Elamur Research⌕
☰
home / papers / 2026 / ChartVerse: Scaling Chart Reasoning via Reliable Programmatic Synthesis from Scratch

ChartVerse: Scaling Chart Reasoning via Reliable Programmatic Synthesis from Scratch

Paper · 2026 · Zheng Liu, Honglin Lin, Xiaoyang Wang et al.

Summary

Open-source vision-language models lag on chart reasoning, mainly because training data is weak. Synthetic charts tend to be simple and repetitive, and automatically generated QA pairs often contain hallucinated or numerically wrong answers. ChartVerse is a code-driven pipeline that builds both charts and reasoning data from scratch.

First, the authors define Rollout Posterior Entropy (RPE), a complexity score:

  • A small VLM reconstructs a chart as plotting code several times.
  • The rendered outputs are compared through CLIP embeddings.
  • Divergent reconstructions indicate a harder chart.

High-RPE charts seed a chart coder that is refined iteratively and samples new, varied chart programs.

Second, QA is built in reverse:

  • Answers are computed from the chart code itself.
  • Questions are then written to fit those answers and checked for consistency.
  • Pairs that a strong model rarely fails are discarded.
  • Chain-of-thought traces are distilled from Qwen3-VL-30B-A3B-Thinking.

Models trained on the resulting datasets beat their own teacher. This shows that verifiable synthetic data can push past the usual distillation limit.

Key points

  • RPE computation: Qwen3-VL-2B-Thinking produces 8 code rollouts at temperature 1.0. The renders are encoded with CLIP, and the spectral entropy of their centered Gram matrix gives the score.
  • Released datasets:
    • ChartVerse-SFT-600K: about 412K charts, 603K QA pairs and 3.9B tokens.
    • ChartVerse-RL-40K, a separate set for reinforcement-learning training.
    • SFT-600K scores highest of all compared corpora on RPE (0.44), color entropy (3.17) and semantic embedding spread (0.51).
  • Models and results: ChartVerse models at 2B, 4B and 8B are initialized from Qwen3-VL Instruct. ChartVerse-4B outperforms Qwen3-VL-8B-Thinking. ChartVerse-8B beats its 30B teacher and comes close to Qwen3-VL-32B-Thinking.
  • RPE as a data filter: With a fixed 100K-sample budget, RPE picks harder data than VLM-as-Judge or Python Code Complexity. Qwen3-VL-4B-Instruct fails 27.6% of RPE-selected samples, versus 21.1% and 23.5% for the other filters. Fine-tuning on RPE-selected data gives the best average score (55.4).
  • QA ablation (average score):
    • Image-based generation: 56.5.
    • Code-based generation: 56.8.
    • Truth-anchored synthesis: 57.4.
    • Truth-anchored synthesis plus fail-rate filtering: 57.8.
  • Resources: Code, model weights and datasets are published at chartverse.github.io.

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

 Institutions

🎓Peking UniversityBeijing · China244 works · 39,359 citations🎓Shanghai Jiao Tong UniversityShanghai · China149 works · 16,161 citations🔬Shanghai Artificial Intelligence LaboratoryShanghai · China117 works · 11,281 citations

 Fields, topics and concepts

📚 Shelves 5The 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 works

🧭 Topics 1Research subjects and languages the work is about.Reasoning · 1,426 works

Cite

Cite as Zheng Liu, Honglin Lin, Xiaoyang Wang, Xin Gao, Yu Li, Mengzhang Cai, Yun Zhu, Zhanping Zhong, Qizhi Pei, Zhuoshi Pan, Xiaoran Shang, Conghui He, Bin Cui, Wentao Zhang and Lijun Wu. 2026. ChartVerse: Scaling Chart Reasoning via Reliable Programmatic Synthesis from Scratch. Proceedings of the 64th Annual Meeting 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.acl-long.344

@inproceedings{liu2026chartverse,
  title = {{ChartVerse: Scaling Chart Reasoning via Reliable Programmatic Synthesis from Scratch}},
  author = {Zheng Liu, Honglin Lin, Xiaoyang Wang, Xin Gao, Yu Li, Mengzhang Cai, Yun Zhu, Zhanping Zhong, Qizhi Pei, Zhuoshi Pan, Xiaoran Shang, Conghui He, Bin Cui, Wentao Zhang and Lijun Wu},
  booktitle = {Proceedings of the 64th Annual Meeting of the Association for Computational Linguistics (Volume 1: Long Papers)},
  year = {2026},
  doi = {10.18653/v1/2026.acl-long.344},
  url = {https://doi.org/10.18653/v1/2026.acl-long.344}
}

References

54 references, 15 of them with a card of ours. ↗ leads to the work, «find» searches for one we have no id for yet.

  1. Anthropic. 2025. Claude 3.5 technical report. Technical report, Anthropic. find
  2. Shuai Bai, Yuxuan Cai, Ruizhe Chen, Keqin Chen, Xionghui Chen, Zesen Cheng, Lianghao Deng, Wei Ding, Chang Gao, Chunjiang Ge, Wenbin Ge, Zhifang Guo, Qidong Huang, Jie Huang, Fei Huang, Binyuan Hui, Shutong Jiang, Zhaohai Li, Mingsheng Li, and 45 others. 2025. Qwen3-vl technical report. Preprint, ar…
  3. Mengzhang Cai, Xin Gao, Yu Li, Honglin Lin, Zheng Liu, Zhuoshi Pan, Qizhi Pei, Xiaoran Shang, Mengyuan Sun, Zinan Tang, Xiaoyang Wang, Zhanping Zhong, Yun Zhu, Dahua Lin, Conghui He, and Lijun Wu. 2025. Opendataarena: A fair and open arena for benchmarking post-training dataset value. Preprint, arXi… — [arXiv:2512.14051](https://arxiv.org/abs/2512.14051) find
  4. Lei Chen, Xuanle Zhao, Zhixiong Zeng, Jing Huang, Yufeng Zhong, and Lin Ma. 2025. Chart-r1: Chain-of-thought supervision and reinforcement for advanced chart reasoner. Preprint, arXiv:2507.15509. — [arXiv:2507.15509](https://arxiv.org/abs/2507.15509) find
  5. Haodong Duan, Xinyu Fang, Junming Yang, Xiangyu Zhao, Yuxuan Qiao, Mo Li, Amit Agarwal, Zhe Chen, Lin Chen, Yuan Liu, Yubo Ma, Hailong Sun, Yifan Zhang, Shiyin Lu, Tack Hwa Wong, Weiyun Wang, Peiheng Zhou, Xiaozhe Li, Chaoyou Fu, and 13 others. 2025. Vlmevalkit: An open-source toolkit for evaluating… — [arXiv:2407.11691](https://arxiv.org/abs/2407.11691) find
  6. Google DeepMind. 2025. Gemini 3 technical report. Technical report, Google. find
  7. Yucheng Han, Chi Zhang, Xin Chen, Xu Yang, Zhibin Wang, Gang Yu, Bin Fu, and Hanwang Zhang. 2023. Chartllama: A multimodal llm for chart understanding and generation. Preprint, arXiv:2311.16483. find
  8. Wei He, Zhiheng Xi, Wanxu Zhao, Xiaoran Fan, Yiwen Ding, Zifei Shan, Tao Gui, Qi Zhang, and Xuanjing Huang. 2025. Distill visual chart reasoning ability from llms to mllms. Preprint, arXiv:2410.18798. find
  9. Muye Huang, Han Lai, Xinyu Zhang, Wenjun Wu, Jie Ma, Lingling Zhang, and Jun Liu. 2025. Evochart: A benchmark and a self-training approach towards real-world chart understanding. Preprint, arXiv:2409.01577. — [arXiv:2409.01577](https://arxiv.org/abs/2409.01577) find
  10. Binyuan Hui, Jian Yang, Zeyu Cui, Jiaxi Yang, Dayiheng Liu, Lei Zhang, Tianyu Liu, Jiajun Zhang, Bowen Yu, Keming Lu, Kai Dang, Yang Fan, Yichang Zhang, An Yang, Rui Men, Fei Huang, Bo Zheng, Yibo Miao, Shanghaoran Quan, and 5 others. 2024. Qwen2.5-coder technical report. Preprint, arXiv:2409.12186.
  11. Caijun Jia, Nan Xu, Jingxuan Wei, Qingli Wang, Lei Wang, Bihui Yu, and Junnan Zhu. 2025. Chartrea-soner: Code-driven modality bridging for long-chain reasoning in chart question answering. Preprint, arXiv:2506.10116. — [arXiv:2506.10116](https://arxiv.org/abs/2506.10116) find
  12. Gongyao Jiang and Qiong Luo. 2025. Chart-coca: Self-improving chart understanding of vision lms via code-driven synthesis and candidate-conditioned answering. Preprint, arXiv:2508.11975. — [arXiv:2508.11975](https://arxiv.org/abs/2508.11975) find
  13. Kushal Kafle, Brian Price, Scott Cohen, and Christopher Kanan. 2018. Dvqa: Understanding data visualizations via question answering. Preprint, arXiv:1801.08163. — [arXiv:1801.08163](https://arxiv.org/abs/1801.08163) find
  14. Samira Ebrahimi Kahou, Vincent Michalski, Adam Atkinson, Akos Kadar, Adam Trischler, and Yoshua Bengio. 2018. Figureqa: An annotated figure dataset for visual reasoning. Preprint, arXiv:1710.07300. find
  15. Jovana Kondic, Pengyuan Li, Dhiraj Joshi, Zexue He, Shafiq Abedin, Jennifer Sun, Ben Wiesel, Eli Schwartz, Ahmed Nassar, Bo Wu, Assaf Arbelle, Aude Oliva, Dan Gutfreund, Leonid Karlinsky, and Rogerio Feris. 2025. Chartgen: Scaling chart understanding via code-guided synthetic chart generation. Prepr… — [arXiv:2507.19492](https://arxiv.org/abs/2507.19492) find
  16. Woosuk Kwon, Zhuohan Li, Siyuan Zhuang, Ying Sheng, Lianmin Zheng, Cody Hao Yu, Joseph E. Gonzalez, Hao Zhang, and Ion Stoica. 2023. Efficient memory management for large language model serving with pagedattention. Preprint, arXiv:2309.06180.
  17. Honglin Lin, Zheng Liu, Yun Zhu, Chonghan Qin, Juekai Lin, Xiaoran Shang, Conghui He, Wentao Zhang, and Lijun Wu. 2026a. Mmfinereason: Closing the multimodal reasoning gap via open data-centric methods. arXiv preprint arXiv:2601.21821. — [arXiv:2601.21821](https://arxiv.org/abs/2601.21821) find
  18. Honglin Lin, Chonghan Qin, Zheng Liu, Qizhi Pei, Yu Li, Zhanping Zhong, Xin Gao, Yanfeng Wang, Conghui He, and Lijun Wu. 2026b. Scientific image synthesis: Benchmarking, methodologies, and downstream utility. arXiv preprint arXiv:2601.17027. — [arXiv:2601.17027](https://arxiv.org/abs/2601.17027) find
  19. Fuxiao Liu, Xiaoyang Wang, Wenlin Yao, Jianshu Chen, Kaiqiang Song, Sangwoo Cho, Yaser Yacoob, and Dong Yu. 2024. Mmc: Advancing multimodal chart understanding with large-scale instruction tuning. Preprint, arXiv:2311.10774.
  20. Shudong Liu, Hongwei Liu, Junnan Liu, Linchen Xiao, Songyang Gao, Chengqi Lyu, Yuzhe Gu, Wenwei Zhang, Derek F. Wong, Songyang Zhang, and Kai Chen. 2025a. Compassverifier: A unified and robust verifier for llms evaluation and outcome reward. Preprint, arXiv:2508.03686.
  21. Zheng Liu, Hao Liang, Bozhou Li, Wentao Xiong, Chong Chen, Conghui He, Wentao Zhang, and Bin Cui. 2025b. Synthvlm: Towards high-quality and efficient synthesis of image-caption datasets for vision-language models. In Proceedings of the 33rd ACM International Conference on Multimedia, pages 12807– 12… find
  22. Zheng Liu, Mengjie Liu, Jingzhou Chen, Jingwei Xu, Bin Cui, Conghui He, and Wentao Zhang. Flare: Fully integration of vision-language representations for deep cross-modal understanding. In The Fourteenth International Conference on Learning Representations. find
  23. Zheng Liu, Mengjie Liu, Jingzhou Chen, Jingwei Xu, Bin Cui, Conghui He, and Wentao Zhang. 2025c. Fusion: Fully integration of vision-language representations for deep cross-modal understanding. arXiv preprint arXiv:2504.09925. — [arXiv:2504.09925](https://arxiv.org/abs/2504.09925) find
  24. Zheng Liu, Mengjie Liu, Siwei Wen, Mengzhang Cai, Bin Cui, Conghui He, and Wentao Zhang. 2025d. From uniform to heterogeneous: Tailoring policy optimization to every token’s nature. arXiv preprint arXiv:2509.16591. — [arXiv:2509.16591](https://arxiv.org/abs/2509.16591) find
  25. Zhuoming Liu, Xiaofeng Gao, Feiyang Niu, Qiaozi Gao, Liu Liu, and Robinson Piramuthu. 2025e. Start: Spatial and textual learning for chart understanding. Preprint, arXiv:2512.07186. — [arXiv:2512.07186](https://arxiv.org/abs/2512.07186) find
  26. Pan Lu, Hritik Bansal, Tony Xia, Jiacheng Liu, Chunyuan Li, Hannaneh Hajishirzi, Hao Cheng, Kai-Wei Chang, Michel Galley, and Jianfeng Gao. 2023. Mathvista: Evaluating mathematical reasoning of foundation models in visual contexts. arXiv preprint arXiv:2310.02255.
  27. Ahmed Masry, Mohammed Saidul Islam, Mahir Ahmed, Aayush Bajaj, Firoz Kabir, Aaryaman Kartha, Md Tahmid Rahman Laskar, Mizanur Rahman, Shadikur Rahman, Mehrad Shahmohammadi, Megh Thakkar, Md Rizwan Parvez, Enamul Hoque, and Shafiq Joty. 2025. Chartqapro: A more diverse and challenging benchmark for c… find
  28. Ahmed Masry, Do Xuan Long, Jia Qing Tan, Shafiq Joty, and Enamul Hoque. 2022. Chartqa: A benchmark for question answering about charts with visual and logical reasoning. Preprint, arXiv:2203.10244.
  29. Ahmed Masry, Mehrad Shahmohammadi, Md Rizwan Parvez, Enamul Hoque, and Shafiq Joty. 2024a. Chartinstruct: Instruction tuning for chart comprehension and reasoning. Preprint, arXiv:2403.09028. find
  30. Ahmed Masry, Megh Thakkar, Aayush Bajaj, Aaryaman Kartha, Enamul Hoque, and Shafiq Joty. 2024b. Chartgemma: Visual instruction-tuning for chart reasoning in the wild. Preprint, arXiv:2407.04172. find
  31. Fanqing Meng, Wenqi Shao, Quanfeng Lu, Peng Gao, Kaipeng Zhang, Yu Qiao, and Ping Luo. 2024. Chartassisstant: A universal chart multimodal language model via chart-to-table pre-training and multitask instruction tuning. Preprint, arXiv:2401.02384. — [arXiv:2401.02384](https://arxiv.org/abs/2401.02384) find
  32. Nitesh Methani, Pritha Ganguly, Mitesh M. Khapra, and Pratyush Kumar. 2020. Plotqa: Reasoning over scientific plots. Preprint, arXiv:1909.00997. — [arXiv:1909.00997](https://arxiv.org/abs/1909.00997) find
  33. Junbo Niu, Zheng Liu, Zhuangcheng Gu, Bin Wang, Linke Ouyang, Zhiyuan Zhao, Tao Chu, Tianyao He, Fan Wu, Qintong Zhang, and 1 others. 2025. Mineru2. 5: A decoupled vision-language model for efficient high-resolution document parsing. arXiv preprint arXiv:2509.22186. — MinerU2.5: A Decoupled Vision-Language Model for Efficient High-Resolution Document Parsing
  34. OpenAI. 2025. Gpt-5 system card. Technical report, OpenAI. find
  35. Alec Radford, Jong Wook Kim, Chris Hallacy, Aditya Ramesh, Gabriel Goh, Sandhini Agarwal, Girish Sastry, Amanda Askell, Pamela Mishkin, Jack Clark, Gretchen Krueger, and Ilya Sutskever. 2021. Learning transferable visual models from natural language supervision. Preprint, arXiv:2103.00020.
  36. Guangming Sheng, Chi Zhang, Zilingfeng Ye, Xibin Wu, Wang Zhang, Ru Zhang, Yanghua Peng, Haibin Lin, and Chuan Wu. 2025. Hybridflow: A flexible and efficient rlhf framework. In Proceedings of the Twentieth European Conference on Computer Systems, EuroSys ’25, page 1279–1297. ACM.
  37. Liyan Tang, Grace Kim, Xinyu Zhao, Thom Lake, Wenxuan Ding, Fangcong Yin, Prasann Singhal, Manya Wadhwa, Zeyu Leo Liu, Zayne Sprague, Ramya Namuduri, Bodun Hu, Juan Diego Rodriguez, Puyuan Peng, and Greg Durrett. 2025. Chartmuseum: Testing visual reasoning capabilities of large vision-language model… — [arXiv:2505.13444](https://arxiv.org/abs/2505.13444) find
  38. Weiyun Wang, Zhangwei Gao, Lixin Gu, Hengjun Pu, Long Cui, Xingguang Wei, Zhaoyang Liu, Linglin Jing, Shenglong Ye, Jie Shao, Zhaokai Wang, Zhe Chen, Hongjie Zhang, Ganlin Yang, Haomin Wang, Qi Wei, Jinhui Yin, Wenhao Li, Erfei Cui, and 56 others. 2025. Internvl3.5: Advancing open-source multimodal…
  39. Zirui Wang, Mengzhou Xia, Luxi He, Howard Chen, Yitao Liu, Richard Zhu, Kaiqu Liang, Xindi Wu, Haotian Liu, Sadhika Malladi, Alexis Chevalier, Sanjeev Arora, and Danqi Chen. 2024. Charxiv: Charting gaps in realistic chart understanding in multimodal llms. Preprint, arXiv:2406.18521. find
  40. Renqiu Xia, Bo Zhang, Hancheng Ye, Xiangchao Yan, Qi Liu, Hongbin Zhou, Zijun Chen, Peng Ye, Min Dou, Botian Shi, Junchi Yan, and Yu Qiao. 2025. Chartx & chartvlm: A versatile benchmark and foundation model for complicated chart reasoning. Preprint, arXiv:2402.12185. — [arXiv:2402.12185](https://arxiv.org/abs/2402.12185) find
  41. Yijia Xiao, Edward Sun, Tianyu Liu, and Wei Wang. 2024. Logicvista: Multimodal llm logical reasoning benchmark in visual contexts. arXiv preprint arXiv:2407.04973. — [arXiv:2407.04973](https://arxiv.org/abs/2407.04973) find
  42. Duo Xu, Hao Cheng, Xin Lin, Zhen Xie, and Hao Wang. 2025a. Chartm 3 : A multi-stage code-driven pipeline for constructing multi-dimensional and multi-step visual reasoning data in chart comprehension. Preprint, arXiv:2511.02415. find
  43. Weiye Xu, Jiahao Wang, Weiyun Wang, Zhe Chen, Wengang Zhou, Aijun Yang, Lewei Lu, Houqiang Li, Xiaohua Wang, Xizhou Zhu, and 1 others. 2025b. Visulogic: A benchmark for evaluating visual reasoning in multi-modal large language models. arXiv preprint arXiv:2504.15279. — [arXiv:2504.15279](https://arxiv.org/abs/2504.15279) find
  44. Zhengzhuo Xu, Sinan Du, Yiyan Qi, Chengjin Xu, Chun Yuan, and Jian Guo. 2024. Chartbench: A benchmark for complex visual reasoning in charts. Preprint, arXiv:2312.15915. — [arXiv:2312.15915](https://arxiv.org/abs/2312.15915) find
  45. Yue Yang, Ajay Patel, Matt Deitke, Tanmay Gupta, Luca Weihs, Andrew Head, Mark Yatskar, Chris Callison-Burch, Ranjay Krishna, Aniruddha Kembhavi, and Christopher Clark. 2025a. Scaling text-rich image understanding via code-guided synthetic multimodal data generation. Preprint, arXiv:2502.14846. find
  46. Yuwei Yang, Zeyu Zhang, Yunzhong Hou, Zhuowan Li, Gaowen Liu, Ali Payani, Yuan-Sen Ting, and Liang Zheng. 2025b. Effective training data synthesis for improving mllm chart understanding. Preprint, arXiv:2508.06492. — [arXiv:2508.06492](https://arxiv.org/abs/2508.06492) find
  47. Kaichen Zhang, Keming Wu, Zuhao Yang, Kairui Hu, Bin Wang, Ziwei Liu, Xingxuan Li, and Lidong Bing. 2025. Openmmreasoner: Pushing the frontiers for multimodal reasoning with an open and general recipe. arXiv preprint arXiv:2511.16334. — [arXiv:2511.16334](https://arxiv.org/abs/2511.16334) find
  48. Liang Zhang, Anwen Hu, Haiyang Xu, Ming Yan, Yichen Xu, Qin Jin, Ji Zhang, and Fei Huang. 2024a. Tinychart: Efficient chart understanding with visual token merging and program-of-thoughts learning. Preprint, arXiv:2404.16635. — [arXiv:2404.16635](https://arxiv.org/abs/2404.16635) find
  49. Renrui Zhang, Dongzhi Jiang, Yichi Zhang, Haokun Lin, Ziyu Guo, Pengshuo Qiu, Aojun Zhou, Pan Lu, Kai-Wei Chang, Yu Qiao, and 1 others. 2024b. Mathverse: Does your multi-modal llm truly see the diagrams in visual math problems? In European Conference on Computer Vision, pages 169–186. Springer.
  50. Xuanle Zhao, Xianzhen Luo, Qi Shi, Chi Chen, Shuo Wang, Zhiyuan Liu, and Maosong Sun. 2025. Chartcoder: Advancing multimodal large language model for chart-to-code generation. Preprint, arXiv:2501.06598.
  51. Chujie Zheng, Shixuan Liu, Mingze Li, Xiong-Hui Chen, Bowen Yu, Chang Gao, Kai Dang, Yuqiong Liu, Rui Men, An Yang, Jingren Zhou, and Junyang Lin. 2025. Group sequence policy optimization. Preprint, arXiv:2507.18071.
  52. Yaowei Zheng, Richong Zhang, Junhao Zhang, Yanhan Ye, Zheyan Luo, Zhangchi Feng, and Yongqiang Ma. 2024. Llamafactory: Unified efficient fine-tuning of 100+ language models. Preprint, arXiv:2403.13372.
  53. Yun Zhu, Xiaoran Shang, Qifan Yu, Kaihang Pan, Boci Peng, Xiaotang Wang, Xiaoyang Wang, Yu Li, Honglin Lin, Zheng Liu, and 1 others. 2026. From factors to methods: A comprehensive survey on visual instruction tuning data selection. Authorea Preprints. find
  54. Chengke Zou, Xingang Guo, Rui Yang, Junyu Zhang, Bin Hu, and Huan Zhang. 2024. Dynamath: A dynamic visual benchmark for evaluating mathematical reasoning robustness of vision language models. arXiv preprint arXiv:2411.00836. find

Updated

↑ back to top

This page in Markdown → https://md.research.elamur.ai/free/en/papers/liu-2026-chartverse-scaling-chart-reasoning-reliable/