Pith. sign in

REVIEW 4 major objections 6 minor 39 references

Active Multimodal Distillation for Few-shot Action Recognition

T0 review · 4 major / 6 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 AMFIR, a framework that actively selects the more reliable of RGB or optical-flow modalities per query sample, claims substantial gains in few-shot action recognition across four benchmarks.

desk verdict Sensible multimodal fusion recipe undone by an undefined free-energy term and a mislabeled loss equation; as written, the central mechanism is not computable. read the letter →

arxiv 2506.13322 v1 pith:YDEFRLUP submitted 2025-06-16 cs.CV cs.AI

Weijia Feng , Yichen Zhu , Ruojia Zhang , Chenyang Wang , Fei Ma , Xiaobao Wang , Xiaobai Li This is my paper · ORCID
classification cs.CVcs.AI
keywords few-shotactionrecognitionactiveinferencemultimodallearningknowledgedistillationRGBandopticalflowmeta-learningadaptivefusionuncertaintyestimation
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 claims that few-shot action recognition benefits from deciding, per query video, which of two modalities—RGB frames or optical flow—is more reliable, rather than fusing them with fixed weights. It introduces AMFIR, a framework whose Active Sample Inference module scores each modality by the maximum value of its class-posterior distribution and uses that score to group samples, to weight a mutual distillation loss that strengthens the weaker modality, and to set adaptive fusion weights at test time. On SSv2, HMDB51, UCF101, and Kinetics-400, the method reports accuracy that exceeds prior unimodal and multimodal baselines, for example 70.6% and 92.3% on SSv2 and 94.9% and 99.1% on UCF101 for 1-shot and 5-shot. A sympathetic reading is that uncertainty-driven modality selection, not extra data, produces the gain. The components are simple enough that they can be ablated individually, and the paper does so.

What carries the argument

The load-bearing mechanism is the Active Sample Inference (ASI) module, built on the absolute certainty $c_i^m = \max_k p_i^m(k \mid x_i^m)$, where $p_i^m$ is the softmax over class prototypes for modality $m$ and query sample $i$. This scalar is computed separately for RGB and optical flow and does three jobs: it routes the query into an RGB-dominant or flow-dominant group, it weights the KL-divergence distillation loss $L^{m\to n}$ that lets the reliable modality teach the unreliable one, and it produces the adaptive fusion weight $\alpha_i^m = c_i^m/(c_i^r + c_i^f)$ for the final prediction. The paper's claim is that this confidence measure, taken from task-specific posterior distributions, captures which modality carries discriminative information for that particular sample in that particular episode.

What would settle it

Construct a test split where the RGB stream is made confidently wrong while optical flow remains correct (for example, by altering textures without changing motion), then check whether AMFIR still routes high-confidence RGB samples to the RGB-dominant group and follows the RGB branch in the final prediction.

Watch

Extended reading notes

Core claim

On the paper's own terms, the central discovery is that a single per-sample scalar—the maximum of a modality-specific softmax posterior over class prototypes—can serve as the organizing signal for multimodal few-shot learning. With only RGB and optical-flow streams, the framework uses this scalar to infer which modality is dominant for each query sample, transfers query-to-prototype similarity knowledge from the dominant to the weaker modality through a KL-divergence distillation, and fuses the final prediction using the same scalar as an adaptive weight. The authors report that the resulting AMFIR framework outperforms existing few-shot action recognition methods on all four tested benchmarks, with the largest reported gains on UCF101 and SSv2.

Load-bearing premise

The framework assumes that when the model is very sure of one modality's answer, that modality is the right one to trust; a confidently wrong guess would mislead the grouping, the distillation, and the final fusion.

Editorial extensions

If this is right

  • On UCF101, the reported 99.1% 5-shot accuracy would leave that benchmark near saturation, so further progress would have to be demonstrated on harder datasets.
  • Because the teacher assignment is made per sample, a single episode can contain both RGB-teacher and flow-teacher samples, which a fixed teacher-student design cannot express.
  • The method adds no extra data or labels beyond the standard support and query episodes, so the reported gains, if real, come from the learning procedure rather than from additional supervision.
  • The three components—grouping, mutual distillation, and adaptive fusion—are modular enough to be inserted into an existing unimodal few-shot recognizer.

Reading between the lines

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

  • Inference: because the paper never gives an operational definition for the free energy $F_i^m$ used in the grouping equation, the active-inference grouping reduces in practice to a comparison of per-modality confidences; replacing the grouping rule with a simple threshold on $c_i^m$ would test whether the theoretical framing adds predictive power.
  • Inference: the same confidence-as-reliability proxy should transfer to other modality pairs, such as audio with video or skeleton with RGB, where posterior calibration patterns differ; a cross-modal test would reveal whether the method is truly modality-agnostic.
  • Inference: because the fusion weight is a per-sample normalized ratio of two confidences, the method implicitly assumes that the two modality streams are calibrated on comparable scales; per-dataset calibration curves would show whether the reported gains survive when the scales are mismatched.
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 / 6 minor

Summary. The paper proposes AMFIR, a multimodal few-shot action recognition framework that combines RGB and optical flow modalities. The method has three main components: an Active Sample Inference (ASI) module that groups query samples into RGB-dominant and flow-dominant groups using a free-energy criterion; an Active Mutual Distillation (AMD) module that transfers knowledge from a reliable modality to an unreliable one; and an Adaptive Multimodal Inference (AMI) module that fuses modality-specific posteriors with confidence-based weights at meta-test time. Experiments on Kinetics-400, SSv2, HMDB51, and UCF101 report large gains over prior unimodal and multimodal baselines, with the authors attributing the gains to the ASI module's active modality selection.

Significance. The problem of multimodal few-shot action recognition is timely, and the paper addresses it with a plausible high-level pipeline: cross-modal distillation, confidence-based fusion, and per-sample modality selection. The experimental scope is broad, covering four benchmarks, and the ablation study in Table 2 shows that each component contributes to the final score. However, the central claimed mechanism is not operational as written: Eq. (3) depends on an undefined free-energy term, and the training objective in Eq. (10) is mathematically the negative entropy rather than cross-entropy, which would push predictions toward uniform distributions. These are load-bearing issues, not presentation problems. The paper also does not provide standard deviations or a confidence-only baseline, so the reported gains cannot currently be attributed to active inference rather than to simple confidence weighting. If the technical issues were corrected and the method re-evaluated, the framework could be a useful contribution, but the manuscript in its current form is not ready for publication.

major comments (4)
  1. [§3.3, Eq. (3)] The free-energy term F^m_i in Eq. (3) is never defined. The text says it is 'free energy derived from Eq. (1)', but Eq. (1) is the POMDP variational free energy and no formula maps the modality-specific posterior p^m_i of Eq. (2) to a scalar F^m_i for a query sample. Consequently, Eq. (3) cannot be evaluated, and since Section 3.4 applies distillation only to samples in G^r and G^f, the undefined quantity controls the teacher-student assignment and the training subset. Figure 2 plots 'free energy' without defining the axis or giving the computation. If F^m_i is meant to be c^m_i, the paper should state this explicitly and justify max-posterior as a reliability measure; otherwise, a concrete definition is required.
  2. [§3.6, Eq. (10)] Eq. (10) is not a cross-entropy loss. As written, L^m_ce = \sum_i \sum_k p^m_i(k) \log p^m_i(k) is the negative Shannon entropy of the predicted posterior, and minimizing it drives p^m_i toward a uniform distribution over classes, which is the opposite of classification. The correct cross-entropy should involve the ground-truth label, e.g., -\sum_i \log p^m_i(y_i). Because Eq. (9) and Eq. (11) optimize this objective during meta-training, this is a load-bearing mathematical error that undermines the reported results.
  3. [§3.3–§3.5] The method's 'active inference' mechanism is confounded with simple confidence weighting. Eq. (5) defines reliability c^m_i as the maximum of the modality-specific posterior, and Eq. (8) uses the same c^m_i to set the adaptive fusion weights. If F^m_i in Eq. (3) is also computed from the same posterior, then the same self-confidence measure determines the grouping, the distillation weights, and the final fusion. The paper needs an ablation that replaces the 'active inference' grouping with a trivial confidence-based grouping (e.g., assign each sample to the modality with higher c^m_i) to show that the claimed active-inference mechanism, rather than confidence weighting alone, produces the reported gains.
  4. [§4.3, Table 1] The experimental section reports only point estimates without standard deviations, confidence intervals, or statistical significance tests. Given that the reported margins over some baselines are small (e.g., UCF101 5-shot: 99.1 vs 99.0), the absence of variance estimates makes the strength of the comparison unclear. The paper should report mean and standard deviation over multiple random episodes or seeds, and should state the number of meta-test tasks used.
minor comments (6)
  1. [Throughout] The framework name is inconsistent: 'AMFIR' appears in the title and abstract, while 'AFMIR' appears in Table 1 and elsewhere, and the method name in Section 3 is also given as 'AFMIR' in one place. Please unify the nomenclature.
  2. [§3.6, Eq. (9)] The notation L^{m→L^n} in Eq. (9) and the text immediately after it appears to be a typo for L^{m→n}; please fix the superscript notation.
  3. [§4.2, Figure 2] Figure 2 plots 'free energy' but provides no axis labels, no units, and no equation for the plotted quantity; please define the vertical axis and state how the quantity is computed.
  4. [§4.4] The phrase 'reduce the thermal accuracy of Kinetics 5 by 7.06%' is unclear; it should likely read 'reduce the accuracy on Kinetics-400 5-shot by 7.06%.'
  5. [§4.1] Please provide details on the optical flow generation algorithm, backbone pretraining protocols, episode construction (number of tasks, class sampling), and whether the reported numbers are averaged over multiple runs; these details are needed to reproduce the results.
  6. [§5] The conclusion claims robustness to 'sensor noise, motion ambiguity, and extreme data scarcity', but none of the experiments in Section 4 introduce sensor noise or corrupted modalities; please either add such experiments or remove the claim.

Circularity Check

0 steps flagged · score 0.0 of 10

No circularity found; the active-inference free energy is under-specified, but no derived quantity reduces to its inputs by construction.

full rationale

The paper's central claimed mechanism, Active Sample Inference (ASI), selects modalities via Eq. (3) using a free energy F^m_i that is said to be 'derived from Eq. (1)', but the paper never defines how the modality-specific posterior of Eq. (2) maps to the scalar F^m_i. This makes the selection rule non-operational and the reported free-energy plots in Figure 2 unverifiable. This is a reproducibility and support problem, not circularity: no equation asserts that F^m_i equals, or is fitted to, the quantity that is later used as the prediction or as the evaluation target. The confidence c^m_i in Eq. (5) is the maximum of the modality-specific posterior. It is used both as a distillation weight in Eq. (6) and as the adaptive fusion weight in Eq. (8). Using the model's own confidence both to define reliability and to weight fusion is a heuristic design choice, but it is not circular: the confidence is computed from distances between query and prototype features, and the final fused prediction in Eq. (7) is a separate softmax over confidence-weighted distances. The final prediction is not fed back into the definition of c^m_i, and no fitted parameter is renamed as a prediction. The method is therefore a confidence-weighted multimodal distillation and fusion scheme whose reported gains may or may not hold, but the derivation chain does not collapse into its inputs. There is no load-bearing self-citation: the self-citations in the introduction and related work concern edge services, AIGC, and caching, and they are not used to justify the active-inference formulation or the distillation/fusion equations. The active-inference background cites external work (Friston, Tschantz et al., etc.), and the AFMAR baseline is by different authors. The similarity of AMFIR's component names to AFMAR's components could raise a novelty concern, but the paper does not reduce its equations to AFMAR's equations, so that is not a circularity finding under the required standard. Overall, the paper is not circular as written. Its main weakness is that the free-energy quantity controlling the central ASI grouping is never operationally defined, which is a completeness/reproducibility issue rather than a self-referential derivation.

Assumptions & free parameters 2 free parameters · 4 assumptions · 0 invented entities

The method introduces no new physical or conceptual entities; the modules are recombinations of standard components. The main unaccounted-for ingredients are the undefined free energy and the confidence-based reliability proxy.

free parameters (2)
  • lambda (distillation balance) = 1.0
    Set uniformly to 1.0 across all benchmarks in Section 4.1; no sensitivity analysis is reported.
  • learning rate gamma = 1e-3
    Chosen for SGD in Section 4.1; no schedule or sensitivity analysis is reported.
assumptions (4)
  • domain assumption Softmax over Euclidean distances yields a meaningful posterior over classes for each modality (Eq 2).
    The entire reliability estimation and fusion scheme depends on treating these softmax outputs as calibrated probabilities; this is not validated.
  • domain assumption The maximum softmax probability (absolute certainty) is a valid proxy for modality reliability.
    Eqs 5, 6, and 8 all rely on this; no external evidence or calibration study is provided.
  • ad hoc to paper The Free Energy Principle / active inference framework applies to modality selection as described.
    Eq (1) is the generic variational free energy, but it is never instantiated for the modality-selection problem; the paper asserts lower free energy implies reliability without defining the quantities.
  • domain assumption Pre-trained backbones (ResNet-50 for RGB, I3D on Charades for flow) transfer to the few-shot benchmarks.
    This is standard practice, but the paper does not report the effect of backbone choice on the comparison with AFMAR.

how reviews work

0 comments
Cite this review

Pith. "Pith review of Active Multimodal Distillation for Few-shot Action Recognition." pith.science (2026). https://pith.science/paper/YDEFRLUP

@misc{pith2026250613322,
  author       = {Pith},
  title        = {Pith review of: Active Multimodal Distillation for Few-shot Action Recognition},
  year         = {2026},
  howpublished = {\url{https://pith.science/paper/YDEFRLUP}},
  note         = {Machine review of arXiv:2506.13322}
}
read the original abstract

Owing to its rapid progress and broad application prospects, few-shot action recognition has attracted considerable interest. However, current methods are predominantly based on limited single-modal data, which does not fully exploit the potential of multimodal information. This paper presents a novel framework that actively identifies reliable modalities for each sample using task-specific contextual cues, thus significantly improving recognition performance. Our framework integrates an Active Sample Inference (ASI) module, which utilizes active inference to predict reliable modalities based on posterior distributions and subsequently organizes them accordingly. Unlike reinforcement learning, active inference replaces rewards with evidence-based preferences, making more stable predictions. Additionally, we introduce an active mutual distillation module that enhances the representation learning of less reliable modalities by transferring knowledge from more reliable ones. Adaptive multimodal inference is employed during the meta-test to assign higher weights to reliable modalities. Extensive experiments across multiple benchmarks demonstrate that our method significantly outperforms existing approaches.

Figures

Figures reproduced from arXiv: 2506.13322 by the authors.

Figure 1
Figure 1. Illustration of our proposed framework in the 3-way 3-shot setting. [PITH_FULL_IMAGE:figures/full_fig_p003_1.png] view at source ↗
Figure 2
Figure 2. The change of free energy. 4.2 Free Energy The experimental results show that, on the SSv2 and Kinetics￾400 datasets, there is a significant fluctuation in the initial free [PITH_FULL_IMAGE:figures/full_fig_p005_2.png] view at source ↗
Figure 4
Figure 4. Comparison with conventional distillation strategies in a 5- [PITH_FULL_IMAGE:figures/full_fig_p007_4.png] view at source ↗
Figures from the paper (1 more)
Figure 3
Figure 3. Figure 3: Comparison results with different numbers of support [PITH_FULL_IMAGE:figures/full_fig_p007_3.png]

Discussion (0). Continue with ORCID to comment.

Reference graph

Works this paper leans on

39 extracted references · 33 canonical work pages

  1. [1]

    Improved few- shot visual classification

    [Bateniet al., 2020 ] Peyman Bateni, Raghav Goyal, Vaden Masrani, Frank Wood, and Leonid Sigal. Improved few- shot visual classification. InProceedings of the IEEE/CVF conference on computer vision and pattern recognition, pages 14493–14502,

    work page 2020

  2. [3]

    Mars: Motion- augmented rgb stream for action recognition

    [Crastoet al., 2019 ] Nieves Crasto, Philippe Weinzaepfel, Karteek Alahari, and Cordelia Schmid. Mars: Motion- augmented rgb stream for action recognition. InProceed- ings of the IEEE/CVF conference on computer vision and pattern recognition, pages 7882–7891,

    work page 2019

  3. [7]

    Exploring collaborative diffusion model inferring for aigc-enabled edge services.IEEE Trans- actions on Cognitive Communications and Networking,

    [Fenget al., 2024 ] Weijia Feng, Ruojia Zhang, Yichen Zhu, Chenyang Wang, Chuan Sun, Xiaoqiang Zhu, Xiang Li, and Tarik Taleb. Exploring collaborative diffusion model inferring for aigc-enabled edge services.IEEE Trans- actions on Cognitive Communications and Networking,

    work page 2024

  4. [10]

    Modality distillation with multiple stream networks for action recognition

    [Garciaet al., 2018 ] Nuno C Garcia, Pietro Morerio, and Vittorio Murino. Modality distillation with multiple stream networks for action recognition. InProceedings of the European Conference on Computer Vision (ECCV), pages 103–118,

    work page 2018

  5. [11]

    The” something something” video database for learning and evaluating vi- sual common sense

    [Goyalet al., 2017 ] Raghav Goyal, Samira Ebrahimi Kahou, Vincent Michalski, Joanna Materzynska, Susanne West- phal, Heuna Kim, Valentin Haenel, Ingo Fruend, Peter Yianilos, Moritz Mueller-Freitag, et al. The” something something” video database for learning and evaluating vi- sual common sense. InProceedings of the IEEE interna- tional conference on comp...

    work page 2017

  6. [15]

    Pro- togan: Towards few shot learning for action recognition

    [Kumar Dwivediet al., 2019 ] Sai Kumar Dwivedi, Vikram Gupta, Rahul Mitra, Shuaib Ahmed, and Arjun Jain. Pro- togan: Towards few shot learning for action recognition. InProceedings of the IEEE/CVF international conference on computer vision workshops, pages 0–0,

    work page 2019

  7. [16]

    Crnet: Cross-reference networks for few-shot segmentation

    [Liuet al., 2020 ] Weide Liu, Chi Zhang, Guosheng Lin, and Fayao Liu. Crnet: Cross-reference networks for few-shot segmentation. InProceedings of the IEEE/CVF conference on computer vision and pattern recognition, pages 4165– 4173,

    work page 2020

  8. [17]

    A review of human emotion syn- thesis based on generative technology.arXiv preprint arXiv:2412.07116,

    [Maet al., 2024a ] Fei Ma, Yukan Li, Yifan Xie, Ying He, Yi Zhang, Hongwei Ren, Zhou Liu, Wei Yao, Fuji Ren, Fei Richard Yu, et al. A review of human emotion syn- thesis based on generative technology.arXiv preprint arXiv:2412.07116,

    arXiv

Show all 39 references
  1. [18]

    Self- paced adversarial training for multimodal few-shot learn- ing

    [Pahdeet al., 2019 ] Frederik Pahde, Oleksiy Ostapenko, Patrick J¨a Hnichen, Tassilo Klein, and Moin Nabi. Self- paced adversarial training for multimodal few-shot learn- ing. In2019 IEEE Winter Conference on Applications of Computer Vision (WACV), pages 218–226. IEEE,

    2019

  2. [19]

    Relational knowledge distillation

    [Parket al., 2019 ] Wonpyo Park, Dongju Kim, Yan Lu, and Minsu Cho. Relational knowledge distillation. InProceed- ings of the IEEE/CVF conference on computer vision and pattern recognition, pages 3967–3976,

    2019

  3. [20]

    Two-stream collaborative learning with spatial- temporal attention for video classification.IEEE Trans- actions on Circuits and Systems for Video Technology, 29(3):773–786,

    [Penget al., 2018 ] Yuxin Peng, Yunzhen Zhao, and Junchao Zhang. Two-stream collaborative learning with spatial- temporal attention for video classification.IEEE Trans- actions on Circuits and Systems for Video Technology, 29(3):773–786,

    2018

  4. [21]

    Temporal- relational crosstransformers for few-shot action recogni- tion

    [Perrettet al., 2021 ] Toby Perrett, Alessandro Masullo, Tilo Burghardt, Majid Mirmehdi, and Dima Damen. Temporal- relational crosstransformers for few-shot action recogni- tion. InProceedings of the IEEE/CVF conference on computer vision and pattern recognition, pages 475–484,

    2021

  5. [22]

    Generating mean- ing: Active inference and generative ai.PsyArXiv

    [Pezzuloet al., 2023 ] Giovanni Pezzulo, Thomas Parr, Paul Cisek, Andy Clark, and Karl Friston. Generating mean- ing: Active inference and generative ai.PsyArXiv. June, 8,

    2023

  6. [24]

    Spatio-temporal relation modeling for few-shot action recognition

    [Thatipelliet al., 2022 ] Anirudh Thatipelli, Sanath Narayan, Salman Khan, Rao Muhammad Anwer, Fahad Shahbaz Khan, and Bernard Ghanem. Spatio-temporal relation modeling for few-shot action recognition. InProceedings of the IEEE/CVF Conference on Computer Vision and Pat- tern R...

    2022

  7. [26]

    Multimodal few-shot learning with frozen language models.Advances in Neural Information Processing Sys- tems, 34:200–212,

    [Tsimpoukelliet al., 2021 ] Maria Tsimpoukelli, Jacob L Menick, Serkan Cabi, SM Eslami, Oriol Vinyals, and Felix Hill. Multimodal few-shot learning with frozen language models.Advances in Neural Information Processing Sys- tems, 34:200–212,

    2021

  8. [27]

    Similarity-preserving knowledge distillation

    [Tung and Mori, 2019] Frederick Tung and Greg Mori. Similarity-preserving knowledge distillation. InProceed- ings of the IEEE/CVF international conference on com- puter vision, pages 1365–1374,

    2019

  9. [28]

    Action recognition with trajectory-pooled deep- convolutional descriptors

    [Wanget al., 2015 ] Limin Wang, Yu Qiao, and Xiaoou Tang. Action recognition with trajectory-pooled deep- convolutional descriptors. InProceedings of the IEEE con- ference on computer vision and pattern recognition, pages 4305–4314,

    2015

  10. [29]

    Hybrid relation guided set matching for few-shot action recognition

    [Wanget al., 2022 ] Xiang Wang, Shiwei Zhang, Zhiwu Qing, Mingqian Tang, Zhengrong Zuo, Changxin Gao, Rong Jin, and Nong Sang. Hybrid relation guided set matching for few-shot action recognition. InProceedings of the IEEE/CVF conference on computer vision and pat- tern recogni...

    2022

  11. [30]

    Ai-tfnet: Active in- ference transfer convolutional fusion network for hyper- spectral image classification.Remote Sensing, 15(5):1292,

    [Wanget al., 2023 ] Jianing Wang, Linhao Li, Yichen Liu, Jinyu Hu, Xiao Xiao, and Bo Liu. Ai-tfnet: Active in- ference transfer convolutional fusion network for hyper- spectral image classification.Remote Sensing, 15(5):1292,

    2023

  12. [31]

    Elevating knowledge-enhanced entity and relation- ship understanding for sarcasm detection.IEEE Transac- tions on Knowledge and Data Engineering,

    [Wanget al., 2025 ] Xiaobao Wang, Yujing Wang, Dongxiao He, Zhe Yu, Yawen Li, Longbiao Wang, Jianwu Dang, and Di Jin. Elevating knowledge-enhanced entity and relation- ship understanding for sarcasm detection.IEEE Transac- tions on Knowledge and Data Engineering,

    2025

  13. [32]

    Active exploration of multimodal complementarity for few-shot action recog- nition

    [Wanyanet al., 2023 ] Yuyang Wanyan, Xiaoshan Yang, Chaofan Chen, and Changsheng Xu. Active exploration of multimodal complementarity for few-shot action recog- nition. InProceedings of the IEEE/CVF Conference on Computer Vision and Pattern Recognition, pages 6492– 6502,

    2023

  14. [33]

    Motion-modulated tem- poral fragment alignment network for few-shot action recognition

    [Wuet al., 2022 ] Jiamin Wu, Tianzhu Zhang, Zhe Zhang, Feng Wu, and Yongdong Zhang. Motion-modulated tem- poral fragment alignment network for few-shot action recognition. InProceedings of the IEEE/CVF Conference on Computer Vision and Pattern Recognition, pages 9151– 9160,

    2022

  15. [34]

    Enriching multimodal sentiment analysis through textual emotional descriptions of visual-audio content

    [Wuet al., 2025 ] Sheng Wu, Dongxiao He, Xiaobao Wang, Longbiao Wang, and Jianwu Dang. Enriching multimodal sentiment analysis through textual emotional descriptions of visual-audio content. InProceedings of the AAAI Con- ference on Artificial Intelligence, volume 39, pages 16...

    2025

  16. [35]

    Human mo- tion video generation: A survey.Authorea Preprints,

    [Xueet al., 2024 ] Haiwei Xue, Xiangyang Luo, Zhanghao Hu, Xin Zhang, Xunzhi Xiang, Yuqin Dai, Jianzhuang Liu, Zhensong Zhang, Minglei Li, Jian Yang, et al. Human mo- tion video generation: A survey.Authorea Preprints,

    2024

  17. [36]

    Aim: Adapting image models for efficient video action recognition.arXiv preprint arXiv:2302.03024,

    [Yanget al., 2023 ] Taojiannan Yang, Yi Zhu, Yusheng Xie, Aston Zhang, Chen Chen, and Mu Li. Aim: Adapting image models for efficient video action recognition.arXiv preprint arXiv:2302.03024,

    2023 arXiv

  18. [37]

    Few- shot action recognition with permutation-invariant atten- tion

    [Zhanget al., 2020 ] Hongguang Zhang, Li Zhang, Xiaojuan Qi, Hongdong Li, Philip HS Torr, and Piotr Koniusz. Few- shot action recognition with permutation-invariant atten- tion. InComputer Vision–ECCV 2020: 16th European Conference, Glasgow, UK, August 23–28, 2020, Proceed- in...

    2020

  19. [38]

    Premier-taco is a few-shot policy learner: Pretraining multitask representation via temporal action- driven contrastive loss

    [Zhenget al., 2024 ] Ruijie Zheng, Yongyuan Liang, Xiyao Wang, Shuang Ma, Hal Daum ´e III, Huazhe Xu, John Langford, Praveen Palanisamy, Kalyan Shankar Basu, and Furong Huang. Premier-taco is a few-shot policy learner: Pretraining multitask representation via temporal action- ...

    2024

  20. [39]

    Compound memory networks for few-shot video classification

    [Zhu and Yang, 2018] Linchao Zhu and Yi Yang. Compound memory networks for few-shot video classification. In Proceedings of the European conference on computer vi- sion (ECCV), pages 751–766, 2018

    2018

  21. [2015]

    The kinetics human action video dataset

    [Kayet al., 2017 ] Will Kay, Joao Carreira, Karen Si- monyan, Brian Zhang, Chloe Hillier, Sudheendra Vijaya- narasimhan, Fabio Viola, Tim Green, Trevor Back, Paul Natsev, et al. The kinetics human action video dataset. arXiv preprint arXiv:1705.06950,

    2017 arXiv

  22. [2016]

    Distilling the knowledge in a neural network.arXiv preprint arXiv:1503.02531,

    [Hinton, 2015] Geoffrey Hinton. Distilling the knowledge in a neural network.arXiv preprint arXiv:1503.02531,

    2015 arXiv

  23. [2017]

    Cross modal distillation for supervision transfer

    [Guptaet al., 2016 ] Saurabh Gupta, Judy Hoffman, and Ji- tendra Malik. Cross modal distillation for supervision transfer. InProceedings of the IEEE conference on com- puter vision and pattern recognition, pages 2827–2836,

    2016

  24. [2018]

    Few-shot object detection with attention- rpn and multi-relation detector

    [Fanet al., 2020 ] Qi Fan, Wei Zhuo, Chi-Keung Tang, and Yu-Wing Tai. Few-shot object detection with attention- rpn and multi-relation detector. InProceedings of the IEEE/CVF conference on computer vision and pattern recognition, pages 4013–4022,

    2020

  25. [2019]

    Learning an augmented rgb representation with cross-modal knowledge distillation for action detection

    [Daiet al., 2021 ] Rui Dai, Srijan Das, and Franc ¸ois Bre- mond. Learning an augmented rgb representation with cross-modal knowledge distillation for action detection. InProceedings of the IEEE/CVF International Conference on Computer Vision, pages 13053–13064,

    2021

  26. [2020]

    Few-shot video classification via temporal alignment

    [Caoet al., 2020 ] Kaidi Cao, Jingwei Ji, Zhangjie Cao, Chien-Yi Chang, and Juan Carlos Niebles. Few-shot video classification via temporal alignment. InProceedings of the IEEE/CVF conference on computer vision and pattern recognition, pages 10618–10627,

    2020

  27. [2021]

    Fast parameter adaptation for few- shot image captioning and visual question answering

    [Donget al., 2018 ] Xuanyi Dong, Linchao Zhu, De Zhang, Yi Yang, and Fei Wu. Fast parameter adaptation for few- shot image captioning and visual question answering. In Proceedings of the 26th ACM international conference on Multimedia, pages 54–62,

    2018

  28. [2022]

    Reinforce- ment learning through active inference.arXiv preprint arXiv:2002.12636,

    [Tschantzet al., 2020 ] Alexander Tschantz, Beren Millidge, Anil K Seth, and Christopher L Buckley. Reinforce- ment learning through active inference.arXiv preprint arXiv:2002.12636,

    2020 arXiv

  29. [2023]

    Adaptive stream processing on edge devices through active inference.arXiv preprint arXiv:2409.17937,

    [Sedlaket al., 2024 ] Boris Sedlak, Victor Casamayor Pu- jol, Andrea Morichetta, Praveen Kumar Donta, and Schahram Dustdar. Adaptive stream processing on edge devices through active inference.arXiv preprint arXiv:2409.17937,

    2024 arXiv

  30. [2024]

    Federated deep reinforcement learning for multimodal content caching in edge-cloud networks.IEEE Transac- tions on Network Science and Engineering,

    [Fenget al., 2025 ] Weijia Feng, Xinyu Zuo, Ruojia Zhang, Yichen Zhu, Chenyang Wang, Jia Guo, and Chuan Sun. Federated deep reinforcement learning for multimodal content caching in edge-cloud networks.IEEE Transac- tions on Network Science and Engineering,

    2025

  31. [2025]

    Depth guided adaptive meta- fusion network for few-shot video recognition

    [Fuet al., 2020 ] Yuqian Fu, Li Zhang, Junke Wang, Yanwei Fu, and Yu-Gang Jiang. Depth guided adaptive meta- fusion network for few-shot video recognition. InPro- ceedings of the 28th ACM International Conference on Multimedia, pages 1142–1151,

    2020

Pith tools

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