{"status":"ok","message-type":"work","message-version":"1.0.0","message":{"indexed":{"date-parts":[[2026,6,18]],"date-time":"2026-06-18T06:52:34Z","timestamp":1781765554616,"version":"3.54.5"},"reference-count":51,"publisher":"Wiley","license":[{"start":{"date-parts":[[2026,6,17]],"date-time":"2026-06-17T00:00:00Z","timestamp":1781654400000},"content-version":"vor","delay-in-days":0,"URL":"http:\/\/creativecommons.org\/licenses\/by-nc\/4.0\/"},{"start":{"date-parts":[[2026,6,17]],"date-time":"2026-06-17T00:00:00Z","timestamp":1781654400000},"content-version":"tdm","delay-in-days":0,"URL":"http:\/\/doi.wiley.com\/10.1002\/tdm_license_1.1"}],"content-domain":{"domain":["onlinelibrary.wiley.com"],"crossmark-restriction":true},"short-container-title":["Computer Graphics Forum"],"abstract":"<jats:title>Abstract<\/jats:title>\n                  <jats:p>Network visualization has traditionally relied on heuristic metrics, such as stress, under the assumption that optimizing them leads to aesthetic and informative layouts. However, no single metric consistently produces the most effective results. A data\u2010driven alternative is to learn from human preferences, where labelers select their favored visualization among multiple layouts of the same graphs. These human\u2010preference labels can then be used to train a generative model that approximates human aesthetic preferences. However, obtaining human labels at scale is costly and time\u2010consuming. As a result, this generative approach has so far been tested only with machine\u2010labeled data [WYHS24]. In this paper, we explore the use of large language models (LLMs) and vision models (VMs) as proxies for human judgment. Through a carefully designed user study involving 27 participants, we curated a large set of human preference labels. We used this data both to better understand human preferences and to bootstrap LLM\/VM labelers. We show that prompt engineering that combines few\u2010shot examples and diverse input formats, such as image embeddings, significantly improves LLM\u2010\u2010human alignment, and additional filtering by the confidence score of the LLM pushes the alignment to human\u2010\u2010human levels. Furthermore, we demonstrate that carefully trained VMs can achieve VM\u2010human alignment at a level comparable to that between human labelers. Our results suggest that AI can feasibly serve as a scalable proxy for human labelers.<\/jats:p>","DOI":"10.1111\/cgf.70456","type":"journal-article","created":{"date-parts":[[2026,6,18]],"date-time":"2026-06-18T06:03:20Z","timestamp":1781762600000},"update-policy":"https:\/\/doi.org\/10.1002\/crossmark_policy","source":"Crossref","is-referenced-by-count":0,"title":["Beauty in the Eye of AI: Aligning LLMs and Vision Models with Human Aesthetics in Network Visualization"],"prefix":"10.1111","author":[{"ORCID":"https:\/\/orcid.org\/0009-0007-6059-9747","authenticated-orcid":false,"given":"X.","family":"Li","sequence":"first","affiliation":[{"name":"Northeastern University  United States"}],"role":[{"vocabulary":"crossref","role":"author"}]},{"ORCID":"https:\/\/orcid.org\/0000-0001-7756-0901","authenticated-orcid":false,"given":"P.","family":"Zhang","sequence":"additional","affiliation":[{"name":"Northeastern University  United States"}],"role":[{"vocabulary":"crossref","role":"author"}]},{"ORCID":"https:\/\/orcid.org\/0000-0003-3674-1428","authenticated-orcid":false,"given":"X.","family":"Wang","sequence":"additional","affiliation":[{"name":"Bosch AI Research  United States"}],"role":[{"vocabulary":"crossref","role":"author"}]},{"ORCID":"https:\/\/orcid.org\/0000-0002-1211-2320","authenticated-orcid":false,"given":"H.","family":"Shen","sequence":"additional","affiliation":[{"name":"The Ohio State University  United States"}],"role":[{"vocabulary":"crossref","role":"author"}]},{"ORCID":"https:\/\/orcid.org\/0000-0003-2017-924X","authenticated-orcid":false,"given":"Y.","family":"Hu","sequence":"additional","affiliation":[{"name":"Northeastern University  United States"}],"role":[{"vocabulary":"crossref","role":"author"}]}],"member":"311","published-online":{"date-parts":[[2026,6,17]]},"reference":[{"key":"e_1_2_9_2_2","unstructured":"ArgyriouE. BekosM. SymvonisA.: Maximizing the total resolution of graphs.Proc. of GD 2010 6502(092010). 2"},{"issue":"6","key":"e_1_2_9_3_2","first-page":"2388","article-title":"Multicriteria scalable graph drawing via stochastic gradient descent, (SGD)2","volume":"28","author":"Ahmed R.","year":"2022","journal-title":"IEEE Transactions on Visualization and Computer Graphics"},{"issue":"1","key":"e_1_2_9_4_2","doi-asserted-by":"crossref","first-page":"7","DOI":"10.1093\/comjnl\/bxz133","article-title":"A heuristic approach towards drawings of graphs with high crossing resolution","volume":"64","author":"Bekos M. A.","year":"2021","journal-title":"The Computer Journal"},{"key":"e_1_2_9_5_2","doi-asserted-by":"crossref","first-page":"303","DOI":"10.1016\/S0925-7721(96)00005-3","article-title":"An experimental comparison of four graph drawing algorithms","volume":"7","author":"Battista G. D.","year":"1997","journal-title":"CGTA: Computational Geometry: Theory and Applications"},{"key":"e_1_2_9_6_2","doi-asserted-by":"crossref","unstructured":"BrandesU. PichC.: Eigensolver methods for progressive multidimensional scaling of large data.LNCS 4372(092006). doi:10.1007\/978-3-540-70904-6_6. 3","DOI":"10.1007\/978-3-540-70904-6_6"},{"key":"e_1_2_9_7_2","unstructured":"ChimaniM. EadesP. EadesP. HongS.-H. HuangW. KleinK. MarnerM. SmithR. T. Thomas B. H.: People prefer less stress and fewer crossings. InInternational Symposium Graph Drawing and Network Visualization(2014). 2"},{"key":"e_1_2_9_8_2","doi-asserted-by":"crossref","first-page":"6","DOI":"10.1109\/PacificVis52677.2021.00009","volume-title":"2021 IEEE 14th Pacific Visualization Symposium (PacificVis)","author":"Cai S.","year":"2021"},{"key":"e_1_2_9_9_2","doi-asserted-by":"crossref","unstructured":"DevkotaS. AhmedR. De LucaF. IsaacsK. E. KobourovS.: Stress-plus-x (SPX) graph layout. InProf. Springer International Symposium on Graph Drawing and Network Visualization(2019) pp.291\u2013304. 2","DOI":"10.1007\/978-3-030-35802-0_23"},{"key":"e_1_2_9_10_2","doi-asserted-by":"publisher","DOI":"10.2312\/evs.20231047"},{"issue":"5","key":"e_1_2_9_11_2","doi-asserted-by":"crossref","first-page":"33","DOI":"10.1109\/MCG.2019.2924636","article-title":"Data2vis: Automatic generation of data visualizations using sequence-to-sequence recurrent neural networks","volume":"39","author":"Dibia V.","year":"2019","journal-title":"IEEE Computer Graphics and Applications"},{"key":"e_1_2_9_12_2","doi-asserted-by":"publisher","DOI":"10.1109\/TVCG.2009.109"},{"key":"e_1_2_9_13_2","doi-asserted-by":"crossref","unstructured":"DidimoW. LiottaG. RomeoS.: Topology-driven force-directed algorithms.Proc. of GD 2010 6502(092010) 165\u2013176. doi:10.1007\/978-3-642-18469-7_15. 2","DOI":"10.1007\/978-3-642-18469-7_15"},{"key":"e_1_2_9_14_2","first-page":"149","article-title":"A heuristic for graph drawing","volume":"42","author":"Eades P.","year":"1984","journal-title":"Congressus Numerantium"},{"key":"e_1_2_9_15_2","doi-asserted-by":"publisher","DOI":"10.7155\/jgaa.00405"},{"issue":"11","key":"e_1_2_9_16_2","first-page":"1129","article-title":"Graph drawing by force-directed placement","volume":"21","author":"Fruchterman T. M.","year":"1991","journal-title":"Software: Practice and experience"},{"key":"e_1_2_9_17_2","doi-asserted-by":"crossref","first-page":"1129","DOI":"10.1002\/spe.4380211102","article-title":"Graph drawing by force directed placement","volume":"21","author":"Fruchterman T. M. J.","year":"1991","journal-title":"Software - Practice and Experience"},{"key":"e_1_2_9_18_2","doi-asserted-by":"crossref","unstructured":"GansnerE. R. KorenY. NorthS.: Graph drawing by stress majorization. InProc. Springer International Symposium on Graph Drawing(2004) pp.239\u2013250. 2 3","DOI":"10.1007\/978-3-540-31843-9_25"},{"key":"e_1_2_9_19_2","doi-asserted-by":"publisher","DOI":"10.1145\/2939672.2939754"},{"key":"e_1_2_9_20_2","doi-asserted-by":"crossref","unstructured":"GiovannangeliL. LalanneF. AuberD. GiotR. BourquiR.: Deep neural network for drawing networks (DNN)2. InProc. Springer International Symposium on Graph Drawing and Network Visualization(2021) pp.375\u2013390. 3","DOI":"10.1007\/978-3-030-92931-2_27"},{"key":"e_1_2_9_21_2","unstructured":"GoodfellowI. J. Pouget-AbadieJ. MirzaM. XuB. Warde-FarleyD. OzairS. CourvilleA. C. BengioY.: Generative adversarial nets. InProc. NeurIPS(2014). 10"},{"key":"e_1_2_9_22_2","unstructured":"Graphviz:.http:\/\/graphviz.org. 3"},{"key":"e_1_2_9_23_2","first-page":"285","volume-title":"Proc. 12th Intl. Symp. Graph Drawing (GD '04)","author":"Hachul S.","year":"2004"},{"key":"e_1_2_9_24_2","first-page":"37","article-title":"Efficient and high quality force-directed graph drawing","volume":"10","author":"Hu Y.","year":"2005","journal-title":"Mathematica Journal"},{"key":"e_1_2_9_25_2","doi-asserted-by":"publisher","DOI":"10.1109\/MCG.2018.2881501"},{"key":"e_1_2_9_26_2","first-page":"770","volume-title":"Proceedings of the IEEE Conference on Computer Vision and Pattern Recognition (CVPR)","author":"He K.","year":"2016"},{"key":"e_1_2_9_26_3","doi-asserted-by":"crossref","unstructured":"doi:10.1109\/CVPR.2016.90. 5","DOI":"10.1109\/CVPR.2016.90"},{"key":"e_1_2_9_27_2","doi-asserted-by":"publisher","DOI":"10.1371\/journal.pone.0098679"},{"key":"e_1_2_9_28_2","doi-asserted-by":"publisher","DOI":"10.1016\/0020-0190(89)90102-6"},{"issue":"1","key":"e_1_2_9_29_2","doi-asserted-by":"crossref","first-page":"665","DOI":"10.1109\/TVCG.2019.2934396","article-title":"A deep generative model for graph layout","volume":"26","author":"Kwon O.-H.","year":"2019","journal-title":"IEEE Transactions on Visualization and Computer Graphics"},{"key":"e_1_2_9_30_2","doi-asserted-by":"crossref","first-page":"169","DOI":"10.1007\/978-3-030-04414-5_12","volume-title":"Graph Drawing and Network Visualization: 26th International Symposium, GD 2018, Barcelona, Spain, September 26\u201328, 2018, Proceedings","author":"Klammler M.","year":"2018"},{"key":"e_1_2_9_31_2","doi-asserted-by":"crossref","unstructured":"KobourovS. G.: Force-directed drawing algorithms.Handbook of Graph Drawing and Visualization(2013) 383\u2013408. doi:10.1201\/b15385-15. 2","DOI":"10.1201\/b15385-15"},{"key":"e_1_2_9_32_2","doi-asserted-by":"crossref","unstructured":"KorenY.: On spectral graph drawing. InCOCOON(2003). 3 5","DOI":"10.1007\/3-540-45071-8_50"},{"key":"e_1_2_9_33_2","doi-asserted-by":"crossref","unstructured":"LiuC. HanY. JiangR. YuanX.: Advisor: Automatic visualization answer for natural-language question on tabular data. InIEEE Pacific Visualization Symposium (PacificVis)(2021) PacificVis '21 IEEE pp.11\u201320. 3","DOI":"10.1109\/PacificVis52677.2021.00010"},{"key":"e_1_2_9_34_2","unstructured":"MooneyG. J. PurchaseH. WybrowM. KobourovS. G. MillerJ.: The perception of stress in graph drawings. InInternational Symposium Graph Drawing and Network Visualization(2024). 2"},{"key":"e_1_2_9_35_2","unstructured":"OquabM. DarcetT. MoutakanniT. VoH. SzafraniecM. KhalidovV. FernandezP. HazizaD. MassaF. El-NoubyA. AssranM. BallasN. GalubaW. HowesR. HuangP.-Y. LiS.-W. MisraI. RabbatM. SharmaV. SynnaeveG. XuH. J\u00e9gouH. MairalJ. LabatutP. JoulinA. BojanowskiP.: Dinov2: Learning robust visual features without supervision.arXiv preprint arXiv:2304.07193(2023). URL:https:\/\/arxiv.org\/abs\/2304.07193. 5 6"},{"key":"e_1_2_9_36_2","doi-asserted-by":"crossref","unstructured":"PurchaseH.: Which aesthetic has the greatest effect on human understanding? InProc. Springer International Symposium on Graph Drawing(1997) pp.248\u2013261. 2","DOI":"10.1007\/3-540-63938-1_67"},{"issue":"2","key":"e_1_2_9_37_2","doi-asserted-by":"crossref","first-page":"147","DOI":"10.1016\/S0953-5438(00)00032-1","article-title":"Effective information visualisation: a study of graph drawing aesthetics and algorithms","volume":"13","author":"Purchase H. C.","year":"2000","journal-title":"Interacting with Computers"},{"key":"e_1_2_9_38_2","doi-asserted-by":"crossref","DOI":"10.1145\/3325861","article-title":"Geometric heuristics for rectilinear crossing minimization","volume":"24","author":"Radermacher M.","year":"2019","journal-title":"ACM J. Exp. Algorithmics"},{"key":"e_1_2_9_39_2","doi-asserted-by":"publisher","DOI":"10.1111\/cgf.14841"},{"key":"e_1_2_9_40_2","unstructured":"Sim\u00e9oniO. VoH. V. SeitzerM. BaldassarreF. OquabM. JoseC. KhalidovV. SzafraniecM. YiS. RamamonjisoaM. MassaF. HazizaD. WehrstedtL. WangJ. DarcetT. MoutakanniT. SentanaL. RobertsC. VedaldiA. TolanJ. BrandtJ. CouprieC. MairalJ. J\u00e9gouH. LabatutP. BojanowskiP.:Dinov3 2025. URL:https:\/\/arxiv.org\/abs\/2508.10104 arXiv:2508.10104. 6"},{"key":"e_1_2_9_41_2","unstructured":"TiezziM. CiravegnaG. GoriM.: Graph neural networks for graph drawing.IEEE Transactions on Neural Networks and Learning Systems(2022). 2 3"},{"key":"e_1_2_9_42_2","doi-asserted-by":"publisher","DOI":"10.1109\/TVCG.2024.3456378"},{"key":"e_1_2_9_42_3","doi-asserted-by":"crossref","unstructured":"doi:10.1109\/TVCG.2024.3456378. 3","DOI":"10.1109\/TVCG.2024.3456378"},{"issue":"1","key":"e_1_2_9_43_2","first-page":"676","article-title":"Deep-drawing: A deep learning approach to graph drawing","volume":"26","author":"Wang Y.","year":"2019","journal-title":"IEEE Transactions on Visualization and Computer Graphics"},{"key":"e_1_2_9_44_2","doi-asserted-by":"publisher","DOI":"10.1109\/TVCG.2019.2934798"},{"issue":"2025","key":"e_1_2_9_45_2","doi-asserted-by":"crossref","first-page":"413","DOI":"10.1007\/s12650-025-01044-x","article-title":"A deep learning approach to evaluate the quality of graph layouts using GNN","volume":"28","author":"Wu X.","year":"2025","journal-title":"Journal of Visualization"},{"key":"e_1_2_9_46_2","doi-asserted-by":"crossref","unstructured":"WuA. XieL. LeeB. WangY. CuiW. QuH.: Learning to automate chart layout configurations using crowdsourced paired comparison. InProceedings of the 2021 CHI Conference on Human Factors in Computing Systems(2021) pp.1\u201313. 4","DOI":"10.1145\/3411764.3445179"},{"key":"e_1_2_9_47_2","doi-asserted-by":"publisher","DOI":"10.1109\/MCG.2021.3093908"},{"key":"e_1_2_9_48_2","doi-asserted-by":"publisher","DOI":"10.1109\/TVCG.2023.3306356"},{"key":"e_1_2_9_49_2","doi-asserted-by":"crossref","first-page":"43","DOI":"10.1016\/j.visinf.2024.04.003","article-title":"Generative AI for Visualization: State of the Art and Future Directions","volume":"8","author":"Ye Y.","year":"2024","journal-title":"Visual Informatics"},{"key":"e_1_2_9_50_2","doi-asserted-by":"publisher","DOI":"10.1109\/TVCG.2018.2859997"}],"container-title":["Computer Graphics Forum"],"original-title":[],"language":"en","link":[{"URL":"https:\/\/onlinelibrary.wiley.com\/doi\/pdf\/10.1111\/cgf.70456","content-type":"application\/pdf","content-version":"vor","intended-application":"text-mining"},{"URL":"https:\/\/onlinelibrary.wiley.com\/doi\/full-xml\/10.1111\/cgf.70456","content-type":"application\/xml","content-version":"vor","intended-application":"text-mining"},{"URL":"https:\/\/onlinelibrary.wiley.com\/doi\/pdf\/10.1111\/cgf.70456","content-type":"unspecified","content-version":"vor","intended-application":"similarity-checking"}],"deposited":{"date-parts":[[2026,6,18]],"date-time":"2026-06-18T06:03:32Z","timestamp":1781762612000},"score":1,"resource":{"primary":{"URL":"https:\/\/onlinelibrary.wiley.com\/doi\/10.1111\/cgf.70456"}},"subtitle":[],"short-title":[],"issued":{"date-parts":[[2026,6,17]]},"references-count":51,"alternative-id":["10.1111\/cgf.70456"],"URL":"https:\/\/doi.org\/10.1111\/cgf.70456","archive":["Portico"],"relation":{},"ISSN":["0167-7055","1467-8659"],"issn-type":[{"value":"0167-7055","type":"print"},{"value":"1467-8659","type":"electronic"}],"subject":[],"published":{"date-parts":[[2026,6,17]]},"assertion":[{"value":"2026-06-17","order":3,"name":"published","label":"Published","group":{"name":"publication_history","label":"Publication History"}}],"article-number":"e70456"}}