Elamur Research⌕
☰
home / papers / 2026 / KidsArtBench: Multi-Dimensional Children's Art Evaluation with Attribute-Aware MLLMs

KidsArtBench: Multi-Dimensional Children's Art Evaluation with Attribute-Aware MLLMs

Paper · 2026 · Mingrui Ye, Chanjin Zheng, Zengyi Yu et al.

Summary

Multimodal LLMs remain weak at judging artistic expression. Aesthetic qualities are abstract, and annotated artwork data is rare. Existing aesthetics datasets usually give a single overall score and focus on adult photos or art. The authors address this gap with KidsArtBench, a set of 1,046 drawings by children aged 5–15, gathered mainly from schools in Eastern China. Experienced art educators scored each piece on nine rubric dimensions, such as Realism, Color Richness, Imagination and Transformation. Some works also carry written formative comments in Chinese, with English translations.

The authors first prompted open-source MLLMs with the rubric and found them clearly weak. They then fine-tuned Qwen2.5-VL-7B with one LoRA adapter per rubric dimension. Regression-Aware Fine-Tuning (RAFT) and Regression-Aware Inference (RAIL) keep the outputs consistent with the ordinal 1–5 scale. This raised average correlation with expert scores from 0.468 to 0.653. Perceptual dimensions improved most, while texture- and concept-related ones stayed harder. The resource offers an open, educator-grounded testbed for formative feedback on creative work.

Key points

  • Dataset and annotation: KidsArtBench has 1,046 artworks. Each dimension is scored 1–5 by at least two of 12 trained educators, and a senior expert resolves disagreements. The data is split 80/10/10 into train, validation and test sets.
  • Rubric: There are nine dimensions grouped into four categories: Formative Creativity, Color Expressiveness, Line Work Richness and Conceptual Thinking. Most scores fall around 3–4.
  • Method: The model uses attribute-specific multi-LoRA with RAFT training and RAIL decoding. On Qwen2.5-VL-7B, correlation rose from 0.468 to 0.653.
  • Error analysis:
    • Artworks from lower grades make up about 60% of the data but account for 88% of high-error cases.
    • Transformation and Line Texture have the highest error rates (61.1% each), compared with 13.9% for Realism.
    • Most errors deviate by only ±1 from the true score.
  • Release: Data and code are available at github.com/bigrayss/KidsArtBench under the MIT License. Use is restricted to research; re-identification and high-stakes grading are prohibited.

Our summary of the paper, in our own words — not its text. The work belongs to its authors; read it at the links above.

 Fields, topics and concepts

📚 Shelves 6The field of science this work belongs to (OpenAlex fields).Computer Science · 16,667 worksArtificial Intelligence · 16,194 worksArts and Humanities · 16,039 worksLanguage and Linguistics · 15,891 worksComputer Vision and Pattern Recognition · 2,869 worksVisual Arts and Performing Arts · 5 works

Cite

Cite as Mingrui Ye, Chanjin Zheng, Zengyi Yu, Chenyu Xiang, Zhixue Zhao, Zheng Yuan and Helen Yannakoudakis. 2026. KidsArtBench: Multi-Dimensional Children's Art Evaluation with Attribute-Aware MLLMs. Proceedings of the 19th Conference of the European Chapter of the Association for Computational Linguistics (Volume 1: Long Papers).

The paper itself, not this page. BibTeX from the library's record of the work. DOI 10.18653/v1/2026.eacl-long.267

@inproceedings{ye2026kidsartbench,
  title = {{KidsArtBench: Multi-Dimensional Children's Art Evaluation with Attribute-Aware MLLMs}},
  author = {Mingrui Ye, Chanjin Zheng, Zengyi Yu, Chenyu Xiang, Zhixue Zhao, Zheng Yuan and Helen Yannakoudakis},
  booktitle = {Proceedings of the 19th Conference of the European Chapter of the Association for Computational Linguistics (Volume 1: Long Papers)},
  year = {2026},
  doi = {10.18653/v1/2026.eacl-long.267},
  url = {https://doi.org/10.18653/v1/2026.eacl-long.267}
}

References

49 references, 9 of them with a card of ours. ↗ leads to the work, «find» searches for one we have no id for yet.

  1. Josh Achiam, Steven Adler, Sandhini Agarwal, Lama Ahmad, Ilge Akkaya, Florencia Leoni Aleman, Diogo Almeida, Janko Altenschmidt, Sam Altman, Shyamal Anadkat, and 1 others. 2023. Gpt-4 technical report. arXiv preprint arXiv:2303.08774.
  2. Ahmad Mustafa Anis, Hasnain Ali, and Saquib Sarfraz. 2025. On the limitations of vision-language models in understanding image transforms. arXiv preprint arXiv:2503.09837. — [arXiv:2503.09837](https://arxiv.org/abs/2503.09837) find
  3. Shuai Bai, Keqin Chen, Xuejing Liu, Jialin Wang, Wenbin Ge, Sibo Song, Kai Dang, Peng Wang, Shijie Wang, Jun Tang, and 1 others. 2025. Qwen2. 5-vl technical report. arXiv preprint arXiv:2502.13923. — Qwen2.5-VL Technical Report
  4. Arne Bewersdorff, Kathrin Seßler, Armin Baur, Enkelejda Kasneci, and Claudia Nerdel. 2023. Assessing student errors in experimentation using artificial intelligence and large language models: A comparative study with human raters. Computers and Education: Artificial Intelligence, 5:100177. find
  5. Anjanava Biswas and Wrick Talukdar. 2024. Robustness of structured data extraction from in-plane rotated documents using multi-modal large language models (llm). Journal of Artificial Intelligence Research. find
  6. Moinak Biswas. 2021. Realism. BioScope: South Asian Screen Studies, 12(1-2):158–161. find
  7. Kuang-Yu Chang, Kung-Hung Lu, and Chu-Song Chen. 2017. Aesthetic critiques generation for photos. In Proceedings of the IEEE international conference on computer vision, pages 3514–3523. find
  8. Olga Denac and 1 others. 2014. The significance and role of aesthetic education in schooling. Creative education, 5(19):1714. find
  9. Keyan Ding, Kede Ma, Shiqi Wang, and Eero P Simoncelli. 2020. Image quality assessment: Unifying structure and texture similarity. IEEE transactions on pattern analysis and machine intelligence, 44(5):2567–2581. find
  10. Yingbei Du. 2020. Research on the transformation and innovation of visual art design form based on digital fusion technology. Applied Mathematics and Nonlinear Sciences. find
  11. Noa Garcia and George Vogiatzis. 2018. How to read paintings: semantic art understanding with multimodal retrieval. In Proceedings of the European Conference on Computer Vision (ECCV) Workshops, pages 0–0. find
  12. Edward J Hu, Yelong Shen, Phillip Wallis, Zeyuan Allen-Zhu, Yuanzhi Li, Shean Wang, Lu Wang, Weizhu Chen, and 1 others. 2022. Lora: Low-rank adaptation of large language models. ICLR, 1(2):3.
  13. Yipo Huang, Xiangfei Sheng, Zhichao Yang, Quan Yuan, Zhichao Duan, Pengfei Chen, Leida Li, Weisi Lin, and Guangming Shi. 2024a. Aesexpert: Towards multi-modality foundation model for image aesthetics perception. In Proceedings of the 32nd ACM International Conference on Multimedia, pages 5911–5920. find
  14. Yipo Huang, Quan Yuan, Xiangfei Sheng, Zhichao Yang, Haoning Wu, Pengfei Chen, Yuzhe Yang, Leida Li, and Weisi Lin. 2024b. Aesbench: An expert benchmark for multimodal large language models on image aesthetics perception. arXiv preprint arXiv:2401.08276. — [arXiv:2401.08276](https://arxiv.org/abs/2401.08276) find
  15. Gabriel Ilharco, Marco Tulio Ribeiro, Mitchell Wortsman, Ludwig Schmidt, Hannaneh Hajishirzi, and Ali Farhadi. 2023. Editing models with task arithmetic. In The Eleventh International Conference on Learning Representations.
  16. Shiqi Jiang, Ning Li, Chen Shi, Liping Guo, Changbo Wang, and Chenhui Li. 2024. Aacp: Aesthetics assessment of children’s paintings based on self-supervised learning. In Proceedings of the AAAI Conference on Artificial Intelligence, volume 38, pages 2534–2542. find
  17. Shu Kong, Xiaohui Shen, Zhe Lin, Radomir Mech, and Charless Fowlkes. 2016. Photo aesthetics ranking network with attributes and content adaptation. In Computer Vision–ECCV 2016: 14th European Conference, Amsterdam, The Netherlands, October 11– 14, 2016, Proceedings, Part I 14, pages 662–679. Springe… find
  18. Ehsan Latif and Xiaoming Zhai. 2024. Fine-tuning chatgpt for automatic scoring. Computers and Education: Artificial Intelligence, 6:100210. find
  19. Unggi Lee, Minji Jeon, Yunseo Lee, Gyuri Byun, Yoorim Son, Jaeyoon Shin, Hongkyu Ko, and Hyeoncheol Kim. 2024. Llava-docent: Instruction tuning with multimodal large language model to support art appreciation education. Computers and Education: Artificial Intelligence, 7:100297. find
  20. Zongxia Li, Xiyang Wu, Hongyang Du, Huy Nghiem, and Guangyao Shi. 2025. Benchmark evaluations, applications, and challenges of large vision language models: A survey. arXiv preprint arXiv:2501.02189. — [arXiv:2501.02189](https://arxiv.org/abs/2501.02189) find
  21. Paul J Locher, Pieter Jan Stappers, and Kees Overbeeke. 1999. An empirical evaluation of the visual rightness theory of pictorial composition. Acta psychologica, 103(3):261–280. find
  22. Michelle J Searle and Lyn M Shulha. 2016. Capturing the imagination: Arts-informed inquiry as a method in program evaluation. Canadian Journal of Program Evaluation, 31(1):34–60. find
  23. Peng Lu, Zhijie Kuang, Xujun Peng, and Ruifan Li. 2015. Discovering harmony: A hierarchical colour harmony model for aesthetics assessment. In Computer Vision–ACCV 2014: 12th Asian Conference on Computer Vision, Singapore, Singapore, November 1-5, 2014, Revised Selected Papers, Part III 12, pages 45… find
  24. Woosuk Seo, Joonyoung Jun, Minki Chun, Hyeonhak Jeong, Sungmin Na, Woohyun Cho, Saeri Kim, and Hyunggu Jung. 2022. Toward an ai-assisted assessment tool to support online art therapy practices: A pilot study. In Proceedings of 20th European Conference on Computer-Supported Cooperative Work. European… find
  25. Michal Lukasik, Zhao Meng, Harikrishna Narasimhan, Yin-Wen Chang, Aditya Krishna Menon, Felix Yu, and Sanjiv Kumar. 2025. Better autoregressive regression with llms via regression-aware fine-tuning. In The Thirteenth International Conference on Learning Representations. find
  26. S Sfarra, C Ibarra-Castanedo, D Ambrosini, D Paoletti, A Bendada, and X Maldague. 2014. Discovering the defects in paintings using non-destructive testing (ndt) techniques and passing through measurements of deformation. Journal of Nondestructive Evaluation, 33:358–383. find
  27. Michal Lukasik, Harikrishna Narasimhan, Aditya Krishna Menon, Felix Yu, and Sanjiv Kumar. 2024. Regression aware inference with LLMs. In Findings of the Association for Computational Linguistics: EMNLP 2024, pages 13667–13678, Miami, Florida, USA. Association for Computational Linguistics. find
  28. Dongyu She, Yu-Kun Lai, Gaoxiong Yi, and Kun Xu. 2021. Hierarchical layout-aware graph convolutional network for unified aesthetics assessment. In Proceedings of the IEEE/CVF conference on computer vision and pattern recognition, pages 8475–8484. find
  29. Hui Mao, Ming Cheung, and James She. 2017. Deepart: Learning joint representations of visual arts. In Proceedings of the 25th ACM international conference on Multimedia, pages 1183–1191. find
  30. Xiangfei Sheng, Leida Li, Pengfei Chen, Jinjian Wu, Weisheng Dong, Yuzhe Yang, Liwu Xu, Yaqian Li, and Guangming Shi. 2023. Aesclip: Multi-attribute contrastive learning for image aesthetics assessment. In Proceedings of the 31st ACM International Conference on Multimedia, pages 1117–1126. find
  31. Louie Meyer, Johanne Engel Aaen, Anitamalina Regitse Tranberg, Peter Kun, Matthias Freiberger, Sebastian Risi, and Anders Sundnes Løvlie. 2024. Algorithmic ways of seeing: Using object detection to facilitate art exploration. In Proceedings of the 2024 CHI Conference on Human Factors in Computing Sy… find
  32. Naila Murray, Luca Marchesotti, and Florent Perronnin. 2012. Ava: A large-scale database for aesthetic visual analysis. In 2012 IEEE conference on computer vision and pattern recognition, pages 2408–2415. IEEE. find
  33. Oksana Pylypchuk, Andrii Polubok, Olga Krivenko, Olena Safronova, Danylo Kosenko, and Nataliia Avdieieva. 2021. Developing an approach to colour assessment of works of art on aim to creating a comfortable and harmonious interior. In 2021 International Conference on Social Sciences and Big Data Appli… find
  34. Sue Robson and Victoria Rowe. 2012. Observing young children’s creative thinking: engagement, involvement and persistence. International Journal of Early Years Education, 20(4):349–364. find
  35. Güneş Salı, Aysel Köksal Akyol, and Gülen Baran. 2014. An analysis of pre-school children’s perception of schoolyard through their drawings. Procedia-Social and Behavioral Sciences, 116:2105–2114. find
  36. Gjorgji Strezoski and Marcel Worring. 2018. Omniart: A large-scale artistic benchmark. ACM Trans. Multimedia Comput. Commun. Appl., 14(4). find
  37. Gemini Team, Petko Georgiev, Ving Ian Lei, Ryan Burnell, Libin Bai, Anmol Gulati, and 1 others. 2024. Gemini 1.5: Unlocking multimodal understanding across millions of tokens of context, 2024. URL https://arxiv.org/abs/2403.05530. ↗
  38. Gemma Team, Aishwarya Kamath, Johan Ferret, Shreya Pathak, Nino Vieillard, Ramona Merhej, Sarah Perrin, Tatiana Matejovicova, Alexandre Ramé, Morgane Rivière, and 1 others. 2025a. Gemma 3 technical report. arXiv preprint arXiv:2503.19786. find
  39. Kimi Team, Angang Du, Bohong Yin, Bowei Xing, Bowen Qu, Bowen Wang, Cheng Chen, Chenlin Zhang, Chenzhuang Du, Chu Wei, and 1 others. 2025b. Kimi-vl technical report. arXiv preprint arXiv:2504.07491.
  40. Qwen Team. 2025. Qwen3 technical report. Preprint, arXiv:2505.09388.
  41. Yiming Wang, Yu Lin, Xiaodong Zeng, and Guannan Zhang. 2023. Multilora: Democratizing lora for better multi-task learning. arXiv preprint arXiv:2311.11501. — [arXiv:2311.11501](https://arxiv.org/abs/2311.11501) find
  42. LLM-Core-Team Xiaomi. 2025. Mimo-vl technical report. Preprint, arXiv:2506.03569.
  43. Weicheng Xing, Tianqing Zhu, Jenny Wang, and Bo Liu. 2024. A survey on mllms in education: Application and future directions. Future Internet, 16(12):1–31. find
  44. Lixiang Yan, Lele Sha, Linxuan Zhao, Yuheng Li, Roberto Martinez-Maldonado, Guanliang Chen, Xinyu Li, Yueqiao Jin, and Dragan Gašević. 2024. Practical and ethical challenges of large language models in education: A systematic scoping review. British Journal of Educational Technology, 55(1):90– 112. find
  45. Duzhen Zhang, Yahan Yu, Jiahua Dong, Chenxing Li, Dan Su, Chenhui Chu, and Dong Yu. 2024. MM-LLMs: Recent advances in MultiModal large language models. In Findings of the Association for Computational Linguistics: ACL 2024, pages 12401– 12430, Bangkok, Thailand. Association for Computational Linguis…
  46. Jiajing Zhang, Yongwei Miao, and Jinhui Yu. 2021. A comprehensive survey on computational aesthetic evaluation of visual art images: Metrics and challenges. IEEE Access, 9:77164–77187. find
  47. Liang Zhao, Eslam Hussam, Jin-Taek Seong, Assem Elshenawy, Mustafa Kamal, and Etaf Alshawarbeh. 2024. Revolutionizing art education: Integrating ai and multimedia for enhanced appreciation teaching. Alexandria Engineering Journal, 93:33–43. find
  48. Chanjin Zheng, Zengyi Yu, Yilin Jiang, Mingzi Zhang, Xunuo Lu, Jing Jin, and Liteng Gao. 2025. Artmentor: Ai-assisted evaluation of artworks to explore multimodal large language models capabilities. In Proceedings of the 2025 CHI Conference on Human Factors in Computing Systems, pages 1–18. find
  49. Hancheng Zhu, Yong Zhou, Leida Li, Yaqian Li, and Yandong Guo. 2021. Learning personalized image aesthetics from subjective and objective attributes. IEEE Transactions on Multimedia, 25:179–190. find

Updated

↑ back to top

This page in Markdown → https://md.research.elamur.ai/free/en/papers/ye-2026-kidsartbench-multi-dimensional-children-s/