Pith. sign in

REVIEW 4 major objections 6 minor 52 references

Listening to the Unspoken: Exploring "365" Aspects of Multimodal Interview Performance Assessment

T0 review · 4 major / 6 minor · reviewed 2026-08-06 · deepseek-v4-flash

Pith's one-line read This paper claims that a shared-compression MLP fusing video, audio, and text, plus a 32-head ensemble averaged over six answers, achieves an MSE of 0.1824 and first place in the AVI Challenge 2025.

desk verdict A challenge-winning system with a clean architecture and a credible engineering win, but the headline MSE is the best of four test-set selections and the validation tables don't add up. read the letter →

arxiv 2507.22676 v3 pith:XSPE5RE5 submitted 2025-07-30 cs.CL cs.MM

classification cs.CLcs.MM
keywords multimodalinterviewassessmentsharedcompressionmultilayerperceptronfusionmulti-labelregressionensemblelearningAVIChallenge2025
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 tries to show that interview performance can be scored automatically at a level that won the AVI Challenge 2025, by fusing three modalities (video, audio, and text) for each of six answers and predicting five job-relevant dimensions. The central move is to reinterpret a multilayer perceptron as an attention-like module and share its 'value' basis vectors across modalities, so all inputs are grounded in one low-dimensional space. On top of that, 32 regression heads per answer are averaged, and then the six answers are averaged, giving a multi-dimensional mean squared error of 0.1824 on the test set. A fair reader would come away with a concrete design that is simple, fully specified, and reproducible, and that outperforms the other six listed challenge systems.

What carries the argument

The load-bearing mechanism is the Multimodal Shared Compression Multilayer Perceptron (MSCMLP), a variant of the classic MLP treated as an attention-like module. The first linear layer of each modality, followed by GeLU, produces activation scores; the second linear layer is a single set of learnable basis vectors shared by all three modalities, and each modality's output is the weighted sum of those basis vectors. Because the output dimension of the shared basis is set below the input dimensions, the fusion both aligns and compresses the 1152-dimensional video, 768-dimensional audio, and 4096-dimensional text embeddings into a unified space. The second mechanism is the two-level ensemble: each of six responses is scored by 32 parallel regression heads, the heads are averaged per response, and the response averages are mean-pooled to produce the final five scores.

What would settle it

On the same AVI 2025 test split, replace the equal-weight average in Eq. (11) with weights learned per response, per dimension, or per head; if any such weighted variant lowers the multi-dimensional average MSE below 0.1824, the paper's pooling strategy is not the optimal use of its own representations, and its headline result would be tied to that choice.

Watch

Extended reading notes

Core claim

The paper's central claim is that its proposed framework—modality-specific extractors (SigLIP2 for video, emotion2vec-plus-seed for audio, SFR-Embedding-Mistral for text), a Shared Compression Multilayer Perceptron for fusion, and a two-level ensemble of 32 regression heads with response-level mean pooling—achieves state-of-the-art performance on the AVI 2025 interview assessment test set, with an average MSE of 0.1824 across integrity, collegiality, social versatility, development orientation, and overall hireability. The authors interpret the MLP's second linear layer as a set of shared basis vectors: each modality computes activation scores against its own first-layer weights, and those scores pick out a weighted combination of the same basis, producing a compressed common embedding. They report that the combination of K-fold cross-validation, max-pooling for both video and audio features, 32 ensemble heads, and mean-pooling across the six responses yields their best test result, and that their validation-set performance is superior to the AVI baseline on four of the five dimensions.

Load-bearing premise

The final prediction is a simple average of every answer and every scoring head, so the whole result depends on those answers and heads being equally informative for each of the five dimensions.

Editorial extensions

If this is right

  • If the 0.1824 test MSE holds, a competitive multimodal interview scorer can be built from frozen pretrained encoders plus a single small fusion MLP, with no fine-tuning of the encoders and no cross-modal attention.
  • The shared-basis design implies one learned linear layer is enough to project heterogeneous embeddings into a common lower-dimensional space where a prediction head can read them, which transfers to other multi-modal regression settings.
  • The two-level averaging implies that repetition — multiple answers per candidate and multiple heads per answer — is itself treated as an accuracy booster, exploiting variance reduction rather than more complex modeling.
  • Per-dimension results on the validation set indicate the largest gains over the AVI baseline are in collegiality, social versatility, development orientation, and overall hireability, while the baseline retains the edge on integrity.

Reading between the lines

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

  • Because Eq. 11 weights every response and head equally, a likely untested improvement is to learn per-question or per-dimension weights; if the six questions differ in how much they reveal about, say, integrity, that pooling is discarding signal the same architecture could use.
  • The MSCMLP attention analogy suggests a direct comparison with low-rank bilinear pooling or cross-attention fusion on the same features; the paper's result implies the benefit may come from compression and a shared basis rather than explicit pairwise interaction, but the paper does not run that comparison.
  • The test-set leaderboard gap to the second-place system is 0.0027 MSE; a resampling or repeated-split analysis would show whether that margin is stable, something the paper does not report.
  • A practical extension would be to apply the same shared-basis fusion to asynchronous video interview screening at scale, where the six responses are fixed and the five dimensions are known to raters — the framework's simplicity keeps deployment costs low.
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 / 6 minor

Summary. The paper proposes a multimodal framework for asynchronous video interview performance assessment, combining three modalities (video, audio, text), six interview responses per candidate, and five target evaluation dimensions. Modality-specific pretrained encoders produce features that are fused by a proposed Shared Compression Multilayer Perceptron (MSCMLP), followed by an ensemble of 32 regression heads whose outputs are averaged across responses. The authors report a test MSE of 0.1824, claiming first place in the AVI Challenge 2025 track, and provide an ablation study over K-fold cross-validation, number of ensemble heads, and modality pooling strategies.

Significance. If the reported result is reliable, the paper is a practically useful challenge-report contribution: the MSCMLP fusion idea is simple and clearly motivated, the equations describing it are internally consistent, and the system reportedly outperforms several competing teams on a shared benchmark. The availability of code is a concrete reproducibility asset. However, the headline claim rests on a single test-set number that was selected over four pooling configurations by test performance, with no variance estimates or significance tests, and several internal numerical inconsistencies in the evaluation tables. The significance is therefore conditional on the authors tightening the experimental reporting.

major comments (4)
  1. [§4.3, Tables 2 and 3] The final configuration (SigLIP2 Max pooling, emotion2vec Max pooling) is selected because it achieved the best test MSE (0.1824), whereas the validation MSE is better for the Mean/Max configuration (0.1673). Thus the reported 0.1824 is the extremum of a four-way search on the test set, not an unbiased estimate of the pipeline's test performance. Moreover, the margin over the runner-up in Table 3 (0.00270) is smaller than the spread among the four pooling configurations in Table 2 (about 0.00607), and no repeated-seed variance, confidence interval, or significance test is provided. To support the first-place claim, please report the selection procedure, give multi-seed means and standard deviations, and state whether the test labels were accessed only once.
  2. [§4.3, Table 4 vs. Tables 1 and 2] The per-dimension validation MSEs for HFUT-VisionXL in Table 4 (0.1678, 0.1619, 0.1567, 0.1403, 0.1606) average to 0.15746, which does not match the validation MSE of 0.1681 reported for the final configuration in Table 2. Additionally, the claimed test improvement from K-fold cross-validation in Table 1 is 0.009598, but the numbers in the table imply 0.193083 − 0.184688 = 0.008395. These arithmetic inconsistencies must be resolved before the evaluation can be considered reliable.
  3. [§4.2 and §4.3] The experimental setup assigns dropout rates to a 'feature adapter,' a 'pure text processing module,' and a 'temporal control module,' but none of these components is defined in Section 3 or shown in Figure 2. The K-fold ablation in Section 4.3 never states the value of K, nor how cross-validation is integrated into training and final prediction. Without these details, the method and the claimed benefit of K-fold cross-validation are not reproducible.
  4. [Eq. (11) and §4.4] The final prediction is an unweighted mean over the six responses and 32 heads. The paper provides no analysis or ablation on whether all responses and all heads are equally informative for each of the five dimensions. If some interview questions carry more signal, or if response order matters, this equal-weight pooling discards information and the reported MSE depends on that choice. Please add an ablation or at least a discussion of per-response and per-head contributions to justify the pooling strategy.
minor comments (6)
  1. [§4.2] The optimizer is written as 'Adaw,' which appears to be a typo for 'AdamW' or 'Adam'; please correct.
  2. [Table 4] Dimension names are inconsistent with Section 1: 'Collegiality' appears as 'Collectiveness,' 'Social versatility' as 'Social Diversity,' and 'Overall hireability' as 'Overall Employability.' Please align the terminology.
  3. [§2] The related-work section claims the method demonstrates 'superior generalization performance and prediction accuracy on multiple standard multi-input multi-label regression datasets,' but the paper evaluates only the AVI2025 dataset; please revise this claim.
  4. [§1 and Abstract] The phrases 'significantly outperforming' and 'significantly superior' are used without statistical tests; please temper them unless significance testing is added.
  5. [§3.2] The word 'learanble' appears in the description of the second linear layer; please fix the typo.
  6. [Table 1] The K-fold column uses '×' and '✓,' but the table caption does not define these symbols or state the K value.

Circularity Check

0 steps flagged · score 0.0 of 10

No circular derivation: the reported MSE is produced by supervised regression with an explicit mean-pooling aggregation rule, not by any input-defines-output construction.

full rationale

The paper's central claim is an empirical challenge result. Features are extracted by pretrained encoders (SigLIP2, emotion2vec, SFR-Mistral) and the MSCMLP weights and regression heads are trained against provided labels using MSE loss (Eq. 12); the final score is the explicit mean over 32 heads and six responses (Eq. 11), an aggregation rule stated independently of the labels. No equation defines a target dimension in terms of its own prediction, and no fitted parameter is renamed as a prediction by construction. The references are to external prior work; no self-citation chain or imported uniqueness theorem is load-bearing. The main validity concern is that the pooling configuration in Table 2 appears to have been selected using test MSE (0.1824 for Max/Max) and then reported in Tables 3 and 4 as the final result; this is an evaluation-bias / data-snooping issue rather than circularity, because the test labels did not enter the training of the regression weights and the reported number is not equivalent to an input by construction. Under the given circularity criteria, no enumerated circular step is present.

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

The central claim rests on frozen pretrained features, a shared projection basis, and equal-weight response pooling. The free parameters are all selected via validation or hand-set, and none are derived from first principles.

free parameters (5)
  • Number of ensemble heads H = 32
    Chosen via ablation on the validation set (Figure 4); the optimum is 32.
  • Compression dimension C = 768
    All modality features are projected to 768 dimensions before fusion (Section 4.2).
  • Dropout rates = 0.2/0.1/0.3
    Set for feature adapter, text module, and temporal control module, respectively (Section 4.2); not derived.
  • Learning rate and batch size = 1e-4, 64
    Optimization hyperparameters (Section 4.2).
  • Pooling strategy per modality = max/max
    Selected by comparing validation/test MSE across four pooling combinations (Table 2).
assumptions (5)
  • domain assumption Pre-trained feature extractors (SigLIP2, emotion2vec, SFR-Mistral) encode task-relevant cues for interview performance
    The whole pipeline relies on frozen embeddings capturing the signal; no fine-tuning or analysis is provided (Section 3.1).
  • domain assumption ASR transcription is accurate enough that text embeddings reflect the verbal content
    The text modality comes from an ASR system [47] with no error analysis (Section 3.1).
  • domain assumption Rater-averaged labels are reliable ground truth
    Labels are means of five raters; rater variance is not reported (Section 4.1).
  • domain assumption A single shared linear basis can align all three modalities
    MSCMLP assumes cross-modal interaction is captured by weighted sums of shared basis vectors (Eq. 7-9).
  • standard math Standard linear algebra and GeLU used in MLPs behave as implemented
    Equations 4-9 assume standard matrix operations.

how reviews work

0 comments
Cite this review

Pith. "Pith review of Listening to the Unspoken: Exploring "365" Aspects of Multimodal Interview Performance Assessment." pith.science (2026). https://pith.science/paper/XSPE5RE5

@misc{pith2026250722676,
  author       = {Pith},
  title        = {Pith review of: Listening to the Unspoken: Exploring "365" Aspects of Multimodal Interview Performance Assessment},
  year         = {2026},
  howpublished = {\url{https://pith.science/paper/XSPE5RE5}},
  note         = {Machine review of arXiv:2507.22676}
}
read the original abstract

Interview performance assessment is essential for determining candidates' suitability for professional positions. To ensure holistic and fair evaluations, we propose a novel and comprehensive framework that explores ``365'' aspects of interview performance by integrating \textit{three} modalities (video, audio, and text), \textit{six} responses per candidate, and \textit{five} key evaluation dimensions. The framework employs modality-specific feature extractors to encode heterogeneous data streams and subsequently fused via a Shared Compression Multilayer Perceptron. This module compresses multimodal embeddings into a unified latent space, facilitating efficient feature interaction. To enhance prediction robustness, we incorporate a two-level ensemble learning strategy: (1) independent regression heads predict scores for each response, and (2) predictions are aggregated across responses using a mean-pooling mechanism to produce final scores for the five target dimensions. By listening to the unspoken, our approach captures both explicit and implicit cues from multimodal data, enabling comprehensive and unbiased assessments. Achieving a multi-dimensional average MSE of 0.1824, our framework secured first place in the AVI Challenge 2025, demonstrating its effectiveness and robustness in advancing automated and multimodal interview performance assessment. The full implementation is available at https://github.com/MSA-LMC/365Aspects.

Figures

Figures reproduced from arXiv: 2507.22676 by the authors.

Figure 1
Figure 1. A conceptual visualization of AVI Challenge 2025 [PITH_FULL_IMAGE:figures/full_fig_p002_1.png] view at source ↗
Figure 2
Figure 2. The proposed framework is illustrated as follows: For each sample, six responses are processed in parallel across [PITH_FULL_IMAGE:figures/full_fig_p004_2.png] view at source ↗
Figure 3
Figure 3. Comparison of Multilayer perceptron (a) and Our [PITH_FULL_IMAGE:figures/full_fig_p004_3.png] view at source ↗
Figures from the paper (1 more)
Figure 4
Figure 4. Figure 4: Ablation on the number of MLPs Ablation on Pooling Methods for Different Modalities.We tested video, audio, and text pooling combinations ( [PITH_FULL_IMAGE:figures/full_fig_p005_4.png]

Discussion (0). Sign in to comment.

Reference graph

Works this paper leans on

52 extracted references · 44 canonical work pages

  1. [1]

    Yunus Abdi, Ömer Küllü, Mehmet Kıvılcım Keleş, and Berk Gökberk. 2023. CPW- DICE: a novel center and pixel-based weighting for damage segmentation. Con- nection Science 35, 1 (2023), 2259115

  2. [2]

    Bagus Tris Atmaja and Akira Sasou. 2023. Ensembling multilingual pre-trained models for predicting multi-label regression emotion share from speech. In 2023 Asia Pacific Signal and Information Processing Association Annual Summit and Conference (APSIPA ASC). IEEE, 1026–1029

  3. [3]

    Caesar Balona. 2024. ActuaryGPT: Applications of large language models to insurance and actuarial work. British Actuarial Journal 29 (2024), e15

  4. [4]

    Jianghui Cai, Min Zhang, Haifeng Yang, Yanting He, Yuqing Yang, Chenhui Shi, Xujun Zhao, and Yaling Xun. 2024. A novel graph-attention based multimodal fusion network for joint classification of hyperspectral image and LiDAR data. Expert Systems with Applications 249 (2024), 123587

  5. [5]

    Jinzhou Cao, Xiangxu Wang, Guanzhou Chen, Wei Tu, Xiaole Shen, Tianhong Zhao, Jiashi Chen, and Qingquan Li. 2025. Disentangling the hourly dynamics of mixed urban function: A multimodal fusion perspective using dynamic graphs. Information Fusion 117 (2025), 102832

  6. [6]

    Meixu Chen, Kai Wang, and Jing Wang. 2024. Advancing Head and Neck Cancer Survival Prediction via Multi-Label Learning and Deep Model Interpretation. ArXiv (2024), arXiv–2405

  7. [7]

    Jacob Devlin, Ming-Wei Chang, Kenton Lee, and Kristina Toutanova. 2019. Bert: Pre-training of deep bidirectional transformers for language understanding. In Proceedings of the 2019 conference of the North American chapter of the association for computational linguistics: human language technologies, volume 1 (long and short papers). 4171–4186

  8. [8]

    Thomas G Dietterich et al. 2002. Ensemble learning. The handbook of brain theory and neural networks 2, 1 (2002), 110–125

Show all 52 references
  1. [9]

    Anuj Dimri, Arindam Paul, Deeptha Girish, Peng Lee, Sardar Afra, and Andrew Jakubowski. 2022. A multi-input multi-label claims channeling system using insurance-based language models. Expert Systems with Applications 202 (2022), 117166

  2. [10]

    Xibin Dong, Zhiwen Yu, Wenming Cao, Yifan Shi, and Qianli Ma. 2020. A survey on ensemble learning. Frontiers of Computer Science 14, 2 (2020), 241–258

  3. [11]

    Florian Eyben, Martin Wöllmer, and Björn Schuller. 2010. Opensmile: the munich versatile and fast open-source audio feature extractor. In Proceedings of the 18th ACM international conference on Multimedia . 1459–1462

  4. [12]

    Chenquan Gan, Xiang Fu, Qingdong Feng, Qingyi Zhu, Yang Cao, and Ye Zhu

  5. [13]

    Sina Ghassemi, Tianyi Zhang, Ward van Breda, Antonis Koutsoumpis, Janneke K Oostrom, Djurre Holtrop, and Reinout E de Vries. 2023. Unsupervised multi- modal learning for dependency-free personality recognition. IEEE transactions on affective computing 15, 3 (2023), 1053–1066

  6. [14]

    Jelena Gorbova, Iiris Lusi, Andre Litvin, and Gholamreza Anbarjafari. 2017. Au- tomated screening of job candidate based on multimodal video processing. In Proceedings of the IEEE conference on computer vision and pattern recognition workshops. 29–35

  7. [15]

    Sepp Hochreiter and Jürgen Schmidhuber. 1997. Long short-term memory.Neural computation 9, 8 (1997), 1735–1780

  8. [16]

    Aaditya Jadhav, Rushikesh Ghodake, Karthik Muralidharan, and G Tarun Varma

  9. [17]

    Kenan Kassab and Alexey Kashevnik. 2024. Personality Traits Estimation Based on Job Interview Video Analysis: Importance of Human Nonverbal Cues Detec- tion. Big Data and Cognitive Computing 8, 12 (2024), 173

  10. [18]

    Onur Köksoy. 2006. Multiresponse robust design: Mean square error (MSE) criterion. Appl. Math. Comput. 175, 2 (2006), 1716–1729

  11. [19]

    Antonis Koutsoumpis, Sina Ghassemi, Janneke K Oostrom, Djurre Holtrop, Ward van Breda, Tianyi Zhang, and Reinout E de Vries. 2024. Beyond traditional inter- views: Psychometric analysis of asynchronous video interviews for personality and interview performance evaluation using...

  12. [20]

    Vivien Kraus, Khalid Benabdeslem, and Bruno Canitia. 2020. Laplacian-based semi-supervised multi-label regression. In 2020 International Joint Conference on Neural Networks (IJCNN). IEEE, 1–8

  13. [21]

    Yann LeCun, Léon Bottou, Yoshua Bengio, and Patrick Haffner. 2002. Gradient- based learning applied to document recognition. Proc. IEEE 86, 11 (2002), 2278– 2324

  14. [22]

    Peishan Li, Weixiao Zhan, Lutao Gao, Shuran Wang, and Linnan Yang. 2025. Multimodal Recommendation System Based on Cross Self-Attention Fusion. Systems 13, 1 (2025), 57

  15. [23]

    Paloma Helena Lira Fontes. 2024. Machine Learning alternative to sparseness: a Radon transform application for multiple and ground roll attenuation. (2024)

  16. [24]

    Jing Liu, Xinghua Tang, Shuanglong Cui, and Xiao Guan. 2022. Predicting the function of rice proteins through Multi-instance Multi-label Learning based on multiple features fusion. Briefings in Bioinformatics 23, 3 (2022), bbac095

  17. [25]

    Kaifang Long, Guoyang Xie, Lianbo Ma, Jiaqi Liu, and Zhichao Lu. 2025. Revisiting multimodal fusion for 3D anomaly detection from an architectural perspective. In Proceedings of the AAAI Conference on Artificial Intelligence, Vol. 39. 12273–12281

  18. [26]

    Jiasen Lu, Dhruv Batra, Devi Parikh, and Stefan Lee. 2019. Vilbert: Pretrain- ing task-agnostic visiolinguistic representations for vision-and-language tasks. Advances in neural information processing systems 32 (2019)

  19. [27]

    Nannan Lu, Zhiyuan Han, Min Han, and Jiansheng Qian. 2024. Bi-stream graph learning based multimodal fusion for emotion recognition in conversation. In- formation Fusion 106 (2024), 102272

  20. [28]

    Ziyang Ma, Zhisheng Zheng, Jiaxin Ye, Jinchao Li, Zhifu Gao, Shiliang Zhang, and Xie Chen. 2023. emotion2vec: Self-supervised pre-training for speech emotion representation. arXiv preprint arXiv:2312.15185 (2023)

  21. [29]

    Sijie Mai, Haifeng Hu, and Songlong Xing. 2020. Modality to modality translation: An adversarial representation learning and graph fusion network for multimodal fusion. In Proceedings of the AAAI conference on artificial intelligence , Vol. 34. 164–172

  22. [30]

    Rui Meng, Ye Liu, Shafiq Rayhan Joty, Caiming Xiong, Yingbo Zhou, and Semih Yavuz. 2024. Sfrembedding-mistral: enhance text retrieval with transfer learning. Salesforce AI Research Blog 3 (2024), 6

  23. [31]

    Ibomoiye Domor Mienye and Yanxia Sun. 2022. A survey of ensemble learning: Concepts, algorithms, applications, and prospects. Ieee Access 10 (2022), 99129– 99149

  24. [32]

    José Morano, Guilherme Aresta, Christoph Grechenig, Ursula Schmidt-Erfurth, and Hrvoje Bogunović. 2024. Deep multimodal fusion of data with heterogeneous dimensionality via projective networks. IEEE Journal of Biomedical and Health Informatics 28, 4 (2024), 2235–2246

  25. [33]

    Laurent Son Nguyen, Alvaro Marcos-Ramiro, Martha Marrón Romera, and Daniel Gatica-Perez. 2013. Multimodal analysis of body communication cues in employ- ment interviews. In Proceedings of the 15th ACM on International conference on multimodal interaction. 437–444

  26. [34]

    Abshishek Rajora, Shubham Gupta, and Suman Kundu. 2025. Cross-Aligned Fusion For Multimodal Understanding. In 2025 IEEE/CVF Winter Conference on Applications of Computer Vision (W ACV). IEEE, 5730–5740

  27. [35]

    David E Rumelhart, Geoffrey E Hinton, and Ronald J Williams. 1986. Learning representations by back-propagating errors. nature 323, 6088 (1986), 533–536

  28. [36]

    Gaurav Sahu and Olga Vechtomova. 2019. Dynamic fusion for multimodal data. arXiv preprint arXiv:1911.03821 (2019)

  29. [37]

    Junyuan Shang, Shenda Hong, Yuxi Zhou, Meng Wu, and Hongyan Li. 2018. Knowledge guided multi-instance multi-label learning via neural networks in medicines prediction. In Asian Conference on Machine Learning . PMLR, 831–846

  30. [38]

    Ethan Steinberg and Peter J Liu. 2019. Using ontologies to improve performance in massively multi-label prediction models. arXiv preprint arXiv:1905.12126 (2019)

  31. [39]

    Jiajia Tang, Kang Li, Xuanyu Jin, Andrzej Cichocki, Qibin Zhao, and Wanzeng Kong. 2021. CTFN: Hierarchical learning for multimodal sentiment analysis using coupled-translation fusion network. In Proceedings of the 59th annual meeting of the association for computational lingui...

  32. [40]

    Michael Tschannen, Alexey Gritsenko, Xiao Wang, Muhammad Ferjad Naeem, Ibrahim Alabdulmohsin, Nikhil Parthasarathy, Talfan Evans, Lucas Beyer, Ye Xia, Basil Mustafa, et al. 2025. Siglip 2: Multilingual vision-language encoders with improved semantic understanding, localization...

  33. [41]

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

  34. [42]

    Wen-chuan Wang, Wei-can Tian, Xiao-xue Hu, Yang-hao Hong, Fu-xin Chai, and Dong-mei Xu. 2024. Dttr: Encoding and decoding monthly runoff prediction model based on deep temporal attention convolution and multimodal fusion. Journal of Hydrology 643 (2024), 131996

  35. [43]

    Yikai Wang, Wenbing Huang, Fuchun Sun, Tingyang Xu, Yu Rong, and Junzhou Huang. 2020. Deep multimodal fusion by channel exchanging.Advances in neural information processing systems 33 (2020), 4835–4845

  36. [44]

    Yibin Wang, Gensheng Pei, and Yusheng Cheng. 2019. Ensemble regression kernel extreme learning machines for multi-instance multi-label learning. In International CCF Conference on Artificial Intelligence . Springer, 226–239

  37. [45]

    Jian-Sheng Wu, Sheng-Jun Huang, and Zhi-Hua Zhou. 2014. Genome-wide pro- tein function prediction through multi-instance multi-label learning. IEEE/ACM Transactions on Computational Biology and Bioinformatics 11, 5 (2014), 891–902

  38. [46]

    Qingyao Wu, Mingkui Tan, Hengjie Song, Jian Chen, and Michael K Ng. 2016. ML-FOREST: A multi-label tree ensemble method for multi-label classification. IEEE transactions on knowledge and data engineering 28, 10 (2016), 2665–2680

  39. [47]

    Dong Yu and Lin Deng. 2016. Automatic speech recognition. Vol. 1. Springer

  40. [48]

    Tianyi Zhang, Antonis Koutsoumpis, Janneke K Oostrom, Djurre Holtrop, Sina Ghassemi, and Reinout E De Vries. 2024. Can large language models assess personality from asynchronous video interviews? A comprehensive evaluation of validity, reliability, fairness, and rating pattern...

  41. [49]

    Fei Zhao, Chengcui Zhang, and Baocheng Geng. 2024. Deep multimodal data fusion. ACM computing surveys 56, 9 (2024), 1–36

  42. [50]

    Xiaoyan Zhu, Jiaxuan Li, Jingtao Ren, Jiayin Wang, and Guangtao Wang. 2023. Dynamic ensemble learning for multi-label classification. Information Sciences 623 (2023), 94–111

  43. [2023]

    AI based multimodal emotion and behavior analysis of interviewee. (2023)

  44. [2024]

    Expert Systems with Applications 242 (2024), 122731

    A multimodal fusion network with attention mechanisms for visual–textual sentiment analysis. Expert Systems with Applications 242 (2024), 122731

Pith tools

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