SoDRA: purifying social graphs with behavior verification for social recommendations

Abstract Social recommendation systems leverage social connections to enhance user modeling. However, real-world social graphs often contain noisy or task-irrelevant edges that do not reflect genuine preference alignment. These noisy connections distort user preference learning and degrade recommendation quality. Purifying such massive, noisy graphs demands high-performance computing (HPC) capabilities. We propose SoDRA (Social Dynamic Refinement and Alignment), a framework that dynamically purifies social graphs via a closed-loop “guess → verify → update” mechanism. The Dynamic Semantic Calibrator (DSC) first re-weights social edges based on behavioral relevance, hypothesizing an optimized graph structure. The Semantic Validator (SV) then computes a differentiable consistency score between social and behavior embeddings. SoDRA employs multi-layer graph convolutional network (GCN) propagation on both user-item and social graphs. The gradient of the recommendation loss flows through this score back to DSC, enabling end-to-end optimization without additional loss terms. Experiments on three real-world datasets demonstrate that SoDRA consistently outperforms state-of-the-art methods across all evaluation metrics, with the best improvement reaching 16.58% in NDCG@20. Additional analysis confirms its robustness in cold-start and long-tail scenarios. The model is implemented with efficient batch-wise processing on a GPU-accelerated platform, demonstrating practical efficiency for social recommendation scenarios. Highlights - A “guess → verify → update” iterative cycle dynamically refines social graphs through behavior-verified influence learning. - A Dynamic Semantic Calibrator (DSC) that re-weights social edges by behavioral relevance, jointly with a Semantic Validator (SV) that computes differentiable semantic consistency scores for end-to-end graph tuning. - Extensive experiments on three real-world datasets demonstrate consistent improvements over state-of-the-art methods, with up to 16.58% NDCG@20 gain and strong robustness in cold-start and long-tail scenarios. Similar content being viewed by others Data availability Data will be made available on request. References Salamat A, Luo X, Jafari A (2021) Heterographrec: a heterogeneous graph-based neural networks for social recommendations. Knowl Based Syst 217:106817. https://doi.org/10.1016/j.knosys.2021.106817 Yan S-R, Zheng X-L, Wang Y et al (2015) A graph-based comprehensive reputation model: exploiting the social context of opinions to enhance trust in social commerce. Inf Sci 318:51–72. https://doi.org/10.1016/j.ins.2014.09.036 Najafabadi MK, Chen R-A, Rezazadeh J et al (2025) From theory to practice: the evolution and comparative analysis of homogeneous vs. heterogeneous graph neural networks in recommender systems. Neurocomputing 624:129446. https://doi.org/10.1016/j.neucom.2025.129446 Han J, Tang Y, Tao Q et al (2024) Dual homogeneity hypergraph motifs with cross-view contrastive learning for multiple social recommendations. ACM Trans Knowl Discov Data 18:158:1-158:24. https://doi.org/10.1145/3653976 Sharma K, Lee Y-C, Nambi S et al (2024) A survey of graph neural networks for social recommender systems. ACM Comput Surv 56:265:1-265:34. https://doi.org/10.1145/3661821 He X, Fan W, Wang R et al (2025) Balancing user preferences by social networks: a condition-guided social recommendation model for mitigating popularity bias. Neural Netw 187:107317. https://doi.org/10.1016/j.neunet.2025.107317 Fan W, Ma Y, Li Q et al (2022) A graph neural network framework for social recommendations. IEEE Trans Knowl Data Eng 34:2033–2047. https://doi.org/10.1109/TKDE.2020.3008732 Wu L, Sun P, Hong R et al (2021) Collaborative neural social recommendation. IEEE Trans Syst Man Cybern Syst 51:464–476. https://doi.org/10.1109/TSMC.2018.2872842 Wu L, Sun P, Fu Y, et al (2019) A Neural Influence Diffusion Model for Social Recommendation. In: Proceedings of the 42nd International ACM SIGIR Conference on Research and Development in Information Retrieval. Association for Computing Machinery, New York, NY, USA, p 235–244 Wu L, Sun P, Hong R, Fu Y, Wang X, Wang M (2019) Socialgcn: an efficient graph convolutional network based model for social recommendation. arXiv preprint arXiv:1811.02815. Abbas K, Xin L, Mingsheng S (2016) discovering items with potential popularity on social media. In: 2016 IEEE 14th Intl Conf on Dependable, Autonomic and Secure Computing, 14th Intl Conf on Pervasive Intelligence and Computing, 2nd Intl Conf on Big Data Intelligence and Computing and Cyber Science and Technology Congress (DASC/PiCom/DataCom/CyberSciTech). p 459–466 Tang J, Sun J, Wang C, Yang Z (2009) Social influence analysis in large-scale networks. In: Proceedings of the 15th ACM SIGKDD international conference on Knowledge discovery and data mining. Association for Computing Machinery, New York, NY, USA, p 807–816 Munagala K, Wang K (2019) improved metric distortion for deterministic social choice rules. In: Proceedings of the 2019 ACM Conference on Economics and Computation. Association for Computing Machinery, New York, NY, USA, p 245–262 Masoumzadeh A, Joshi J (2012) Preserving structural properties in edge-perturbing anonymization techniques for social networks. IEEE Trans Dependable Secure Comput 9:877–889. https://doi.org/10.1109/TDSC.2012.65 Sun S, Yu T, Xu J et al (2023) GraphIQA: learning distortion graph representations for blind image quality assessment. IEEE Trans Multimed 25:2912–2925. https://doi.org/10.1109/TMM.2022.3152942 Liao J, Zhou W, Luo F et al (2022) SocialLGN: light graph convolution network for social recommendation. Inf Sci 589:595–607. https://doi.org/10.1016/j.ins.2022.01.001 Chen G, Xia L, Huang C (2025) LightGNN: simple graph neural network for recommendation. In: Proceedings of the Eighteenth ACM International Conference on Web Search and Data Mining. Association for Computing Machinery, New York, NY, USA, p 549–558 Liu Y, Xia L, Huang C (2024) SelfGNN: self-supervised graph neural networks for sequential recommendation. In: Proceedings of the 47th International ACM SIGIR Conference on Research and Development in Information Retrieval. Association for Computing Machinery, New York, NY, USA, p 1609–1618 Corso G, Stark H, Jegelka S et al (2024) Graph neural networks. Nat Rev Methods Primers 4:1–13. https://doi.org/10.1038/s43586-024-00294-7 Wu Z, Pan S, Chen F et al (2021) A comprehensive survey on graph neural networks. IEEE Trans Neural Netw Learn Syst 32:4–24. https://doi.org/10.1109/TNNLS.2020.2978386 Fan W, Ma Y, Li Q, et al (2019) graph neural networks for social recommendation. In: The World Wide Web Conference. Association for Computing Machinery, New York, NY, USA, p 417–426 Guo Z, Wang H (2021) A deep graph neural network-based mechanism for social recommendations. IEEE Trans Ind Inform 17:2776–2783. https://doi.org/10.1109/TII.2020.2986316 Hu B, Zhou N, Zhou Q et al (2020) DiffNet: a learning to compare deep network for product recognition. IEEE Access 8:19336–19344. https://doi.org/10.1109/ACCESS.2020.2967090 Yang Y, Wu L, Liao Y, et al (2025) invariance matters: empowering social recommendation via graph invariant learning. In: Proceedings of the 48th International ACM SIGIR Conference on Research and Development in Information Retrieval. Association for Computing Machinery, New York, NY, USA, p 2038–2047 He D, Wang T, Zhai L et al (2023) Adversarial representation mechanism learning for network embedding. IEEE Trans Knowl Data Eng 35:1200–1213. https://doi.org/10.1109/TKDE.2021.3103193 Wang X, He X, Wang M, et al (2019) Neural graph collaborative filtering. In: Proceedings of the 42nd International ACM SIGIR Conference on Research and Development in Information Retrieval. Association for Computing Machinery, New York, NY, USA, pp 165–174 He X, Deng K, Wang X, et al (2020) LightGCN: simplifying and powering graph convolution network for recommendation. In: Proceedings of the 43rd International ACM SIGIR Conference on Research and Development in Information Retrieval. Association for Computing Machinery, New York, NY, USA, pp 639–648 Tang X, Liu Y, He X, et al (2022) friend story ranking with edge-contextual local graph convolutions. In: Proceedings of the Fifteenth ACM International Conference on Web Search and Data Mining. Association for Computing Machinery, New York, NY, USA, pp 1007–1015 Qiao W, Wu W, Liu S et al (2025) SauronEyes: disentangling voluminous logs to unveil camouflaged attack intentions. IEEE Trans Inf Forensics Secur. https://doi.org/10.1109/TIFS.2025.3618381 Shao T, Zeng W, Zhao X (2025) DSHCL: dual-state hypergraph contrastive learning for information diffusion prediction. IEEE Trans Knowl Data Eng 37:5158–5170. https://doi.org/10.1109/TKDE.2025.3581419 Lim S, Lee H, Kim H, et al (2023) ZTLS: a DNS-based approach to zero round trip delay in TLS handshake. In: Proceedings of the ACM Web Conference 2023. Association for Computing Machinery, New York, NY, USA, p 2360–2370 Sun H, Li X, Su D, et al (2024) Towards data-centric machine learning on directed graphs: a Survey Hu W, Wu J, Qian Q (2025) GAFExplainer: global view explanation of graph neural networks through attribute augmentation and fusion embedding. IEEE Trans Knowl Data Eng 37:2569–2583. https://doi.org/10.1109/TKDE.2025.3539989 Zhou X, Shen Z (2023) A Tale of Two Graphs: Freezing and denoising graph structures for multimodal recommendation. In: Proceedings of the 31st ACM International Conference on Multimedia. Association for Computing Machinery, New York, NY, USA, p 935–943 Wang W, Xu Y, Feng F, et al (2023) Diffusion recommender model. In: Proceedings of the 46th International ACM SIGIR Conference on Research and Development in Information Retrieval. Association for Computing Machinery, New York, NY, USA, p 832–841 Yu P, Tan Z, Lu G, Bao B-K (2023) Multi-view graph convolutional network for multimedia recommendation. In: Proceedings of the 31st ACM International Conference on Multimedia. Association for Computing Machinery, New York, NY, USA, p 6576–6585 Zhang Y, Chen Z, Cheng C-H, et al (2025) Trustworthy text-to-image diffusion models: a timely and focused survey Ren X, Wei W, Xia L, Huang C (2025) A comprehensive survey on self-supervised learning for recommendation. ACM Comput Surv. https://doi.org/10.1145/3746280 Louizos C, Welling M, Kingma DP (2018) Learning sparse neural networks through l0 regularization Li Z, Xia L, Huang C (2024) RecDiff: diffusion model for social recommendation. In: Proceedings of the 33rd ACM International Conference on Information and Knowledge Management. Association for Computing Machinery, New York, NY, USA, p 1346–1355 Liu C, Zhang J, Wang S et al (2025) Score-based generative diffusion models for social recommendations. IEEE Trans Knowl Data Eng 37:6666–6679. https://doi.org/10.1109/TKDE.2025.3600103 Chen W, Zhang Y, Li H et al (2025) Dual-domain collaborative denoising for social recommendation. IEEE Trans Comput Soc Syst 12:2736–2751. https://doi.org/10.1109/TCSS.2025.3529706 Wang L, Wang W, Xiao X, Li Q (2025) contrastive learning augmented social recommendations Rendle S, Freudenthaler C, Gantner Z, Schmidt-Thieme L (2012) BPR: bayesian personalized ranking from implicit feedback Acknowledgements This work was supported by National Natural Science Foundation of China (No. 72501197), Chengdu Philosophy and Social Sciences Planning Project (No. 25CS090), and Natural Science Foundation of Sichuan Province (No. 2024NSFSC1059). Funding National Natural Science Foundation of China (No. 72501197), Chengdu Philosophy and Social Sciences Planning Project (No. 25CS090), and Natural Science Foundation of Sichuan Province (No. 2024NSFSC1059). Author information Authors and Affiliations Contributions Y.G. and H.S. contributed equally to this work and share first authorship. Y.G. helped in conceptualization; methodology design; formal analysis of model components; review & editing; and funding acquisition. H.S. worked in software implementation (PyTorch-based model coding); data curation (collection, preprocessing, and splitting of Amazon, Ciao, and LastFM datasets); validation (reproducing baseline methods and ablation studies); investigation (hyperparameter tuning and performance analysis); writing—original draft; and visualization (tables and result plots). X.L. helped in methodology refinement; theoretical analysis; formal analysis of cold-start scenarios; and supervision of experimental reproducibility. Y.Z. helped in investigation of related work; data curation; validation (cross-dataset generalization tests); and formatting and reference management. M.C. worked in software optimization; validation; visualization; assistance in code documentation; and open-source preparation. W.W. (corresponding author) helped in overall research direction and problem framing; methodology oversight; supervision of the entire project; project administration; and correspondence with journal editors. All authors reviewed and approved the final manuscript. Corresponding author Ethics declarations Conflict of interest The authors declare that they have no known competing financial interests or personal relationships that could have appeared to influence the work reported in this paper. Additional information Publisher's Note Springer Nature remains neutral with regard to jurisdictional claims in published maps and institutional affiliations. Rights and permissions Springer Nature or its licensor (e.g. a society or other partner) holds exclusive rights to this article under a publishing agreement with the author(s) or other rightsholder(s); author self-archiving of the accepted manuscript version of this article is solely governed by the terms of such publishing agreement and applicable law. About this article Cite this article Guo, Y., Hou, S., Li, X. et al. SoDRA: purifying social graphs with behavior verification for social recommendations. J Supercomput 82, 743 (2026). https://doi.org/10.1007/s11227-026-08879-7 Received: Accepted: Published: Version of record: DOI: https://doi.org/10.1007/s11227-026-08879-7