weights = lambda1=lambda2=lambda3=1.0, lambda4=lambda5=0.1
    Section 3.2.3 fixes the weights by hand; the balance between reconstruction, anchor, joint, GAN, and physics losses affects all reported metrics.
  • Diffusion sampling timesteps = 1000 training, 25 inference
    Section 4 sets these values; the 25-step DPM-Solver++ schedule is a practical choice that influences fidelity and is not defended by an experiment.
  • Filtering Module sampling ranges = fn sampled from U(Kmin, Kmax); fs sampled from U(4, floor(N/fn))
    Section 3.3 and Algorithm 1 define the anchor sparsity distribution by hand-set ranges; this distribution is part of the curriculum design and affects the training data seen by the model.
assumptions (4)
  • standard math DDPM forward/reverse process and DPM-Solver++ solver are valid for motion prediction
    Section 3.1 invokes the diffusion literature and DPM-Solver++ without re-derivation; this is accepted background in diffusion-based generation.
  • domain assumption Trajectory and sparse anchor postures can be decoupled into global displacement and local pose subspaces
    Section 1 and Section 3.2.2 assume this decomposition; if the two controls interact strongly, the separate encoders and the trajectory-only initial generator may not capture the motion.
  • domain assumption Dense-to-sparse anchor curriculum improves convergence relative to direct sparse training
    Section 3.3 asserts that sparse training is unstable due to insufficient supervision; Table 2 supports it empirically on one dataset, but it is a training-method belief, not a theorem.
  • ad hoc to paper Replacing text encoders of baselines with two linear layers yields a fair comparison
    Section 4 'Implement Details' makes this choice to enable trajectory plus anchor conditioning; it assumes the adaptation does not handicap prior methods.

how reviews work

0 comments
Cite this review

Pith. "Pith review of PMG: Progressive Motion Generation via Sparse Anchor Postures Curriculum Learning." pith.science (2026). https://pith.science/paper/VU4F2ZC4

@misc{pith2026250416722,
  author       = {Pith},
  title        = {Pith review of: PMG: Progressive Motion Generation via Sparse Anchor Postures Curriculum Learning},
  year         = {2026},
  howpublished = {\url{https://pith.science/paper/VU4F2ZC4}},
  note         = {Machine review of arXiv:2504.16722}
}
read the original abstract

In computer animation, game design, and human-computer interaction, synthesizing human motion that aligns with user intent remains a significant challenge. Existing methods have notable limitations: textual approaches offer high-level semantic guidance but struggle to describe complex actions accurately; trajectory-based techniques provide intuitive global motion direction yet often fall short in generating precise or customized character movements; and anchor poses-guided methods are typically confined to synthesize only simple motion patterns. To generate more controllable and precise human motions, we propose \textbf{ProMoGen (Progressive Motion Generation)}, a novel framework that integrates trajectory guidance with sparse anchor motion control. Global trajectories ensure consistency in spatial direction and displacement, while sparse anchor motions only deliver precise action guidance without displacement. This decoupling enables independent refinement of both aspects, resulting in a more controllable, high-fidelity, and sophisticated motion synthesis. ProMoGen supports both dual and single control paradigms within a unified training process. Moreover, we recognize that direct learning from sparse motions is inherently unstable, we introduce \textbf{SAP-CL (Sparse Anchor Posture Curriculum Learning)}, a curriculum learning strategy that progressively adjusts the number of anchors used for guidance, thereby enabling more precise and stable convergence. Extensive experiments demonstrate that ProMoGen excels in synthesizing vivid and diverse motions guided by predefined trajectory and arbitrary anchor frames. Our approach seamlessly integrates personalized motion with structured guidance, significantly outperforming state-of-the-art methods across multiple control scenarios.

Figures

Figures reproduced from arXiv: 2504.16722 by the authors.

Figure 1
Figure 1. The left panel illustrates the synthesized results of the ProMoGen framework, under the control of both predefined [PITH_FULL_IMAGE:figures/full_fig_p001_1.png] view at source ↗
Figure 2
Figure 2. The SAP-CL training strategy(left panel) and the structure of ProMoGen method(right panel). [PITH_FULL_IMAGE:figures/full_fig_p002_2.png] view at source ↗
Figure 3
Figure 3. In this figure, the horizontal axis represents FID, [PITH_FULL_IMAGE:figures/full_fig_p003_3.png] view at source ↗
Figures from the paper (4 more)
Figure 4
Figure 4. Figure 4: This figure illustrates the trajectory-based, sparse pose-guided motion generation process of our ProMoGen. In the [PITH_FULL_IMAGE:figures/full_fig_p005_4.png]
Figure 5
Figure 5. Figure 5: The figure illustrates the comparative generation effects of various modules. Under identical trajectory and anchor [PITH_FULL_IMAGE:figures/full_fig_p006_5.png]
Figure 6
Figure 6. Figure 6: The two figures above display the changing of FID [PITH_FULL_IMAGE:figures/full_fig_p006_6.png]
Figure 7
Figure 7. Figure 7: The figure presents a series of ablation experiments [PITH_FULL_IMAGE:figures/full_fig_p006_7.png]

Discussion (0). Continue with ORCID to comment.

Reference graph

Works this paper leans on

70 extracted references · 23 canonical work pages

  1. [1]

    Nikos Athanasiou, Alpár Cseke, Markos Diomataris, Michael J Black, and Gül Varol. 2024. MotionFix: Text-driven 3d human motion editing. InSIGGRAPH Asia 2024 Conference Papers. 1–11

    work page 2024

  2. [2]

    Nikos Athanasiou, Mathis Petrovich, Michael J Black, and Gül Varol. 2023. SINC: Spatial composition of 3D human motions for simultaneous action generation. In Proceedings of the IEEE/CVF International Conference on Computer Vision . 9984– 9995

    work page 2023

  3. [3]

    Yoshua Bengio, Jérôme Louradour, Ronan Collobert, and Jason Weston. 2009. Curriculum learning. In Proceedings of the 26th annual international conference on machine learning. 41–48

    2009

  4. [4]

    Ling-Hao Chen, Shunlin Lu, Wenxun Dai, Zhiyang Dou, Xuan Ju, Jingbo Wang, Taku Komura, and Lei Zhang. 2024. Pay Attention and Move Better: Harnessing Attention for Interactive Motion Generation and Training-free Editing. arXiv preprint arXiv:2410.18977 (2024)

    arXiv 2024

  5. [5]

    Rui Chen, Mingyi Shi, Shaoli Huang, Ping Tan, Taku Komura, and Xuelin Chen

  6. [6]

    Kiran Chhatre, Nikos Athanasiou, Giorgio Becherini, Christopher Peters, Michael J Black, Timo Bolkart, et al . 2024. Emotional speech-driven 3d body animation via disentangled latent diffusion. In Proceedings of the IEEE/CVF Con- ference on Computer Vision and Pattern Recognition . 1942–1953

    2024

  7. [7]

    Seunggeun Chi, Hyung-gun Chi, Hengbo Ma, Nakul Agarwal, Faizan Siddiqui, Karthik Ramani, and Kwonjoon Lee. 2024. M2d2m: Multi-motion generation from text with discrete diffusion models. In European Conference on Computer Vision. Springer, 18–36

    work page 2024

  8. [8]

    Yuchen Chu and Zeshi Yang. 2024. Real-time Diverse Motion In-betweening with Space-time Control. In Proceedings of the 17th ACM SIGGRAPH Conference on Motion, Interaction, and Games . 1–8

    work page 2024

Show all 70 references
  1. [9]

    Setareh Cohan, Guy Tevet, Daniele Reda, Xue Bin Peng, and Michiel van de Panne

  2. [10]

    Peishan Cong, Ziyi Wang, Zhiyang Dou, Yiming Ren, Wei Yin, Kai Cheng, Yujing Sun, Xiaoxiao Long, Xinge Zhu, and Yuexin Ma. 2024. Laserhuman: language- guided scene-aware human motion generation in free environment.arXiv preprint arXiv:2403.13307 (2024)

    2024 arXiv

  3. [11]

    In ACM SIGGRAPH 2024 Conference Papers

    Flexible motion in-betweening with diffusion models. In ACM SIGGRAPH 2024 Conference Papers. 1–9

    2024

  4. [12]

    Minyue Dai, Jingbo Wang, Ke Fan, Bin Ji, Haoyu Zhao, Junting Dong, and Bo Dai. 2025. Towards Synthesized and Editable Motion In-Betweening Through Part-Wise Phase Representation. arXiv:2503.08180 [cs.CV] https://arxiv.org/abs/ 2503.08180

    2025 arXiv

  5. [13]

    Rishabh Dabral, Muhammad Hamza Mughal, Vladislav Golyanik, and Christian Theobalt. 2023. Mofusion: A framework for denoising-diffusion-based motion synthesis. In Proceedings of the IEEE/CVF conference on computer vision and pattern recognition. 9760–9770

    2023

  6. [14]

    Canxuan Gang and Yiran Wang. 2025. Human Motion Prediction, Reconstruction, and Generation. arXiv preprint arXiv:2502.15956 (2025)

    2025 arXiv

  7. [15]

    Wenxun Dai, Ling-Hao Chen, Jingbo Wang, Jinpeng Liu, Bo Dai, and Yansong Tang. 2024. Motionlcm: Real-time controllable motion generation via latent consistency model. In European Conference on Computer Vision . Springer, 390– 408

    2024

  8. [16]

    Chuan Guo, Shihao Zou, Xinxin Zuo, Sen Wang, Wei Ji, Xingyu Li, and Li Cheng

  9. [17]

    Chuan Guo, Yuxuan Mu, Muhammad Gohar Javed, Sen Wang, and Li Cheng. 2024. Momask: Generative masked modeling of 3d human motions. In Proceedings of the IEEE/CVF Conference on Computer Vision and Pattern Recognition . 1900–1910

    2024

  10. [18]

    Martin Heusel, Hubert Ramsauer, Thomas Unterthiner, Bernhard Nessler, and Sepp Hochreiter. 2017. Gans trained by a two time-scale update rule converge to a local nash equilibrium. Advances in neural information processing systems 30 (2017)

    2017

  11. [19]

    Jonathan Ho, Ajay Jain, and Pieter Abbeel. 2020. Denoising diffusion probabilistic models. Advances in neural information processing systems 33 (2020), 6840–6851

    2020

  12. [20]

    Ziyan Guo, Zeyu Hu, Na Zhao, and De Wen Soh. 2025. MotionLab: Unified Human Motion Generation and Editing via the Motion-Condition-Motion Paradigm. arXiv preprint arXiv:2502.02358 (2025)

    2025 arXiv

  13. [21]

    Yiheng Huang, Hui Yang, Chuanchen Luo, Yuxi Wang, Shibiao Xu, Zhaoxiang Zhang, Man Zhang, and Junran Peng. 2024. Stablemofusion: Towards robust and efficient diffusion-based motion generation framework. In Proceedings of the 32nd ACM International Conference on Multimedia . 224–232

    2024

  14. [22]

    Zhongyu Jiang, Wenhao Chai, Zhuoran Zhou, Cheng-Yen Yang, Hsiang-Wei Huang, and Jenq-Neng Hwang. 2025. PackDiT: Joint Human Motion and Text Generation via Mutual Prompting. arXiv preprint arXiv:2501.16551 (2025)

    2025 arXiv

  15. [23]

    Jonathan Ho, Tim Salimans, Alexey Gritsenko, William Chan, Mohammad Norouzi, and David J Fleet. 2022. Video diffusion models. Advances in Neu- ral Information Processing Systems 35 (2022), 8633–8646

    2022

  16. [24]

    Kacper Kania, Marek Kowalski, and Tomasz Trzciński. 2021. Trajevae: Control- lable human motion generation from trajectories. arXiv preprint arXiv:2104.00351 (2021)

    2021 arXiv

  17. [25]

    Korrawe Karunratanakul, Konpat Preechakul, Supasorn Suwajanakorn, and Siyu Tang. 2023. Guided motion diffusion for controllable human motion synthesis. In Proceedings of the IEEE/CVF International Conference on Computer Vision . 2151– 2162

    2023

  18. [26]

    Gaurav Kamath, Sebastian Schuster, Sowmya Vajjala, and Siva Reddy. 2024. Scope ambiguities in large language models. Transactions of the Association for Compu- tational Linguistics 12 (2024), 738–754

    2024

  19. [27]

    Hyuhng Joon Kim, Youna Kim, Cheonbok Park, Junyeob Kim, Choonghyun Park, Kang Min Yoo, Sang-goo Lee, and Taeuk Kim. 2024. Aligning language models to explicitly handle ambiguity. arXiv preprint arXiv:2404.11972 (2024)

    2024 arXiv

  20. [28]

    Jihoon Kim, Taehyun Byun, Seungyoun Shin, Jungdam Won, and Sungjoon Choi

  21. [29]

    Aryan Keluskar, Amrita Bhattacharjee, and Huan Liu. 2024. Do LLMs Understand Ambiguity in Text? A Case Study in Open-world Question Answering. In 2024 IEEE International Conference on Big Data (BigData) . IEEE, 7485–7490

    2024

  22. [30]

    So Young Lee, Russell Scheinberg, Amber Shore, and Ameeta Agrawal. 2025. Mul- tilingual Relative Clause Attachment Ambiguity Resolution in Large Language Models. arXiv preprint arXiv:2503.02971 (2025)

    2025 arXiv

  23. [31]

    Fenglin Liu, Shen Ge, Yuexian Zou, and Xian Wu. 2022. Competence-based multimodal curriculum learning for medical report generation. arXiv preprint arXiv:2206.14579 (2022)

    2022 arXiv

  24. [32]

    Pattern Recognition 132 (2022), 108894

    Conditional motion in-betweening. Pattern Recognition 132 (2022), 108894

    2022

  25. [33]

    Jisu Kim and Juhwan Lee. 2024. Strategic Data Ordering: Enhancing Large Language Model Performance through Curriculum Learning. arXiv preprint arXiv:2405.07490 (2024)

    2024 arXiv

  26. [34]

    Xin Ma, Yaohui Wang, Gengyun Jia, Xinyuan Chen, Ziwei Liu, Yuan-Fang Li, Cunjian Chen, and Yu Qiao. 2024. Latte: Latent diffusion transformer for video generation. arXiv preprint arXiv:2401.03048 (2024)

    2024 arXiv

  27. [35]

    Binyamin Manela and Armin Biess. 2022. Curriculum learning with hindsight experience replay for sequential object manipulation tasks. Neural Networks 145 (2022), 260–270

    2022

  28. [36]

    Shunlin Lu, Ling-Hao Chen, Ailing Zeng, Jing Lin, Ruimao Zhang, Lei Zhang, and Heung-Yeung Shum. 2023. Humantomato: Text-aligned whole-body motion generation. arXiv preprint arXiv:2310.12978 (2023)

    2023 arXiv

  29. [37]

    Sihan Ma, Qiong Cao, Jing Zhang, and Dacheng Tao. 2024. Contact-aware human motion generation from textual descriptions. arXiv preprint arXiv:2403.15709 (2024)

    2024 arXiv

  30. [38]

    Mathis Petrovich, Michael J Black, and Gül Varol. 2022. Temos: Generating diverse human motions from textual descriptions. In European Conference on Computer Vision. Springer, 480–497

    2022

  31. [39]

    Haoxuan Qu, Ziyan Guo, and Jun Liu. 2024. GPT-Connect: Interaction between Text-Driven Human Motion Generator and 3D Scenes in a Training-free Manner. arXiv preprint arXiv:2403.14947 (2024)

    2024 arXiv

  32. [40]

    Muhammad A Muttaqien, Ayanori Yorozu, and Akihisa Ohya. 2024. Mobile Robots through Task-Based Human Instructions using Incremental Curriculum Learning. In 2024 IEEE International Conference on Cybernetics and Intelligent Systems (CIS) and IEEE International Conference on Robo...

    2024

  33. [41]

    William Peebles and Saining Xie. 2023. Scalable diffusion models with transform- ers. In Proceedings of the IEEE/CVF international conference on computer vision . 4195–4205

    2023

  34. [42]

    Jiaming Song, Chenlin Meng, and Stefano Ermon. 2020. Denoising diffusion implicit models. arXiv preprint arXiv:2010.02502 (2020)

    2020 arXiv

  35. [43]

    Petru Soviany, Radu Tudor Ionescu, Paolo Rota, and Nicu Sebe. 2021. Curriculum self-paced learning for cross-domain object detection.Computer Vision and Image Understanding 204 (2021), 103166

    2021

  36. [44]

    Yonatan Shafir, Guy Tevet, Roy Kapon, and Amit H Bermano. 2023. Human motion diffusion as a generative prior. arXiv preprint arXiv:2303.01418 (2023)

    2023 arXiv

  37. [45]

    Mengyi Shan, Lu Dong, Yutao Han, Yuan Yao, Tao Liu, Ifeoma Nwogu, Guo-Jun Qi, and Mitch Hill. 2024. Towards open domain text-driven synthesis of multi-person motions. In European Conference on Computer Vision . Springer, 67–86

    2024

  38. [46]

    Ashish Vaswani, Noam Shazeer, Niki Parmar, Jakob Uszkoreit, Llion Jones, Aidan N Gomez, Łukasz Kaiser, and Illia Polosukhin. 2017. Attention is all you need. Advances in neural information processing systems 30 (2017)

    2017

  39. [47]

    Weilin Wan, Zhiyang Dou, Taku Komura, Wenping Wang, Dinesh Jayaraman, and Lingjie Liu. 2024. Tlcontrol: Trajectory and language control for human motion synthesis. In European Conference on Computer Vision . Springer, 37–54

    2024

  40. [48]

    Guy Tevet, Brian Gordon, Amir Hertz, Amit H Bermano, and Daniel Cohen-Or

  41. [49]

    In European Conference on Computer Vision

    Motionclip: Exposing human motion generation to clip space. In European Conference on Computer Vision . Springer, 358–374

  42. [50]

    Guy Tevet, Sigal Raab, Brian Gordon, Yonatan Shafir, Daniel Cohen-Or, and Amit H Bermano. 2022. Human motion diffusion model. arXiv preprint Conference’17, July 2017, Washington, DC, USA Yingjie Xi, Jian Jun Zhang, and Xiaosong Yang arXiv:2209.14916 (2022)

    2022 arXiv

  43. [51]

    Zan Wang, Yixin Chen, Baoxiong Jia, Puhao Li, Jinlu Zhang, Jingze Zhang, Tengyu Liu, Yixin Zhu, Wei Liang, and Siyuan Huang. 2024. Move as you say interact as you can: Language-guided human motion generation with scene affordance. In Proceedings of the IEEE/CVF Conference on C...

    2024

  44. [52]

    Zhiming Wang, Ning Ge, and Jianhua Lu. 2025. Motion In-betweening with Spatial and Temporal Transformers. IEEE Transactions on Circuits and Systems for Video Technology (2025), 1–1. doi:10.1109/TCSVT.2025.3526236

    2025

  45. [53]

    Congyi Wang. 2023. T2m-hifigpt: generating high quality human motion from textual descriptions with residual discrete representations. arXiv preprint arXiv:2312.10628 (2023)

    2023 arXiv

  46. [54]

    Linji Wang, Zifan Xu, Peter Stone, and Xuesu Xiao. 2024. Grounded curriculum learning. arXiv preprint arXiv:2409.19816 (2024)

    2024 arXiv

  47. [55]

    Yin Wang, Mu Li, Jiapeng Liu, Zhiying Leng, Frederick WB Li, Ziyao Zhang, and Xiaohui Liang. 2025. Fg-T2M++: LLMs-Augmented Fine-Grained Text Driven Human Motion Generation. arXiv preprint arXiv:2502.05534 (2025)

    2025 arXiv

  48. [56]

    Ziming Cheng Jiangfeiyang Wang Yihao Liao, Yiyu Fu. 2024. AnimationGPT:An AIGC tool for generating game combat motion assets. https://github.com/ fyyakaxyy/AnimationGPT

    2024

  49. [57]

    Bowen Zhang, Yidong Wang, Wenxin Hou, Hao Wu, Jindong Wang, Manabu Okumura, and Takahiro Shinozaki. 2021. Flexmatch: Boosting semi-supervised learning with curriculum pseudo labeling. Advances in neural information pro- cessing systems 34 (2021), 18408–18419

    2021

  50. [58]

    Yiming Xie, Varun Jampani, Lei Zhong, Deqing Sun, and Huaizu Jiang. 2024. OmniControl: Control Any Joint at Any Time for Human Motion Generation. arXiv:2310.08580 [cs.CV] https://arxiv.org/abs/2310.08580

    2024 arXiv

  51. [59]

    Ling Yang, Zhilong Zhang, Yang Song, Shenda Hong, Runsheng Xu, Yue Zhao, Wentao Zhang, Bin Cui, and Ming-Hsuan Yang. 2023. Diffusion models: A comprehensive survey of methods and applications. Comput. Surveys 56, 4 (2023), 1–39

    2023

  52. [60]

    Zhuoyi Yang, Jiayan Teng, Wendi Zheng, Ming Ding, Shiyu Huang, Jiazheng Xu, Yuanming Yang, Wenyi Hong, Xiaohan Zhang, Guanyu Feng, et al . 2024. Cogvideox: Text-to-video diffusion models with an expert transformer. arXiv preprint arXiv:2408.06072 (2024)

    2024 arXiv

  53. [61]

    Zeyu Zhang, Yiran Wang, Wei Mao, Danning Li, Rui Zhao, Biao Wu, Zirui Song, Bohan Zhuang, Ian Reid, and Richard Hartley. 2025. Motion Anything: Any to Motion Generation. arXiv preprint arXiv:2503.06955 (2025)

    2025 arXiv

  54. [62]

    Yangyang Zhao, Zhenyu Wang, and Zhenhua Huang. 2021. Automatic curriculum learning with over-repetition penalty for dialogue policy learning. In Proceedings of the AAAI Conference on Artificial Intelligence , Vol. 35. 14540–14548

    2021

  55. [63]

    Jianrong Zhang, Yangsong Zhang, Xiaodong Cun, Yong Zhang, Hongwei Zhao, Hongtao Lu, Xi Shen, and Ying Shan. 2023. Generating human motion from textual descriptions with discrete representations. In Proceedings of the IEEE/CVF conference on computer vision and pattern recogniti...

    2023

  56. [64]

    Mingyuan Zhang, Huirong Li, Zhongang Cai, Jiawei Ren, Lei Yang, and Ziwei Liu

  57. [66]

    Xuan Zhang, Pamela Shapiro, Gaurav Kumar, Paul McNamee, Marine Carpuat, and Kevin Duh. 2019. Curriculum learning for domain adaptation in neural machine translation. arXiv preprint arXiv:1905.05816 (2019)

    2019 arXiv

  58. [69]

    Xukun Zhou, Fengxin Li, Ming Chen, Yan Zhou, Pengfei Wan, Di Zhang, Hongyan Liu, Jun He, and Zhaoxin Fan. 2025. ExGes: Expressive Human Motion Re- trieval and Modulation for Audio-Driven Gesture Synthesis. arXiv preprint arXiv:2503.06499 (2025)

    2025

  59. [70]

    Wentao Zhu, Xiaoxuan Ma, Dongwoo Ro, Hai Ci, Jinlu Zhang, Jiaxin Shi, Feng Gao, Qi Tian, and Yizhou Wang. 2023. Human motion generation: A survey.IEEE Transactions on Pattern Analysis and Machine Intelligence 46, 4 (2023), 2430–2449

    2023

  60. [2022]

    In Pro- ceedings of the IEEE/CVF Conference on Computer Vision and Pattern Recognition (CVPR)

    Generating Diverse and Natural 3D Human Motions From Text. In Pro- ceedings of the IEEE/CVF Conference on Computer Vision and Pattern Recognition (CVPR). 5152–5161

  61. [2023]

    Advances in Neural Information Processing Systems 36 (2023), 13981–13992

    Finemogen: Fine-grained spatio-temporal motion generation and editing. Advances in Neural Information Processing Systems 36 (2023), 13981–13992

    2023

  62. [2024]

    In ACM SIG- GRAPH 2024 Conference Papers

    Taming diffusion probabilistic models for character control. In ACM SIG- GRAPH 2024 Conference Papers . 1–10

    2024

Pith tools

Reviewed August 16, 2026 · model on record in the stance chip above.