Pith. sign in

REVIEW 4 major objections 5 minor 1 cited by

Human-in-Context: Unified Cross-Domain 3D Human Motion Modeling via In-Context Learning

T0 review · 4 major / 5 minor · reviewed 2026-08-15 · deepseek-v4-flash

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

Pith's one-line read Single model, trained once, handles ten 3D human-motion tasks across pose and mesh, beating prior domain-specific and cross-domain models by 21.8% on average.

desk verdict A coherent extension of PiC with real architectural ideas, but the headline improvement rests on under-specified re-implementations of baselines, including at least one broken one. read the letter →

arxiv 2508.10897 v1 pith:C6SDXUPQ submitted 2025-08-14 cs.CV

Mengyuan Liu , Xinshun Wang , Zhongbin Fang , Deheng Ye , Xia Li , Tao Tang , Songtao Wu , Xiangtai Li
show 1 more author
Ming-Hsuan Yang
This is my paper · ORCID
classification cs.CV
keywords 3Dhumanmotionmodelingin-contextlearningcross-domainmodelposeandmeshrepresentationpromptsamplingpredictionunifiedout-of-domaingeneralization
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 sets out to prove that a single 3D human-motion model can be trained once, with no domain-specific heads and no multi-stage pipeline, and still handle ten tasks spanning two modalities (pose and mesh) and four datasets. It positions HiC as the first fully unified cross-domain 3D human motion model, with in-context learning—showing the model one input–output example, or prompt, that tells it which task to perform—as the unification mechanism. Two design choices carry the argument: a max-min similarity prompt sampling that selects representative anchors matched to each query, and a dual-branch network, X-Fusion Net, that injects prompt context into the query path while aggregating dependencies at multiple levels. The paper reports that this single model outperforms prior domain-specific and cross-domain models by 21.8% on average (and its own pose-only predecessor PiC by 9.9%) on both the three in-domain datasets and the unseen 3DPW benchmark. If right, the practical payoff is that one deployable model could replace a zoo of task-specific and dataset-specific motion models.

What carries the argument

Three mechanisms carry the argument. (1) A unified cross-modal representation: pose and mesh sequences are padded with virtual joints and zero shape parameters so every task input and output lives in $\mathbb{R}^{F\times J\times C}$, making all ten tasks share one interface. (2) Max-min similarity prompt sampling (SPS): each training sequence is embedded in a relative motion similarity space anchored at the canonical T-pose; anchors are chosen iteratively by grouping unsampled sequences by maximum similarity to existing anchors and then sampling the sequence with minimum similarity to its group representative, giving $K=800$ anchors that cover dense and sparse regions; at inference the query retrieves its closest hard anchor plus a learnable soft anchor. (3) X-Fusion Net: a dual query/prompt branch network whose blocks perform multi-level context aggregation—self-attention for global embedding-space dependencies, graph convolution for local graph-space dependencies, and a state-space model for local state-space dependencies—followed by a cross-level context update that compresses the levels into influence scores and combines them with a learned frame-wise and joint-wise weighted sum, with prompt context added into the query branch each layer.

What would settle it

Train each baseline twice: once under the paper's unified single-process protocol and once with its native domain-specific heads and multi-stage training, then compare on the same ten tasks and on 3DPW. If the unified-protocol baselines lose substantially to their native versions, or if HiC's margin shrinks when baselines are given their own optimizers and loss schedules, the claim that the architecture and prompting strategy cause the improvement is falsified.

Watch

Extended reading notes

Core claim

The central claim is that cross-domain 3D human motion modeling can be formulated as in-context learning: the task is communicated entirely by a prompt pair $[\text{input}, \text{target}]$ drawn from the same domain as the query, and the model performs the corresponding task on the query without any task-specific head or fine-tuning. To make this work across modalities, the paper unifies pose and mesh sequences into one tensor format: pose joints are position coordinates, SMPL mesh joints are axis-angle rotation vectors, and gaps are filled with virtual joints and zero shape parameters. The paper then claims that two additions make the unified model succeed: max-min similarity prompt sampling, which spreads a small set of hard anchors across the data distribution and retrieves the closest one for each query, and X-Fusion Net, whose dual-branch architecture aggregates context with self-attention, graph convolution, and state-space models and updates features with learned cross-level weights. The reported evidence is a consistent win on all ten tasks on AMASS, Human3.6M, and FreeMan, and on the out-of-domain 3DPW dataset, with an average improvement of 21.8% over domain-specific and cross-domain baselines and 9.9% over PiC.

Load-bearing premise

The central comparison rests on the assumption that baseline models can be fairly re-implemented under the authors' unified setting—one training run, no domain-specific heads, one shared task formulation—so the reported 21.8% gap reflects model quality rather than the authors' choices of training recipes, hyperparameters, or losses for the baselines.

Editorial extensions

If this is right

  • A single trained model can serve ten tasks across pose and mesh modalities, replacing per-task heads and multi-stage pipelines.
  • Out-of-domain generalization improves: training on AMASS, Human3.6M, and FreeMan transfers to unseen 3DPW better than every compared baseline on all ten tasks.
  • Distribution-aware prompt retrieval helps more than random prompts: SPS beats random, farthest-point, and cluster-based sampling on both in-domain and out-of-domain benchmarks.
  • Dynamic cross-level weighting in X-Fusion Net outperforms static averaging of aggregation levels, and all three aggregators—self-attention, graph convolution, and the state-space model—contribute to the gain.
  • Scaling data roughly 21x while adding mesh tasks and a fourth dataset improves rather than degrades the model's generality.

Reading between the lines

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

  • An untested extension follows directly: the same unified tensor format could absorb additional modalities, such as RGB video or point clouds, by adding a modality-specific encoder in front of the shared in-context model, potentially yielding a single human-motion foundation model.
  • The anchor set is a non-parametric memory; replacing exhaustive similarity search with approximate nearest-neighbor retrieval could let the anchor count and training scale grow far beyond the 800 used here, which the ablations suggest would help up to a point.
  • The max-min sampling objective is task-agnostic and could be transferred to other sequential domains, such as robotic manipulation trajectories, animal motion, or hand motion, wherever a prompt pair can define the task.
  • A decisive experiment the paper does not run: test on a fifth dataset with a different body model or joint topology (for example, SMPL-X) to see whether virtual-joint padding generalizes across body definitions or degrades silently.
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

4 major / 5 minor

Summary. The paper proposes Human-in-Context (HiC), a single model trained once to handle ten 3D human motion tasks across pose and mesh modalities and four datasets (AMASS, Human3.6M, FreeMan, and the out-of-domain 3DPW). The method extends the authors' previous Pose-in-Context (PiC) with a max-min similarity prompt sampling strategy that retrieves hard anchors and paired soft anchors, and with X-Fusion Net, a dual-branch network using self-attention, graph convolution, and state-space aggregation followed by cross-level context update. The paper claims consistent improvements over re-implemented baselines, reporting a 9.9% gain over PiC and a 21.8% average gain over prior domain-specific and cross-domain models on MPJPE/MPVE metrics. The main evidence is in Tables 3 and 4, supported by ablations on anchors, soft anchors, sampling, fusion, dimensions, and layers. Code and models are promised at a public repository.

Significance. If the empirical claims hold, HiC would be a useful step toward a single fully unified 3D human motion model, and its design choices (prompt sampling with hard/soft anchors, multi-level context fusion) are reasonable and well motivated. The paper also contributes a larger in-context formulation than PiC, covering two modalities and ten tasks, and it releases code and models. The significance, however, is contingent on the credibility of the baseline comparison. The current evaluation is weakened by the fact that all baselines are re-implemented by the authors under a protocol described in one sentence, that the main comparison is against the authors' own prior work (PiC), and that some re-implemented baseline entries (PoseRetNet on mesh tasks) are two orders of magnitude worse than other methods, which suggests a non-functional adaptation rather than a fair comparison. The central quantitative claims therefore need substantial verification before the paper can be accepted.

major comments (4)
  1. [Section 5, Tables 3 and 4] The claimed 21.8% average improvement is not credible as reported because several baseline entries appear non-functional. PoseRetNet is listed with MPVE values of 478.51 and 481.95 on AMASS mesh recovery and future mesh recovery, 328.24 and 343.60 on Human3.6M, and 314.26 and 311.91 on 3DPW, while every other method on those same entries is in the 43-145 mm range. Including these entries in an average comparison inflates the headline margin. The paper never defines the formula for the 21.8% average (which tasks/datasets are averaged, whether the average is over all 50 baseline-task entries, and whether non-functional entries are included). The authors should report per-baseline averages, exclude or separately analyze non-functional task adaptations, and give the exact formula.
  2. [Section 5, Experimental Setting] The baseline re-implementation protocol is described only as 'all models are re-implemented to align with the setting of unified cross-domain 3D human motion modeling, i.e., one-time training for all tasks and datasets without any domain-specific model heads.' This is insufficient to establish that the comparison reflects model quality rather than the authors' implementation choices. No per-baseline architecture adaptation, loss weighting, training schedule, data augmentation, or compute budget is given. The suspicion is reinforced by MotionBERT obtaining 98.36 mm on H3.6M pose estimation, far above the published native-task performance of that method. The authors should provide per-baseline implementation details and, where possible, validate that each re-implementation reproduces its published native-task result before it is used in the unified setting.
  3. [Section 4.2, Eq. (4)] The similarity measure used for anchor sampling and prompt retrieval computes a Euclidean distance between raw sequence values, but those values are semantically heterogeneous: pose entries are 3D position coordinates in millimeters, while mesh entries are axis-angle rotation vectors in radians. Directly comparing these quantities in a single distance is not meaningfully justified, especially for the claimed cross-modal prompt retrieval where a pose query may be compared with mesh anchors. The paper should either define a common normalized or semantically aligned representation before applying Eq. (4), or demonstrate empirically that the retrieval behavior is robust to this unit mismatch.
  4. [Tables 3-9] No error bars or multiple-seed results are reported for any experiment, despite the fact that prompt retrieval, mask generation, and training involve stochastic choices. Since the paper's central claim is a set of consistent margins (9.9% over PiC, 21.8% over the field), the absence of variance estimates makes it impossible to tell whether the reported gaps are statistically significant. Reporting at least mean and standard deviation over three seeds for the main tables and the key ablations (Tables 5-8) is necessary to support the strength of these claims.
minor comments (5)
  1. [Section 5] There are typos in the experimental section: 'perfromance evalaution' should be 'performance evaluation', and the Figure 6 caption contains 'Recvoery' instead of 'Recovery'.
  2. [Figure 9 and Figure 10 captions] The captions describe ablations of 'multi-level context update', but the component being ablated is the multi-level context aggregation in X-Fusion blocks; the wording should be aligned with Section 4.3.2.
  3. [Eq. (2)] The notation [E(QD_j), U] for the query feature says concatenation is along the temporal axis, but the dimensions of E(QD_j) and the TUP U are not specified; please clarify how the temporal lengths match.
  4. [Section 4.2] The description of the relative motion similarity space is intuitive but not fully specified; to make the anchor sampling reproducible, the paper should state how the canonical T-body is represented in both pose and mesh forms and how sequences of different lengths are handled.
  5. [Table 2] The row for Future Mesh Recovery lists the output as {X_mesh_{F+1:2F}, beta} while the input is 2D pose from frames 1:F, but the text describing future mesh recovery in Section 3.1 does not explicitly say that the output is future frames; please make the formulation consistent.

Circularity Check

0 steps flagged · score 0.0 of 10

No significant circularity: HiC's reported gains are empirical measurements against tabulated baselines; the baseline re-implementation protocol is an evaluation-fairness concern, not a circular derivation.

full rationale

The paper's central claims are empirical rather than derived: HiC is trained once and directly evaluated on AMASS, Human3.6M, FreeMan, and 3DPW with MPJPE/MPVE metrics. The comparison against MotionBERT, PoseRetNet, TCPFormer, and HoT uses external methods, and the comparison against PiC is a tabulated re-implementation of the authors' own prior work, not an imported theorem or a fitted value renamed as a prediction. The proposed components are defined in a self-contained way: Eq. (4) defines the similarity function, Algorithm 1 specifies the anchor sampling, Eqs. (12)-(18) define the X-Fusion block, and the ablations (Tables 5-9) directly test each design choice against alternatives. I do not find any step in which a predicted quantity is identical by construction to an input or fitted parameter. The strongest concern in the paper is experimental: the re-implementation protocol is described in one sentence in Section 5, and Table 3 shows PoseRetNet reporting mesh MPVEs of 478.51 and 481.95 on AMASS, 328.24 and 343.60 on Human3.6M, and 314.26 and 311.91 on 3DPW, far outside the 43-145 mm range of all other methods, while the 21.8% average formula is never defined. This could inflate the headline margin, but it is a correctness and fairness issue rather than a circularity one: the HiC numbers are direct measurements, the model is not defined in terms of the baselines, and no load-bearing argument reduces to a self-citation. Accordingly, the appropriate finding is no significant circularity.

Assumptions & free parameters 6 free parameters · 5 assumptions · 0 invented entities

The paper introduces no new physical or conceptual entities; it proposes a model architecture, a prompting strategy, and an evaluation protocol. The main load-bearing assumptions are representational (unifying positions and rotations), procedural (the similarity space and anchor sampling), and evaluative (baseline re-implementations).

free parameters (6)
  • Number of anchors K = 800
    Chosen through ablation (Table 5) as the best trade-off between generalization and efficiency. This is a hyperparameter fitted to the validation/test performance.
  • Number of layers K = 8
    Selected through ablation (Table 9) balancing performance and parameter count.
  • Hidden feature dimension H = 128
    Selected through ablation (Figure 10) as the best-performing dimension.
  • Mask ratio = 40%
    A hand-chosen hyperparameter for all masking-based tasks (joint completion and motion in-between). No ablation is shown for this value.
  • Sequence length F = 16 frames
    Fixed for all tasks and datasets. The impact of this choice on task difficulty is not analyzed.
  • Number of joints J = 24 joints
    Used as the unified joint count after padding. It reflects the chosen mesh joint regressor rather than a property of all datasets.
assumptions (5)
  • domain assumption The unified formulation treats 3D joint positions and SMPL axis-angle rotations as directly comparable and interchangeable representations.
    Section 3.1 pads both representations to the same shape and treats them as a unified X1:F tensor. The similarity computation in Eq. (4) then computes Euclidean distances between joint positions and rotation vectors, which are geometrically different quantities.
  • domain assumption The relative motion similarity space, built on Euclidean distances between raw sequences, is a meaningful space for identifying representative anchors.
    Section 4.2 defines similarity as average Euclidean distance between raw sequences. There is no evidence that this raw-feature distance captures task-relevant distributional structure across modalities and datasets.
  • domain assumption In-context learning transfers from randomly selected prompt examples to novel queries in the same domain.
    This is the core premise of the in-context learning setup in Section 3.2 and is not proven, only demonstrated empirically.
  • ad hoc to paper The baselines are fairly re-implemented under the unified setting.
    Section 5 states all baselines are re-implemented to align with the unified setting. The paper gives no details on how each baseline was adapted, tuned, or verified against its original published results.
  • domain assumption SMPL mesh ground truth generated by MoSh for Human3.6M is accurate enough for evaluation.
    Section 3.1 follows prior work in applying MoSh to obtain SMPL parameters. Errors in this fitted representation propagate to both training targets and evaluation values.

how reviews work

0 comments
Cite this review

Pith. "Pith review of Human-in-Context: Unified Cross-Domain 3D Human Motion Modeling via In-Context Learning." pith.science (2026). https://pith.science/paper/C6SDXUPQ

@misc{pith2026250810897,
  author       = {Pith},
  title        = {Pith review of: Human-in-Context: Unified Cross-Domain 3D Human Motion Modeling via In-Context Learning},
  year         = {2026},
  howpublished = {\url{https://pith.science/paper/C6SDXUPQ}},
  note         = {Machine review of arXiv:2508.10897}
}
read the original abstract

This paper aims to model 3D human motion across domains, where a single model is expected to handle multiple modalities, tasks, and datasets. Existing cross-domain models often rely on domain-specific components and multi-stage training, which limits their practicality and scalability. To overcome these challenges, we propose a new setting to train a unified cross-domain model through a single process, eliminating the need for domain-specific components and multi-stage training. We first introduce Pose-in-Context (PiC), which leverages in-context learning to create a pose-centric cross-domain model. While PiC generalizes across multiple pose-based tasks and datasets, it encounters difficulties with modality diversity, prompting strategy, and contextual dependency handling. We thus propose Human-in-Context (HiC), an extension of PiC that broadens generalization across modalities, tasks, and datasets. HiC combines pose and mesh representations within a unified framework, expands task coverage, and incorporates larger-scale datasets. Additionally, HiC introduces a max-min similarity prompt sampling strategy to enhance generalization across diverse domains and a network architecture with dual-branch context injection for improved handling of contextual dependencies. Extensive experimental results show that HiC performs better than PiC in terms of generalization, data scale, and performance across a wide range of domains. These results demonstrate the potential of HiC for building a unified cross-domain 3D human motion model with improved flexibility and scalability. The source codes and models are available at https://github.com/BradleyWang0416/Human-in-Context.

Figures

Figures reproduced from arXiv: 2508.10897 by the authors.

Figure 1
Figure 1. Motivation (left) and performance overview (right) for our proposed Pose-in-Context and Human-in-Context. Compared to [PITH_FULL_IMAGE:figures/full_fig_p002_1.png] view at source ↗
Figure 2
Figure 2. Pipeline of the proposed Human-in-Context (HiC), with highlights on its difference from PiC. On the data end, from various [PITH_FULL_IMAGE:figures/full_fig_p005_2.png] view at source ↗
Figure 3
Figure 3. Illustration of the prompting strategy in Human-in-Context. First, a relative motion similarity space is constructed on all motion [PITH_FULL_IMAGE:figures/full_fig_p006_3.png] view at source ↗
Figures from the paper (7 more)
Figure 4
Figure 4. Figure 4: Illustration of the prompt retrieval in HiC (top) and PiC [PITH_FULL_IMAGE:figures/full_fig_p007_4.png]
Figure 5
Figure 5. Figure 5: Illustration of the X-Fusion block. The X-Fusion block applies multi-level context aggregation and cross-level context update to [PITH_FULL_IMAGE:figures/full_fig_p009_5.png]
Figure 6
Figure 6. Figure 6: Qualitative results for future mesh recovery (left) and motion in-between (mesh) (right). Significant improvements are highlighted [PITH_FULL_IMAGE:figures/full_fig_p011_6.png]
Figure 7
Figure 7. Figure 7: Visualization of features and weights produced in two inference processes. In each inference, a different prompt is paired with the [PITH_FULL_IMAGE:figures/full_fig_p012_7.png]
Figure 8
Figure 8. Figure 8: Visual comparison between our proposed max-min similarity prompt sampling, farthest point sampling, and random sampling. [PITH_FULL_IMAGE:figures/full_fig_p012_8.png]
Figure 9
Figure 9. Figure 9: Ablations on multi-level context update on AMASS [PITH_FULL_IMAGE:figures/full_fig_p015_9.png]
Figure 10
Figure 10. Figure 10: Ablations on feature dimension on AMASS (left) and [PITH_FULL_IMAGE:figures/full_fig_p015_10.png]

Discussion (0). Continue with ORCID to comment.

Forward citations

Cited by 1 Pith paper

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

  1. Superman: Unifying Skeleton and Vision for Human Motion Perception and Generation

    cs.CV 2026-02 conditional novelty 6.0 of 10

    A single MLLM trained with a vision-guided hybrid VQ-VAE tokenizer reports state-of-the-art or competitive results for 3D pose estimation, motion prediction, and motion in-betweening on Human3.6M and 3DPW.

Reference graph

Works this paper leans on

92 extracted references · 69 canonical work pages · cited by 1 Pith paper

  1. [1]

    Enhanced skeleton visualization for view invariant human action recognition,

    M. Liu, H. Liu, and C. Chen, “Enhanced skeleton visualization for view invariant human action recognition,” PR, 2017. 1, 3

    2017

  2. [2]

    3d human pose estimation= 2d pose estimation+ matching,

    C.-H. Chen and D. Ramanan, “3d human pose estimation= 2d pose estimation+ matching,” in CVPR, 2017. 1, 3

    2017

  3. [3]

    Pose2mesh: graph convolutional network for 3d human pose and mesh recovery from a 2d human pose,

    H. Choi, G. Moon, and K. M. Lee, “Pose2mesh: graph convolutional network for 3d human pose and mesh recovery from a 2d human pose,” in ECCV, 2020. 1, 2, 3

    2020

  4. [4]

    Recognizing human actions as the evolution of pose estimation maps,

    M. Liu and J. Yuan, “Recognizing human actions as the evolution of pose estimation maps,” in CVPR, 2018. 1

    2018

  5. [5]

    Milnet: multiplex interactive learning network for rgb-t semantic segmentation,

    J. Liu, H. Liu, X. Li, J. Ren, and X. Xu, “Milnet: multiplex interactive learning network for rgb-t semantic segmentation,” IEEE T-IP, 2025. 1

    2025

  6. [6]

    Dynamic dense graph convolutional network for skeleton-based human motion prediction,

    X. Wang, W. Zhang, C. Wang, Y . Gao, and M. Liu, “Dynamic dense graph convolutional network for skeleton-based human motion prediction,” IEEE T-IP, 2024. 1, 3

    2024

  7. [7]

    Symbiotic graph neural networks for 3d skeleton-based human action recognition and motion prediction,

    M. Li, S. Chen, X. Chen, Y . Zhang, Y . Wang, and Q. Tian, “Symbiotic graph neural networks for 3d skeleton-based human action recognition and motion prediction,” IEEE T-PAMI, 2021. 1, 3

    2021

  8. [8]

    Tcpformer: learning temporal correlation with implicit pose proxy for 3d human pose estimation,

    J. Liu, M. Liu, H. Liu, and W. Li, “Tcpformer: learning temporal correlation with implicit pose proxy for 3d human pose estimation,” in AAAI, 2025. 1, 10, 11

    2025

Show all 92 references
  1. [9]

    Hourglass tokenizer for efficient transformer-based 3d human pose estimation,

    W. Li, M. Liu, H. Liu, P. Wang, J. Cai, and N. Sebe, “Hourglass tokenizer for efficient transformer-based 3d human pose estimation,” inCVPR, 2024. 1, 3, 10, 11

    2024

  2. [10]

    Feature boosting network for 3d pose estimation,

    J. Liu, H. Ding, A. Shahroudy, L.-Y . Duan, X. Jiang, G. Wang, and A. C. Kot, “Feature boosting network for 3d pose estimation,” IEEE T-PAMI,

  3. [11]

    Arts: semi-analytical regressor using disentangled skeletal representations for human mesh recovery from videos,

    T. Tang, H. Liu, Y . You, T. Wang, and W. Li, “Arts: semi-analytical regressor using disentangled skeletal representations for human mesh recovery from videos,” in ACM MM, 2024. 1, 2, 3, 4

    2024

  4. [12]

    Attention is all you need,

    A. Vaswani, N. Shazeer, N. Parmar, J. Uszkoreit, L. Jones, A. N. Gomez, Ł. Kaiser, and I. Polosukhin, “Attention is all you need,” NeurIPS, 2017. 1 16

    2017

  5. [13]

    An image is worth 16x16 words: transformers for image recognition at scale,

    A. Dosovitskiy, “An image is worth 16x16 words: transformers for image recognition at scale,” arXiv preprint arXiv:2010.11929, 2020. 1, 3

    2010 arXiv

  6. [14]

    Vision mamba: efficient visual representation learning with bidirectional state space model,

    L. Zhu, B. Liao, Q. Zhang, X. Wang, W. Liu, and X. Wang, “Vision mamba: efficient visual representation learning with bidirectional state space model,” arXiv preprint arXiv:2401.09417, 2024. 1, 3

    2024 arXiv

  7. [15]

    Bert: pre-training of deep bidirectional transformers for language understanding,

    J. Devlin, M.-W. Chang, K. Lee, and K. Toutanova, “Bert: pre-training of deep bidirectional transformers for language understanding,” arXiv preprint arXiv:1810.04805, 2018. 1, 3

    2018 arXiv

  8. [16]

    Language models are few-shot learners,

    T. Brown, B. Mann, N. Ryder, M. Subbiah, J. D. Kaplan, P. Dhariwal, A. Neelakantan, P. Shyam, G. Sastry, A. Askell et al., “Language models are few-shot learners,” NeurIPS, 2020. 1, 3

    2020

  9. [17]

    Spatial temporal graph convolutional networks for skeleton-based action recognition,

    S. Yan, Y . Xiong, and D. Lin, “Spatial temporal graph convolutional networks for skeleton-based action recognition,” in AAAI, 2018. 1

    2018

  10. [18]

    Motionbert: a unified perspective on learning human motion representations,

    W. Zhu, X. Ma, Z. Liu, L. Liu, W. Wu, and Y . Wang, “Motionbert: a unified perspective on learning human motion representations,” in ICCV,

  11. [19]

    Unified pose sequence modeling,

    L. G. Foo, T. Li, H. Rahmani, Q. Ke, and J. Liu, “Unified pose sequence modeling,” in CVPR, 2023. 1, 3

    2023

  12. [20]

    Large motion model for unified multi-modal motion generation,

    M. Zhang, D. Jin, C. Gu, F. Hong, Z. Cai, J. Huang, C. Zhang, X. Guo, L. Yang, Y . Heet al., “Large motion model for unified multi-modal motion generation,” in ECCV, 2025. 1, 3

    2025

  13. [21]

    Unihcp: a unified model for human-centric perceptions,

    Y . Ci, Y . Wang, M. Chen, S. Tang, L. Bai, F. Zhu, R. Zhao, F. Yu, D. Qi, and W. Ouyang, “Unihcp: a unified model for human-centric perceptions,” in CVPR, 2023. 1, 3

    2023

  14. [22]

    Macdiff: unified skeleton modeling with masked conditional diffusion,

    L. Wu, L. Lin, J. Zhang, Y . Ma, and J. Liu, “Macdiff: unified skeleton modeling with masked conditional diffusion,” in ECCV, 2024. 1

    2024

  15. [23]

    History repeats itself: Human motion prediction via motion attention,

    W. Mao, M. Liu, and M. Salzmann, “History repeats itself: Human motion prediction via motion attention,” in ECCV, 2020. 1, 3

    2020

  16. [24]

    Gcnext: towards the unity of graph convolutions for human motion prediction,

    X. Wang, Q. Cui, C. Chen, and M. Liu, “Gcnext: towards the unity of graph convolutions for human motion prediction,” in AAAI, 2024. 1

    2024

  17. [25]

    Towards accurate 3d human motion prediction from incomplete observations,

    Q. Cui and H. Sun, “Towards accurate 3d human motion prediction from incomplete observations,” in CVPR, 2021. 1, 3

    2021

  18. [26]

    What makes good in-context examples for gpt-3?

    J. Liu, D. Shen, Y . Zhang, B. Dolan, L. Carin, and W. Chen, “What makes good in-context examples for gpt-3?” arXiv preprint arXiv:2101.06804,

    arXiv

  19. [27]

    What makes good examples for visual in-context learning?

    Y . Zhang, K. Zhou, and Z. Liu, “What makes good examples for visual in-context learning?” NeurIPS, 2024. 1, 3

    2024

  20. [28]

    Images speak in images: a generalist painter for in-context visual learning,

    X. Wang, W. Wang, Y . Cao, C. Shen, and T. Huang, “Images speak in images: a generalist painter for in-context visual learning,” in CVPR, 2023. 1, 3

    2023

  21. [29]

    Explore in-context learning for 3d point cloud understanding,

    Z. Fang, X. Li, X. Li, J. M. Buhmann, C. C. Loy, and M. Liu, “Explore in-context learning for 3d point cloud understanding,” NeurIPS, 2023. 1, 3

    2023

  22. [30]

    Skeleton-in-context: unified skeleton sequence modeling with in-context learning,

    X. Wang, Z. Fang, X. Li, X. Li, C. Chen, and M. Liu, “Skeleton-in-context: unified skeleton sequence modeling with in-context learning,” in CVPR,

  23. [31]

    On human motion prediction using recurrent neural networks,

    J. Martinez, M. J. Black, and J. Romero, “On human motion prediction using recurrent neural networks,” in CVPR, 2017. 2, 3

    2017

  24. [32]

    Convolutional sequence to sequence model for human dynamics,

    C. Li, Z. Zhang, W. S. Lee, and G. H. Lee, “Convolutional sequence to sequence model for human dynamics,” in CVPR, 2018. 2, 3

    2018

  25. [33]

    Multiscale spatio-temporal graph neural networks for 3d skeleton-based motion prediction,

    M. Li, S. Chen, Y . Zhao, Y . Zhang, Y . Wang, and Q. Tian, “Multiscale spatio-temporal graph neural networks for 3d skeleton-based motion prediction,” IEEE T-IP, 2021. 2, 3

    2021

  26. [34]

    Learning trajectory dependen- cies for human motion prediction,

    W. Mao, M. Liu, M. Salzmann, and H. Li, “Learning trajectory dependen- cies for human motion prediction,” in ICCV, 2019. 2, 3

    2019

  27. [35]

    Spatiotemporal co- attention recurrent neural networks for human-skeleton motion prediction,

    X. Shu, L. Zhang, G.-J. Qi, W. Liu, and J. Tang, “Spatiotemporal co- attention recurrent neural networks for human-skeleton motion prediction,” IEEE T-PAMI, 2021. 2

    2021

  28. [36]

    A simple yet effective baseline for 3d human pose estimation,

    J. Martinez, R. Hossain, J. Romero, and J. J. Little, “A simple yet effective baseline for 3d human pose estimation,” in ICCV, 2017. 2

    2017

  29. [37]

    Mixste: seq2seq mixed spatio-temporal encoder for 3d human pose estimation in video,

    J. Zhang, Z. Tu, J. Yang, Y . Chen, and J. Yuan, “Mixste: seq2seq mixed spatio-temporal encoder for 3d human pose estimation in video,” in CVPR,

  30. [38]

    Mhformer: multi- hypothesis transformer for 3d human pose estimation,

    W. Li, H. Liu, H. Tang, P. Wang, and L. Van Gool, “Mhformer: multi- hypothesis transformer for 3d human pose estimation,” in CVPR, 2022. 2, 3

    2022

  31. [39]

    Diffpose: toward more reliable 3d pose estimation,

    J. Gong, L. G. Foo, Z. Fan, Q. Ke, H. Rahmani, and J. Liu, “Diffpose: toward more reliable 3d pose estimation,” in CVPR, 2023. 2

    2023

  32. [40]

    App: adaptive pose pooling for 3d human pose estimation from videos,

    J. Zhang, M. Liu, H. Liu, G. Wang, and W. Li, “App: adaptive pose pooling for 3d human pose estimation from videos,” in ACM MM, 2024. 2

    2024

  33. [41]

    Finepose: fine-grained prompt-driven 3d human pose estimation via diffusion models,

    J. Xu, Y . Guo, and Y . Peng, “Finepose: fine-grained prompt-driven 3d human pose estimation via diffusion models,” in CVPR, 2024. 2, 3

    2024

  34. [42]

    Hybrik: a hybrid analytical-neural inverse kinematics solution for 3d human pose and shape estimation,

    J. Li, C. Xu, Z. Chen, S. Bian, L. Yang, and C. Lu, “Hybrik: a hybrid analytical-neural inverse kinematics solution for 3d human pose and shape estimation,” in CVPR, 2021. 2, 3, 4

    2021

  35. [43]

    Skeleton2mesh: kinematics prior injected unsupervised human mesh recovery,

    Z. Yu, J. Wang, J. Xu, B. Ni, C. Zhao, M. Wang, and W. Zhang, “Skeleton2mesh: kinematics prior injected unsupervised human mesh recovery,” in ICCV, 2021. 2

    2021

  36. [44]

    Convolutional sequence generation for skeleton-based action synthesis,

    S. Yan, Z. Li, Y . Xiong, H. Yan, and D. Lin, “Convolutional sequence generation for skeleton-based action synthesis,” in ICCV, 2019. 2, 3

    2019

  37. [45]

    Generative tweening: long-term inbetweening of 3d human motions,

    Y . Zhou, J. Lu, C. Barnes, J. Yang, S. Xiang et al. , “Generative tweening: long-term inbetweening of 3d human motions,” arXiv preprint arXiv:2005.08891, 2020. 2

    2005 arXiv

  38. [46]

    Convolutional autoencoders for human motion infilling,

    M. Kaufmann, E. Aksan, J. Song, F. Pece, R. Ziegler, and O. Hilliges, “Convolutional autoencoders for human motion infilling,” in 3DV, 2020. 2, 3

    2020

  39. [47]

    Human motion prediction via spatio-temporal inpainting,

    A. Hernandez, J. Gall, and F. Moreno-Noguer, “Human motion prediction via spatio-temporal inpainting,” in ICCV, 2019. 2

    2019

  40. [48]

    Coupled action recognition and pose estimation from multiple views,

    A. Yao, J. Gall, and L. Van Gool, “Coupled action recognition and pose estimation from multiple views,” IJCV, 2012. 2, 3

    2012

  41. [49]

    SMPL: a skinned multi-person linear model,

    M. Loper, N. Mahmood, J. Romero, G. Pons-Moll, and M. J. Black, “SMPL: a skinned multi-person linear model,” ACM TOG, 2015. 2, 3, 6

    2015

  42. [50]

    Keep it smpl: automatic estimation of 3d human pose and shape from a single image,

    F. Bogo, A. Kanazawa, C. Lassner, P. Gehler, J. Romero, and M. J. Black, “Keep it smpl: automatic estimation of 3d human pose and shape from a single image,” in ECCV, 2016. 2, 3, 4

    2016

  43. [51]

    Unsupervised learning of view-invariant action representations,

    J. Li, Y . Wong, Q. Zhao, and M. S. Kankanhalli, “Unsupervised learning of view-invariant action representations,” in NeurIPS, 2018. 2

    2018

  44. [52]

    Motion guided 3d pose estimation from videos,

    J. Wang, S. Yan, Y . Xiong, and D. Lin, “Motion guided 3d pose estimation from videos,” in ECCV, 2020. 2

    2020

  45. [53]

    Human3. 6m: large scale datasets and predictive methods for 3d human sensing in natural environments,

    C. Ionescu, D. Papava, V . Olaru, and C. Sminchisescu, “Human3. 6m: large scale datasets and predictive methods for 3d human sensing in natural environments,” IEEE T-PAMI, 2013. 2, 4, 10, 11

    2013

  46. [54]

    Recovering accurate 3d human pose in the wild using imus and a moving camera,

    T. V on Marcard, R. Henschel, M. J. Black, B. Rosenhahn, and G. Pons- Moll, “Recovering accurate 3d human pose in the wild using imus and a moving camera,” in ECCV, 2018. 2, 4, 10, 11, 13

    2018

  47. [55]

    Amass: archive of motion capture as surface shapes,

    N. Mahmood, N. Ghorbani, N. F. Troje, G. Pons-Moll, and M. J. Black, “Amass: archive of motion capture as surface shapes,” in ICCV, 2019. 2, 4, 10, 11

    2019

  48. [56]

    Freeman: towards benchmarking 3d human pose estimation under real-world conditions,

    J. Wang, F. Yang, B. Li, W. Gou, D. Yan, A. Zeng, Y . Gao, J. Wang, Y . Jing, and R. Zhang, “Freeman: towards benchmarking 3d human pose estimation under real-world conditions,” in CVPR, 2024. 2, 4, 10, 11

    2024

  49. [57]

    Ntu rgb+ d: a large scale dataset for 3d human activity analysis,

    A. Shahroudy, J. Liu, T.-T. Ng, and G. Wang, “Ntu rgb+ d: a large scale dataset for 3d human activity analysis,” in CVPR, 2016. 3

    2016

  50. [58]

    Ntu rgb+ d 120: a large-scale benchmark for 3d human activity understanding,

    J. Liu, A. Shahroudy, M. Perez, G. Wang, L.-Y . Duan, and A. C. Kot, “Ntu rgb+ d 120: a large-scale benchmark for 3d human activity understanding,” IEEE T-PAMI, 2019. 3

    2019

  51. [59]

    Disentangling and unifying graph convolutions for skeleton-based action recognition,

    Z. Liu, H. Zhang, Z. Chen, Z. Wang, and W. Ouyang, “Disentangling and unifying graph convolutions for skeleton-based action recognition,” in CVPR, 2020. 3

    2020

  52. [60]

    Vg4d: vision- language model goes 4d video recognition,

    Z. Deng, X. Li, X. Li, Y . Tong, S. Zhao, and M. Liu, “Vg4d: vision- language model goes 4d video recognition,” in ICRA, 2024. 3

    2024

  53. [61]

    A comprehensive study of weight sharing in graph networks for 3d human pose estimation,

    K. Liu, R. Ding, Z. Zou, L. Wang, and W. Tang, “A comprehensive study of weight sharing in graph networks for 3d human pose estimation,” in ECCV, 2020. 3

    2020

  54. [62]

    Progressively generating better initial guesses towards next stages for high-quality human motion prediction,

    T. Ma, Y . Nie, C. Long, Q. Zhang, and G. Li, “Progressively generating better initial guesses towards next stages for high-quality human motion prediction,” in CVPR, 2022. 3

    2022

  55. [63]

    Graph stacked hourglass networks for 3d human pose estimation,

    T. Xu and W. Takano, “Graph stacked hourglass networks for 3d human pose estimation,” in CVPR, 2021. 3

    2021

  56. [64]

    Back to mlp: a simple baseline for human motion prediction,

    W. Guo, Y . Du, X. Shen, V . Lepetit, X. Alameda-Pineda, and F. Moreno- Noguer, “Back to mlp: a simple baseline for human motion prediction,” in WACV, 2023. 3

    2023

  57. [65]

    3d human pose estimation in video with temporal convolutions and semi-supervised training,

    D. Pavllo, C. Feichtenhofer, D. Grangier, and M. Auli, “3d human pose estimation in video with temporal convolutions and semi-supervised training,” in CVPR, 2019. 3

    2019

  58. [66]

    Co- occurrence feature learning for skeleton based action recognition using regularized deep lstm networks,

    W. Zhu, C. Lan, J. Xing, W. Zeng, Y . Li, L. Shen, and X. Xie, “Co- occurrence feature learning for skeleton based action recognition using regularized deep lstm networks,” in AAAI, 2016. 3

    2016

  59. [67]

    Spatio-temporal lstm with trust gates for 3d human action recognition,

    J. Liu, A. Shahroudy, D. Xu, and G. Wang, “Spatio-temporal lstm with trust gates for 3d human action recognition,” in ECCV, 2016. 3

    2016

  60. [68]

    Exploiting spatial-temporal relationships for 3d pose estimation via graph convolutional networks,

    Y . Cai, L. Ge, J. Liu, J. Cai, T.-J. Cham, J. Yuan, and N. M. Thalmann, “Exploiting spatial-temporal relationships for 3d pose estimation via graph convolutional networks,” in ICCV, 2019. 3

    2019

  61. [69]

    Learning dynamic relationships for 3d human motion prediction,

    Q. Cui, H. Sun, and F. Yang, “Learning dynamic relationships for 3d human motion prediction,” in CVPR, 2020. 3

    2020

  62. [70]

    3d human pose estimation with spatial and temporal transformers,

    C. Zheng, S. Zhu, M. Mendieta, T. Yang, C. Chen, and Z. Ding, “3d human pose estimation with spatial and temporal transformers,” in ICCV,

  63. [71]

    Graformer: graph-oriented transformer for 3d pose estimation,

    W. Zhao, W. Wang, and Y . Tian, “Graformer: graph-oriented transformer for 3d pose estimation,” in CVPR, 2022. 3 17

    2022

  64. [72]

    Motionagformer: enhancing 3d human pose estimation with a transformer-gcnformer network,

    S. Mehraban, V . Adeli, and B. Taati, “Motionagformer: enhancing 3d human pose estimation with a transformer-gcnformer network,” in WACV,

  65. [73]

    Mamba: linear-time sequence modeling with selective state spaces,

    A. Gu and T. Dao, “Mamba: linear-time sequence modeling with selective state spaces,” arXiv preprint arXiv:2312.00752, 2023. 3

    2023 arXiv

  66. [74]

    Pose magic: efficient and temporally consistent human pose estimation with a hybrid mamba- gcn network,

    X. Zhang, Q. Bao, Q. Cui, W. Yang, and Q. Liao, “Pose magic: efficient and temporally consistent human pose estimation with a hybrid mamba- gcn network,” arXiv preprint arXiv:2408.02922, 2024. 3

    2024 arXiv

  67. [75]

    Simba: mamba augmented u-shiftgcn for skeletal action recognition in videos,

    S. Chaudhuri and S. Bhattacharya, “Simba: mamba augmented u-shiftgcn for skeletal action recognition in videos,”arXiv preprint arXiv:2404.07645,

    arXiv

  68. [76]

    Omg-seg: is one model good enough for all segmentation?

    X. Li, H. Yuan, W. Li, H. Ding, S. Wu, W. Zhang, Y . Li, K. Chen, and C. C. Loy, “Omg-seg: is one model good enough for all segmentation?” in CVPR, 2024. 3

    2024

  69. [77]

    Seggpt: towards segmenting everything in context,

    X. Wang, X. Zhang, Y . Cao, W. Wang, C. Shen, and T. Huang, “Seggpt: towards segmenting everything in context,” in ICCV, 2023. 3

    2023

  70. [78]

    Learning to reconstruct 3d human pose and shape via model-fitting in the loop,

    N. Kolotouros, G. Pavlakos, M. J. Black, and K. Daniilidis, “Learning to reconstruct 3d human pose and shape via model-fitting in the loop,” in ICCV, 2019. 3, 4

    2019

  71. [79]

    2d/3d pose estimation and action recognition using multitask deep learning,

    D. C. Luvizon, D. Picard, and H. Tabia, “2d/3d pose estimation and action recognition using multitask deep learning,” in CVPR, 2018. 3

    2018

  72. [80]

    A unified 3d human motion synthesis model via conditional variational auto-encoder,

    Y . Cai, Y . Wang, Y . Zhu, T.-J. Cham, J. Cai, J. Yuan, J. Liu, C. Zheng, S. Yan, H. Ding et al., “A unified 3d human motion synthesis model via conditional variational auto-encoder,” in ICCV, 2021. 3

    2021

  73. [81]

    Learning to retrieve prompts for in-context learning,

    O. Rubin, J. Herzig, and J. Berant, “Learning to retrieve prompts for in-context learning,” arXiv preprint arXiv:2112.08633, 2021. 3

    2021 arXiv

  74. [82]

    Learning transferable visual models from natural language supervision,

    A. Radford, J. W. Kim, C. Hallacy, A. Ramesh, G. Goh, S. Agarwal, G. Sastry, A. Askell, P. Mishkin, J. Clark et al., “Learning transferable visual models from natural language supervision,” in ICML, 2021. 3

    2021

  75. [83]

    In-context learning unlocked for diffusion models,

    Z. Wang, Y . Jiang, Y . Lu, Y . Shen, P. He, W. Chen, Z. Wang, and M. Zhou, “In-context learning unlocked for diffusion models,” arXiv preprint arXiv:2305.01115, 2023. 3

    2023 arXiv

  76. [84]

    Visual prompting via image inpainting,

    A. Bar, Y . Gandelsman, T. Darrell, A. Globerson, and A. Efros, “Visual prompting via image inpainting,” NeurIPS, 2022. 3

    2022

  77. [85]

    Explore in-context segmentation via latent diffusion models,

    C. Wang, X. Li, H. Ding, L. Qi, J. Zhang, Y . Tong, C. C. Loy, and S. Yan, “Explore in-context segmentation via latent diffusion models,”AAAI, 2025. 3

    2025

  78. [86]

    Towards large-scale 3d representation learning with multi-dataset point prompt training,

    X. Wu, Z. Tian, X. Wen, B. Peng, X. Liu, K. Yu, and H. Zhao, “Towards large-scale 3d representation learning with multi-dataset point prompt training,” CVPR, 2024. 3

    2024

  79. [87]

    Exploring effec- tive factors for improving visual in-context learning,

    Y . Sun, Q. Chen, J. Wang, J. Wang, and Z. Li, “Exploring effec- tive factors for improving visual in-context learning,” arXiv preprint arXiv:2304.04748, 2023. 3

    2023

  80. [88]

    Mosh: motion and shape capture from sparse markers

    M. Loper, N. Mahmood, and M. J. Black, “Mosh: motion and shape capture from sparse markers.” ACM TOG, 2014. 4

    2014

  81. [89]

    End-to-end recovery of human shape and pose,

    A. Kanazawa, M. J. Black, D. W. Jacobs, and J. Malik, “End-to-end recovery of human shape and pose,” in CVPR, 2018. 4

    2018

  82. [90]

    Pymaf: 3d human pose and shape regression with pyramidal mesh alignment feedback loop,

    H. Zhang, Y . Tian, X. Zhou, W. Ouyang, Y . Liu, L. Wang, and Z. Sun, “Pymaf: 3d human pose and shape regression with pyramidal mesh alignment feedback loop,” in CVPR, 2021. 4

    2021

  83. [91]

    3d human pose estimation via non-causal retentive networks,

    K. Zheng, F. Lu, Y . Lv, L. Zhang, C. Guo, and J. Wu, “3d human pose estimation via non-causal retentive networks,” in ECCV, 2025. 10, 11 Mengyuan Liu received his Ph.D. degree from the School of Electrical Engineering and Com- puter Science, Peking University, China. He was a...

    2025

  84. [2023]

    1, 3, 4, 6, 8, 10, 11

Pith tools

Reviewed August 15, 2026 · model on record in the stance chip above.