Pith. sign in

REVIEW 4 major objections 5 minor 1 cited by

IMoRe: Implicit Program-Guided Reasoning for Human Motion Q&A

T0 review · 4 major / 5 minor · reviewed 2026-08-06 · deepseek-v4-flash

Pith's one-line read Implicit program-guided reasoning, not handcrafted modules, carries human motion question answering to new best accuracies.

desk verdict Solid architecture paper for motion QA, but the headline SOTA claim is confounded by the backbone; the internal ablation is the part to trust. read the letter →

arxiv 2508.01984 v1 pith:NQK4XMLY submitted 2025-08-04 cs.CV

classification cs.CV
keywords humanmotionquestionansweringprogram-guidedreasoningimplicitneuro-symbolicMemory-Attention-CompositionnetworkVisionTransformerBabel-QAHuMMan-QA
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

This paper argues that human motion question answering does not need handcrafted functional modules, and that implicit program-guided reasoning can do the job better. The authors propose IMoRe, a memory-based reasoning network that conditions every reasoning step on a structured program's functions rather than on soft attention over question words, and that reads from multiple layers of a pretrained motion Vision Transformer. On the Babel-QA benchmark its strongest setting reaches 64.0% overall accuracy, against 57.8% for the explicit neuro-symbolic baseline NSPose, and it retains most of that advantage when programs are predicted rather than given. The authors also introduce a HuMMan-QA dataset and report that IMoRe generalizes to it, reaching 73.0% versus 69.1% for NSPose. If correct, the contribution is a single adaptable reasoning architecture that removes the manual module design bottleneck while keeping the interpretability of program structure.

What carries the argument

The load-bearing object is the implicit program-guided reasoning module, a MAC-style Memory-Attention-Composition cell with control, read, and write operations, but with the control signal taken directly from structured program functions instead of inferred by soft attention over question words. At step $i+1$ the module first attends to previous memory states using program function $P_{i+1}$ as the query, then reads from a knowledge pool of projected multi-level features $F'_m = \{h'_1, \dots, h'_M\}$ extracted from six middle layers and the final layer of a pretrained motion Vision Transformer. The program's dependency structure, such as $\text{relate}(\text{before})$ depending on $\text{filter}(\text{left})$, tells the network which earlier memory states matter, and the program-guided reading mechanism chooses which feature level to attend to, so action queries can use high-level semantics while body-part queries use fine-grained local detail. In the IMoRe II variant, 224-frame windows are randomly sampled from each sequence and the best-scoring of five test windows is kept, a protocol the paper uses because the highest-logit segment should contain the queried concept. This replaces the handcrafted modules of explicit neuro-symbolic execution while preserving an interpretable step-by-step trace.

What would settle it

Run IMoRe and NSPose on Babel-QA with the same motion encoder and with programs replaced by random or empty programs; if IMoRe's margin over NSPose persists without meaningful program information, the reported gain is not driven by program guidance. A second check is to report IMoRe II's accuracy for a single randomly chosen 224-frame window instead of the best of five; if single-window accuracy falls much below 64.0%, part of the headline result comes from window selection rather than reasoning.

Watch

Extended reading notes

Core claim

The central claim is that symbolic program functions should serve as a conditioning signal for a unified implicit reasoning network, not as executable instructions for handcrafted modules. In IMoRe, the program P is read step by step: each reasoning step attends to the relevant previous memory states and to motion features selected from a multi-level knowledge pool built from six intermediate layers plus the final output of the motion ViT. This replaces MAC's soft attention over question words with explicit program guidance, and replaces NSPose's separate functional modules with one memory network. The paper reports that this design yields 64.0% overall accuracy on Babel-QA in the ground-truth program setting (IMoRe II), beats NSPose's 57.8%, and that with predicted programs the drop is roughly 2.5 points for IMoRe II whereas NSPose drops about 10.6 points. On the new HuMMan-QA dataset IMoRe reaches 73.0% versus 69.1% for NSPose, which the authors take as evidence that the approach generalizes across datasets. The paper also argues that implicit reasoning's resilience to noisy program predictions is a structural advantage over explicit execution, where one mistaken step can cascade.

Load-bearing premise

The load-bearing premise is that every question comes with a structured program P whose functions align with the reasoning steps, and that the pretrained motion transformer's layers genuinely separate high-level action semantics from fine-grained body-part detail; if either fails, the program signal and the reading mechanism lose their value.

Editorial extensions

If this is right

  • If the central claim holds, new motion QA systems can be built without designing one module per operation type; a single reasoning stack conditioned on programs handles action, direction, and body-part questions.
  • Predicted-program robustness becomes a real deployment advantage: the paper reports only a 1-2% drop for IMoRe under predicted programs while the explicit baseline NSPose loses roughly ten points, implying the implicit network can absorb noisy control signals.
  • The multi-level reading mechanism implies that a single pretrained motion transformer can supply both the global action semantics and the local body-part cues needed for fine-grained reasoning, so feature-level selection is part of the reasoning itself.
  • The introduced HuMMan-QA dataset, with 2066 training, 524 validation, and 533 test questions generated with GPT-4o and manually verified, provides a second benchmark that future motion QA work can use to test generalization.

Reading between the lines

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

  • Editorial extension: the same program-conditioned implicit reasoning recipe could apply to other temporal reasoning tasks, such as video QA or instruction following, whenever a structured program can be supplied.
  • Editorial extension: the best-of-five random-frame selection used in IMoRe II is an optimistic oracle; reporting single-window accuracy and the variance across the five runs would show how much of the 64.0% gain depends on that selection rather than on reasoning.
  • Editorial extension: a controlled noise experiment on programs, varying the fraction of randomly replaced functions, would directly test the paper's claim that implicit reasoning degrades gracefully while explicit execution cascades.
  • Editorial extension: the attention visualization of feature-level selection suggests a testable prediction—if lower ViT layers are removed from the knowledge pool, body-part accuracy should drop more than action accuracy.
Share X Bluesky LinkedIn Reddit HN

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 IMoRe, an implicit program-guided reasoning framework for human motion question answering. Instead of executing handcrafted neuro-symbolic modules as in NSPose, IMoRe uses a MAC-style iterative memory network whose reasoning steps are conditioned on structured program functions, together with a program-guided reading mechanism that selects from multi-level features of a pretrained motion ViT. The authors evaluate on the Babel-QA benchmark and introduce a new HuMMan-QA dataset. They report state-of-the-art results: IMoRe II reaches 64.0% overall on Babel-QA versus 57.8% for NSPose, and 73.0% versus 69.1% on HuMMan-QA. They also report robustness to predicted programs, an ablation study, and comparisons with video-language model baselines.

Significance. If validated, the paper makes a useful architectural contribution: replacing handcrafted explicit modules with program-conditioned implicit reasoning, while retaining interpretability through program structure and intermediate attention maps. The release of code and a new HuMMan-QA dataset, the inclusion of a predicted-program setting, and the addition of VLM baselines are all strengths. However, the central claim that implicit program-guided reasoning outperforms explicit neuro-symbolic execution is not yet isolated from two confounding factors: the motion encoder is not matched across comparisons, and the best IMoRe II result uses a test-time best-of-five selection rule not applied to baselines. These issues require additional experiments before the headline performance gain can be attributed to the reasoning architecture.

major comments (4)
  1. [§4.3, Table 3] The ablation intended to control for the motion encoder does not isolate the reasoning contribution. Row A, 'ViT + NSPose', drops to 0.440 overall, below NSPose's native 0.578 in Table 1, and the paper interprets this as evidence that the gain comes from the reasoning module. However, NSPose's explicit modules were designed for the original joint-level encoder, and patch-based ViT features are likely incompatible with them; the row is therefore a strawman baseline. The paper provides no experiment running IMoRe with NSPose's encoder and no fair adaptation of NSPose to ViT features, so the 3.1–6.2 point gap in Table 1 may be substantially due to the stronger backbone. Please add an encoder-matched comparison or explicitly qualify the claim.
  2. [§4, Implementation Details] The headline IMoRe II result uses a test-time best-of-five selection rule: 'During validation and testing, we execute five runs for the same sequence and select the run with the highest logits scores.' This selection is not applied to any baseline or to IMoRe I, so the Table 1 comparison between IMoRe II (0.640) and NSPose (0.578) is not apples-to-apples. Since the selection is an oracle-like operation over random 224-frame windows, it can only help IMoRe II. Please report IMoRe II without highest-logit selection, and/or apply the same best-of-k procedure to baselines that can use it. In addition, the three-run averages are reported without standard deviations or significance tests; with gaps as small as 3.1 points, this leaves the reader unable to assess whether differences are reliable.
  3. [§4, Implementation Details; §4.1, Table 1] The predicted-program robustness claim is not fully supported. The text says only that 'we adapt the coarse-to-fine two-stage program generation network from [9]' to predict programs, with no details about training data, supervision, or predicted-program accuracy on Babel-QA. It is also unclear whether the NSPose* numbers in Table 1 use the same predicted programs as IMoRe I* and IMoRe II*. Without these details, the conclusion that implicit reasoning degrades only 1–2% while explicit reasoning degrades more under noisy programs cannot be evaluated. Please describe the program predictor, report its accuracy, and state whether both methods consume identical predicted programs.
  4. [§4, Datasets] The new HuMMan-QA dataset is a core contribution, but its label quality is not documented enough to support the generalization claim. The paper states that GPT-4o generated labels 'which were then meticulously refined and manually verified against the actual motion sequences by the authors.' No verification protocol, inter-annotator agreement, or per-question-type error analysis is provided. Since the labels for action, direction, and body part come from a model plus manual checking rather than from existing annotations, the benchmark's reliability needs quantitative support. Please include annotation statistics and, if possible, a small human-study comparison. Also clarify how ground-truth programs are defined for HuMMan-QA questions.
minor comments (5)
  1. [§3.3, Eq. (3)] The sentence preceding Eq. (3) says the attention uses 'the key and query', but it should say 'the key and value'; Eq. (3) itself correctly specifies K=hm, V=hm.
  2. [Table 3 vs Table 1] Row D of Table 3 reports 0.000 for the body-part 'BTW' condition, while Table 1 reports 0.083 for IMoRe II in the same condition; these should be reconciled if Row D is intended to be the full IMoRe II model.
  3. [Title page] The affiliations use the number 1 for both the Institute of High-Performance Computing and the Centre for Frontier AI Research; the second affiliation should be numbered 2.
  4. [Figure 3] Figure 3 appears to contain repeated panels and inconsistent captions (for example, multiple 'Query Action (Relation: After)' blocks with the same question); please replace with a clean set of distinct qualitative examples.
  5. [Throughout] The paper alternates between 'NS-Pose' and 'NSPose'; please use a single consistent name for the baseline.

Circularity Check

0 steps flagged · score 0.0 of 10

No significant circularity: the central result is an empirical architecture comparison on external benchmarks, and the program input does not by construction encode the answer label.

full rationale

The paper's claimed contribution is an empirical architecture evaluated on Babel-QA and a newly constructed HuMMan-QA dataset, not a formal derivation whose conclusion is built into its premises. IMoRe's inputs are the motion sequence, the question text, and a structured program; the answer label appears only in the standard cross-entropy training loss (Eq. 7), never as a direct input. The program-guided reasoning module (Eqs. 2, 3, 5) is a learned attention mechanism over memory states and multi-level ViT features; no equation reduces the predicted answer to the program alone. The main comparisons in Tables 1 and 2 are against NSPose, direct feature-fusion baselines, and video-language models, all evaluated on held-out test questions. The only references with overlapping authors are [17] and [29], both cited in the related-work survey as examples of neuro-symbolic reasoning; they are not load-bearing for the paper's architecture or results, and no uniqueness theorem or ansatz is imported from them. The 'ViT + NSPose' ablation (Table 3, row A) is a potential experimental confound, since replacing NSPose's encoder with the ViT may not be a fair adaptation, but that is a correctness and robustness concern, not circularity: the paper does not define the reasoning advantage in terms of that baseline. The predicted-program experiments use an external predictor [9] and are presented as robustness checks, not as the source of the main claim. Overall, no circular step can be exhibited with the required specificity, so the appropriate finding is no significant circularity.

Assumptions & free parameters 2 free parameters · 3 assumptions · 0 invented entities

The central claim rests on standard supervised learning plus two domain assumptions: that a structured program is available for each question, and that ViT layer depth corresponds to semantic granularity. The hand-chosen numbers with the largest effect on the headline are the six selected ViT layers and the five-window highest-logit inference in IMoRe II. No new entity is introduced.

free parameters (2)
  • Number of ViT layers in the multi-level knowledge pool = M=6 layers (0, 2, 4, 6, 8, 11) plus final encoder output
    Hand-selected in Section 4, not swept; determines the feature pool used by the program-guided reading mechanism.
  • Test-time random segment runs with highest-logit selection = K=5
    IMoRe II samples five random 224-frame windows at validation and test and keeps the run with the highest logits, a hand-chosen inference policy that can inflate the reported accuracy.
assumptions (3)
  • domain assumption A structured program P is available for every QA pair, either as ground truth or from an external predictor.
    Stated in Section 3.1; every reasoning step attends to program functions, so without P the model cannot execute.
  • domain assumption Pretrained motion ViT layers form a semantic hierarchy: high layers for action-level semantics, low layers for body-part detail.
    Assumed in Section 3.4 to justify the multi-level knowledge pool; only attention visualizations support it.
  • domain assumption The ground-truth programs in Babel-QA and the generated programs in HuMMan-QA are sufficiently correct to train the implicit reasoner.
    The training loss in Eq. 7 conditions on P as a given; noisy or incorrect P would mislead the memory states.

how reviews work

0 comments
Cite this review

Pith. "Pith review of IMoRe: Implicit Program-Guided Reasoning for Human Motion Q&A." pith.science (2026). https://pith.science/paper/NQK4XMLY

@misc{pith2026250801984,
  author       = {Pith},
  title        = {Pith review of: IMoRe: Implicit Program-Guided Reasoning for Human Motion Q&A},
  year         = {2026},
  howpublished = {\url{https://pith.science/paper/NQK4XMLY}},
  note         = {Machine review of arXiv:2508.01984}
}
read the original abstract

Existing human motion Q\&A methods rely on explicit program execution, where the requirement for manually defined functional modules may limit the scalability and adaptability. To overcome this, we propose an implicit program-guided motion reasoning (IMoRe) framework that unifies reasoning across multiple query types without manually designed modules. Unlike existing implicit reasoning approaches that infer reasoning operations from question words, our model directly conditions on structured program functions, ensuring a more precise execution of reasoning steps. Additionally, we introduce a program-guided reading mechanism, which dynamically selects multi-level motion representations from a pretrained motion Vision Transformer (ViT), capturing both high-level semantics and fine-grained motion cues. The reasoning module iteratively refines memory representations, leveraging structured program functions to extract relevant information for different query types. Our model achieves state-of-the-art performance on Babel-QA and generalizes to a newly constructed motion Q\&A dataset based on HuMMan, demonstrating its adaptability across different motion reasoning datasets. Code and dataset are available at: https://github.com/LUNAProject22/IMoRe.

Figures

Figures reproduced from arXiv: 2508.01984 by the authors.

Figure 1
Figure 1. Illustration of the Program-Guided Reasoning Process [PITH_FULL_IMAGE:figures/full_fig_p001_1.png] view at source ↗
Figure 2
Figure 2. (a) Overall pipeline of our Program-Guided Motion Reasoning Framework. Motion and text features are extracted using a [PITH_FULL_IMAGE:figures/full_fig_p003_2.png] view at source ↗
Figure 3
Figure 3. Qualitative comparison between our method (IMoRe) and NSPose on motion sequence reasoning tasks with different query types: [PITH_FULL_IMAGE:figures/full_fig_p008_3.png] view at source ↗
Figures from the paper (2 more)
Figure 1
Figure 1. Figure 1: Visualization of feature level selection. [PITH_FULL_IMAGE:figures/full_fig_p011_1.png]
Figure 2
Figure 2. Figure 2: Visualization of concept localization. 11 [PITH_FULL_IMAGE:figures/full_fig_p011_2.png]

Discussion (0). Sign in to comment.

Forward citations

Cited by 1 Pith paper

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

  1. LLaMo: Scaling Pretrained Language Models for Unified Motion Understanding and Generation with Continuous Autoregressive Tokens

    cs.CV 2026-02 unverdicted novelty 6.0 of 10

    LLaMo scales pretrained LLMs for unified motion-language tasks by encoding motion into continuous causal latents and adding a flow-matching head for real-time autoregressive generation and captioning.

Reference graph

Works this paper leans on

40 extracted references · 31 canonical work pages · cited by 1 Pith paper

  1. [9]

    Coarse-to-fine decoding for neural semantic parsing

    Li Dong and Mirella Lapata. Coarse-to-fine decoding for neural semantic parsing. arXiv preprint arXiv:1805.04793,

  2. [1]

    Neural module networks

    Jacob Andreas, Marcus Rohrbach, Trevor Darrell, and Dan Klein. Neural module networks. In Proceedings of the IEEE conference on computer vision and pattern recognition, pages 39–48, 2016. 3

  3. [2]

    Dynamic motion representation for human action recogni- tion

    Sadjad Asghari-Esfeden, Mario Sznaier, and Octavia Camps. Dynamic motion representation for human action recogni- tion. In Proceedings of the IEEE/CVF winter conference on applications of computer vision, pages 557–566, 2020. 2

  4. [3]

    Skelemotion: A new representation of skeleton joint sequences based on motion information for 3d action recognition

    Carlos Caetano, Jessica Sena, Franc ¸ois Br´emond, Jefersson A Dos Santos, and William Robson Schwartz. Skelemotion: A new representation of skeleton joint sequences based on motion information for 3d action recognition. In 2019 16th IEEE international conference on advanced video and signal based surveillance (AVSS), pages 1–8. IEEE, 2019

  5. [4]

    Jolo-gcn: mining joint-centered light-weight information for skeleton-based action recognition

    Jinmiao Cai, Nianjuan Jiang, Xiaoguang Han, Kui Jia, and Jiangbo Lu. Jolo-gcn: mining joint-centered light-weight information for skeleton-based action recognition. In Pro- ceedings of the IEEE/CVF winter conference on applications of computer vision, pages 2735–2744, 2021. 2

  6. [5]

    Meta module network for compositional visual reasoning

    Wenhu Chen, Zhe Gan, Linjie Li, Yu Cheng, William Wang, and Jingjing Liu. Meta module network for compositional visual reasoning. In Proceedings of the IEEE/CVF Winter Conference on Applications of Computer Vision, pages 655– 664, 2021. 3

  7. [6]

    Grounding physical concepts of objects and events through dynamic visual reasoning

    Zhenfang Chen, Jiayuan Mao, Jiajun Wu, Kwan-Yee Kenneth Wong, Joshua B Tenenbaum, and Chuang Gan. Grounding physical concepts of objects and events through dynamic visual reasoning. arXiv preprint arXiv:2103.16564, 2021. 3

  8. [7]

    Skeleton-based action recognition with shift graph convolutional network

    Ke Cheng, Yifan Zhang, Xiangyu He, Weihan Chen, Jian Cheng, and Hanqing Lu. Skeleton-based action recognition with shift graph convolutional network. In Proceedings of the IEEE/CVF conference on computer vision and pattern recognition, pages 183–192, 2020. 2

Show all 40 references
  1. [8]

    Potion: Pose motion representation for action recognition

    Vasileios Choutas, Philippe Weinzaepfel, J ´erˆome Revaud, and Cordelia Schmid. Potion: Pose motion representation for action recognition. In Proceedings of the IEEE conference on computer vision and pattern recognition, pages 7024–7033,

  2. [10]

    Revisiting skeleton-based action recognition

    Haodong Duan, Yue Zhao, Kai Chen, Dahua Lin, and Bo Dai. Revisiting skeleton-based action recognition. In Proceedings of the IEEE/CVF conference on computer vision and pattern recognition, pages 2969–2978, 2022. 2

  3. [11]

    Motion question answering via modular motion programs

    Mark Endo, Joy Hsu, Jiaman Li, and Jiajun Wu. Motion question answering via modular motion programs. In Inter- national Conference on Machine Learning, pages 9312–9328. PMLR, 2023. 1, 2, 3, 4, 6, 7, 8

  4. [12]

    Skeleton- based action segmentation with multi-stage spatial-temporal graph convolutional neural networks

    Benjamin Filtjens, Bart Vanrumste, and Peter Slaets. Skeleton- based action segmentation with multi-stage spatial-temporal graph convolutional neural networks. IEEE Transactions on Emerging Topics in Computing, 12(1):202–212, 2022. 2

  5. [13]

    Generating diverse and natural 3d human motions from text

    Chuan Guo, Shihao Zou, Xinxin Zuo, Sen Wang, Wei Ji, Xingyu Li, and Li Cheng. Generating diverse and natural 3d human motions from text. In Proceedings of the IEEE/CVF conference on computer vision and pattern recognition, pages 5152–5161, 2022. 2

  6. [14]

    3d concept grounding on neural fields

    Yining Hong, Yilun Du, Chunru Lin, Josh Tenenbaum, and Chuang Gan. 3d concept grounding on neural fields. Ad- vances in Neural Information Processing Systems, 35:7769– 7782, 2022. 3

  7. [15]

    Ns3d: Neuro-symbolic grounding of 3d objects and relations

    Joy Hsu, Jiayuan Mao, and Jiajun Wu. Ns3d: Neuro-symbolic grounding of 3d objects and relations. In Proceedings of the IEEE/CVF Conference on Computer Vision and Pattern Recognition, pages 2614–2623, 2023. 3

  8. [16]

    Compositional attention networks for machine reasoning

    Drew A Hudson and Christopher D Manning. Compositional attention networks for machine reasoning. arXiv preprint arXiv:1803.03067, 2018. 2, 3, 6, 8

  9. [17]

    Learning to reason iteratively and parallelly for complex visual reasoning scenarios

    Shantanu Jaiswal, Debaditya Roy, Basura Fernando, and Che- ston Tan. Learning to reason iteratively and parallelly for complex visual reasoning scenarios. In Advances in Neu- ral Information Processing Systems, pages 137965–137998. Curran Associates, Inc., 2024. 2

  10. [18]

    Clevr: A diagnostic dataset for compositional language and elementary visual reasoning

    Justin Johnson, Bharath Hariharan, Laurens Van Der Maaten, Li Fei-Fei, C Lawrence Zitnick, and Ross Girshick. Clevr: A diagnostic dataset for compositional language and elementary visual reasoning. In Proceedings of the IEEE conference on computer vision and pattern recognitio...

  11. [19]

    Self-supervised pre-training and contrastive representation learning for multiple-choice video qa

    Seonhoon Kim, Seohyeong Jeong, Eunbyul Kim, Inho Kang, and Nojun Kwak. Self-supervised pre-training and contrastive representation learning for multiple-choice video qa. In Pro- ceedings of the AAAI Conference on Artificial Intelligence , pages 13171–13179, 2021. 2

  12. [20]

    Do rules still rule? comprehensive evaluation of a rule-based question generation system

    Bernardo Leite and Henrique Lopes Cardoso. Do rules still rule? comprehensive evaluation of a rule-based question generation system. In CSEDU (2), pages 27–38, 2023. 2

  13. [21]

    Blip: Bootstrapping language-image pre-training for unified vision- language understanding and generation

    Junnan Li, Dongxu Li, Caiming Xiong, and Steven Hoi. Blip: Bootstrapping language-image pre-training for unified vision- language understanding and generation. In International conference on machine learning, pages 12888–12900. PMLR,

  14. [22]

    Vila: On pre-training for visual language models

    Ji Lin, Hongxu Yin, Wei Ping, Pavlo Molchanov, Mohammad Shoeybi, and Song Han. Vila: On pre-training for visual language models. InProceedings of the IEEE/CVF conference on computer vision and pattern recognition , pages 26689– 26699, 2024. 2

  15. [23]

    Pku-mmd: A large scale benchmark for skeleton-based human action understanding

    Chunhui Liu, Yueyu Hu, Yanghao Li, Sijie Song, and Jiaying Liu. Pku-mmd: A large scale benchmark for skeleton-based human action understanding. In Proceedings of the workshop on visual analysis in smart and connected communities, pages 1–8, 2017. 2

  16. [24]

    Roberta: A robustly optimized bert pretraining approach

    Yinhan Liu, Myle Ott, Naman Goyal, Jingfei Du, Mandar Joshi, Danqi Chen, Omer Levy, Mike Lewis, Luke Zettle- moyer, and Veselin Stoyanov. Roberta: A robustly optimized bert pretraining approach. arXiv preprint arXiv:1907.11692,

  17. [25]

    Learning to answer questions from image using convolutional neural network

    Lin Ma, Zhengdong Lu, and Hang Li. Learning to answer questions from image using convolutional neural network. In Proceedings of the AAAI conference on artificial intelligence,

  18. [26]

    Ask your neurons: A neural-based approach to answering ques- tions about images

    Mateusz Malinowski, Marcus Rohrbach, and Mario Fritz. Ask your neurons: A neural-based approach to answering ques- tions about images. In Proceedings of the IEEE international conference on computer vision, pages 1–9, 2015. 2

  19. [27]

    The neuro-symbolic concept learner: Interpreting scenes, words, and sentences from natural super- vision

    Jiayuan Mao, Chuang Gan, Pushmeet Kohli, Joshua B Tenen- baum, and Jiajun Wu. The neuro-symbolic concept learner: Interpreting scenes, words, and sentences from natural super- vision. arXiv preprint arXiv:1904.12584, 2019. 2

  20. [28]

    Spartqa:: A textual question an- swering benchmark for spatial reasoning

    Roshanak Mirzaee, Hossein Rajaby Faghihi, Qiang Ning, and Parisa Kordjmashidi. Spartqa:: A textual question an- swering benchmark for spatial reasoning. arXiv preprint arXiv:2104.05832, 2021. 2

  21. [29]

    Neuro symbolic knowledge reasoning for procedural video question answer- ing

    Thanh-Son Nguyen, Hong Yang, Tzeh Yuan Neoh, Hao Zhang, Ee Yeo Keat, and Basura Fernando. Neuro symbolic knowledge reasoning for procedural video question answer- ing. arXiv preprint arXiv:2503.14957, 2025. 3

  22. [30]

    The kit motion-language dataset

    Matthias Plappert, Christian Mandery, and Tamim Asfour. The kit motion-language dataset. Big data, 4(4):236–252,

  23. [31]

    Learning transferable visual models from natural language supervision

    Alec Radford, Jong Wook Kim, Chris Hallacy, Aditya Ramesh, Gabriel Goh, Sandhini Agarwal, Girish Sastry, Amanda Askell, Pamela Mishkin, Jack Clark, Gretchen Krueger, and Ilya Sutskever. Learning transferable visual models from natural language supervision. In International Con...

  24. [32]

    Locate: End-to-end localization of actions in 3d with transformers

    Jiankai Sun, Bolei Zhou, Michael J Black, and Arjun Chan- drasekaran. Locate: End-to-end localization of actions in 3d with transformers. arXiv preprint arXiv:2203.10719, 2022. 2

  25. [33]

    Motionclip: Exposing human motion gen- eration to clip space

    Guy Tevet, Brian Gordon, Amir Hertz, Amit H Bermano, and Daniel Cohen-Or. Motionclip: Exposing human motion gen- eration to clip space. In European Conference on Computer Vision, pages 358–374. Springer, 2022. 1, 2

  26. [34]

    Spatial temporal graph convolutional networks for skeleton-based action recog- nition

    Sijie Yan, Yuanjun Xiong, and Dahua Lin. Spatial temporal graph convolutional networks for skeleton-based action recog- nition. In Proceedings of the AAAI conference on artificial intelligence, 2018. 2

  27. [35]

    Efficient dense labelling of human activity sequences from wearables using fully convolutional networks

    Rui Yao, Guosheng Lin, Qinfeng Shi, and Damith C Ranas- inghe. Efficient dense labelling of human activity sequences from wearables using fully convolutional networks. Pattern Recognition, 78:252–266, 2018. 2

  28. [36]

    Qa-gnn: Reasoning with language models and knowledge graphs for question answering

    Michihiro Yasunaga, Hongyu Ren, Antoine Bosselut, Percy Liang, and Jure Leskovec. Qa-gnn: Reasoning with language models and knowledge graphs for question answering. arXiv preprint arXiv:2104.06378, 2021. 2

  29. [37]

    Neural-symbolic vqa: Dis- entangling reasoning from vision and language understanding

    Kexin Yi, Jiajun Wu, Chuang Gan, Antonio Torralba, Push- meet Kohli, and Josh Tenenbaum. Neural-symbolic vqa: Dis- entangling reasoning from vision and language understanding. Advances in neural information processing systems, 31, 2018. 2, 3

  30. [38]

    Exploring vision transformers for 3d human motion-language models with motion patches

    Qing Yu, Mikihiro Tanaka, and Kent Fujiwara. Exploring vision transformers for 3d human motion-language models with motion patches. In Proceedings of the IEEE/CVF Con- ference on Computer Vision and Pattern Recognition, pages 937–946, 2024. 2, 4, 5

  31. [39]

    Finemogen: Fine-grained spatio- temporal motion generation and editing

    Mingyuan Zhang, Huirong Li, Zhongang Cai, Jiawei Ren, Lei Yang, and Ziwei Liu. Finemogen: Fine-grained spatio- temporal motion generation and editing. Advances in Neural Information Processing Systems, 36:13981–13992, 2023. 2, 6

  32. [40]

    Video instruction tuning with synthetic data

    Yuanhan Zhang, Jinming Wu, Wei Li, Bo Li, Zejun Ma, Ziwei Liu, and Chunyuan Li. Video instruction tuning with synthetic data. arXiv preprint arXiv:2410.02713, 2024. 2 10 Supplementary Material for IMoRe: Implicit Program-Guided Reasoning for Human Motion Q&A Comparison with vi...

Pith tools

Reviewed August 6, 2026 · model on record in the stance chip above.