pith. machine review for the scientific record. sign in

arxiv: 2604.15813 · v1 · submitted 2026-04-17 · 💻 cs.DB

Recognition: unknown

Exploring Agentic Visual Analytics: A Co-Evolutionary Framework of Roles and Workflows

Tianqi Luo , Leixian Shen , Yuyu Luo
Authors on Pith no claims yet

Pith reviewed 2026-05-10 07:30 UTC · model grok-4.3

classification 💻 cs.DB
keywords agentic visual analyticsLLM-driven agentsrole-workflow taxonomyvisual analytics pipelinehuman-AI collaborationautonomy levelsco-evolutionary frameworkdesign guidelines
0
0 comments X

The pith

A co-evolutionary framework tracks how rising agent autonomy in visual analytics forces humans to move from direct operators to strategic supervisors.

A machine-rendered reading of the paper's core claim, the machinery that carries it, and where it could break.

The paper surveys 55 agentic visual analytics systems built around large language model agents that can plan, run, check, and refine data visualizations on their own. It introduces a framework that examines how greater agent independence changes the human's part in the process from hands-on work to high-level direction. The framework includes a taxonomy of four agent roles—planner, creator, reviewer, and context manager—placed against the standard stages of a visual analytics pipeline. Analysis of the systems reveals consistent patterns of trade-offs in how much control agents take versus how much humans retain. From this the authors extract practical guidelines for designing such systems and point to open questions for further study.

Core claim

The paper's central claim is that agentic visual analytics requires a co-evolutionary view in which agent autonomy and human roles develop together. Surveying 55 systems shows that agents take on four distinct roles—PLANNER for setting goals and steps, CREATOR for generating visualizations and code, REVIEWER for evaluating outputs, and CONTEXT MANAGER for maintaining data and history—and that these roles align with established visual analytics pipeline stages. The framework identifies recurring trade-offs along autonomy level, role assignment, and workflow structure, which in turn support concrete design guidelines.

What carries the argument

The role-workflow taxonomy that defines four agentic roles (PLANNER, CREATOR, REVIEWER, CONTEXT MANAGER) and aligns each with stages of the traditional visual analytics pipeline.

If this is right

  • Higher agent autonomy shifts humans from performing low-level operations to setting goals and reviewing results.
  • The four roles together cover the complete visual analytics pipeline from initial planning to ongoing context handling.
  • Designers must navigate explicit trade-offs between autonomy, role distribution, and workflow integration.
  • The observed patterns yield a set of actionable guidelines for building future agentic systems.
  • Future work can follow the paper's outlined research directions to extend the framework.

Where Pith is reading between the lines

These are editorial extensions of the paper, not claims the author makes directly.

  • The taxonomy could be applied directly to classify and compare any new agentic system released after the survey.
  • Context management may prove especially important for long-running sessions that maintain data history across multiple interactions.
  • The co-evolutionary lens might extend to other agentic domains such as code generation or scientific discovery pipelines.
  • Testing the framework against real user studies would reveal whether the predicted role shifts match observed behavior.

Load-bearing premise

The 55 systems chosen for the survey represent the full range of current agentic visual analytics work and the four-role taxonomy captures all essential role dynamics without major omissions.

What would settle it

A newly released agentic visual analytics system that performs core tasks through a role outside the four defined categories or in which human involvement does not decrease as agent autonomy rises.

Figures

Figures reproduced from arXiv: 2604.15813 by Leixian Shen, Tianqi Luo, Yuyu Luo.

Figure 1
Figure 1. Figure 1: Framework of agentic visual analytics systems, illustrating how four agentic roles, i.e., Planner, Creator, Reviewer, and Context [PITH_FULL_IMAGE:figures/full_fig_p001_1.png] view at source ↗
Figure 2
Figure 2. Figure 2: The Co-Evolutionary Framework: as AI agent roles evolve from simple assistance (Level 1) to strategic orchestration (Level 4), human roles correspondingly shift from direct command to high-level supervision. The integration of agentic capabilities introduces novel computational challenges, such as managing persistent analytical context, grounding multi-modal visual feedback, and mitigating code hallucinati… view at source ↗
Figure 3
Figure 3. Figure 3: Paper count statistics by autonomous levels. Note that the data [PITH_FULL_IMAGE:figures/full_fig_p002_3.png] view at source ↗
Figure 4
Figure 4. Figure 4: Example corpus of paradigm innovations in agentic VA systems. (a) Input Data Schemas: Shifting from static serialization to programming [PITH_FULL_IMAGE:figures/full_fig_p005_4.png] view at source ↗
Figure 5
Figure 5. Figure 5: Representative Human-AI Interaction Interfaces in Agentic Visual Analytics. [PITH_FULL_IMAGE:figures/full_fig_p006_5.png] view at source ↗
read the original abstract

Agentic visual analytics (VA) represents an emerging class of systems in which large language model (LLM)-driven agents autonomously plan, execute, evaluate, and iterate across the full visual analytics pipeline. By shifting users from low-level tool operations to high-level analytical goals expressed through natural language, these systems are fundamentally transforming how humans interact with data. However, the rapid proliferation of such systems in recent years has outpaced our understanding of their design landscape. Two intertwined problems remain open: how do autonomous agents reshape the traditional VA pipeline, and how must human involvement adapt as agent autonomy increases? To address these questions, this paper presents a comprehensive survey of 55 primary agentic VA systems and introduces a co-evolutionary framework. This framework is essential because it jointly analyzes the progression of agent autonomy alongside the necessary shift in human roles from manual operators to strategic supervisors. Within this framework, we define a role-workflow taxonomy that aligns four key agentic roles (PLANNER, CREATOR, REVIEWER, and CONTEXT MANAGER) and maps them onto established VA pipeline stages. Our analysis uncovers recurring trade-offs along three foundational axes: autonomy levels, agentic roles, and the VA workflow. We consolidate these findings into actionable design guidelines and outline future research directions for agentic visual analytics. A web-based interactive browser of our co-evolutionary framework, including the corpus and design guidelines, is available at agenticva.github.io/AgenticVA/.

Editorial analysis

A structured set of objections, weighed in public.

Desk editor's note, referee report, simulated authors' rebuttal, and a circularity audit. Tearing a paper down is the easy half of reading it; the pith above is the substance, this is the friction.

Referee Report

2 major / 2 minor

Summary. The paper surveys 55 primary agentic visual analytics systems and introduces a co-evolutionary framework that jointly analyzes the progression of agent autonomy with shifts in human roles from manual operators to strategic supervisors. It defines a role-workflow taxonomy aligning four agentic roles (PLANNER, CREATOR, REVIEWER, and CONTEXT MANAGER) with established VA pipeline stages, identifies recurring trade-offs along axes of autonomy levels, agentic roles, and VA workflows, consolidates findings into design guidelines, and provides an interactive web-based browser of the framework and corpus.

Significance. If the survey corpus is representative and the taxonomy logically consistent without major omissions, the framework would provide significant value by offering a structured, co-evolutionary lens for understanding how LLM-driven agents reshape visual analytics pipelines and human involvement. The open interactive resource strengthens its utility for researchers and practitioners seeking actionable design insights in this emerging area.

major comments (2)
  1. [Survey of 55 Systems] The section describing the survey of the 55 primary agentic VA systems provides no explicit inclusion/exclusion criteria, search strategy, databases queried, temporal bounds, or inter-rater reliability measures. This directly undermines the central claim that the four-role taxonomy and co-evolutionary framework comprehensively capture the field's dynamics, as selection bias could result in omitted roles or incomplete trade-off axes requiring framework revision.
  2. [Role-Workflow Taxonomy] The derivation of the role-workflow taxonomy (PLANNER, CREATOR, REVIEWER, CONTEXT MANAGER) and its mapping to VA pipeline stages lacks any reported validation process or quantitative assessment of coverage across the surveyed systems. Without this, it is unclear whether the taxonomy is exhaustive or if unaccounted roles/workflows exist that would necessitate revisions to the co-evolutionary framing.
minor comments (2)
  1. [Abstract] The abstract references the survey of 55 systems and the framework but omits any mention of the selection methodology or validation approach, which would better ground the claims for readers.
  2. [Web Resource] The paper could include a short description of the features and navigation of the interactive browser (agenticva.github.io/AgenticVA/) to help readers explore the corpus and guidelines without needing to visit the site immediately.

Simulated Author's Rebuttal

2 responses · 0 unresolved

We thank the referee for their thoughtful and constructive comments, which help us improve the transparency and rigor of the survey methodology and taxonomy derivation. We address each major comment below and will incorporate the suggested additions in the revised manuscript.

read point-by-point responses

  1. Referee: [Survey of 55 Systems] The section describing the survey of the 55 primary agentic VA systems provides no explicit inclusion/exclusion criteria, search strategy, databases queried, temporal bounds, or inter-rater reliability measures. This directly undermines the central claim that the four-role taxonomy and co-evolutionary framework comprehensively capture the field's dynamics, as selection bias could result in omitted roles or incomplete trade-off axes requiring framework revision.

    Authors: We acknowledge that the manuscript does not currently include an explicit methodology section detailing the survey process. In the revised version, we will add a dedicated 'Survey Methodology' section that specifies: the search strategy (keywords such as 'LLM agent visual analytics' and 'agentic data visualization', queried across Google Scholar, arXiv, ACM DL, and IEEE Xplore); inclusion criteria (systems that employ LLM-driven agents to autonomously perform tasks across the VA pipeline, published 2022–2024); exclusion criteria (non-LLM systems, purely theoretical work, or non-VA applications); temporal bounds; and the selection procedure, including duplicate removal and consensus-based inclusion decisions among authors. We will also report any measures of agreement. These additions will allow readers to evaluate potential bias and better support the framework's claims. revision: yes

  2. Referee: [Role-Workflow Taxonomy] The derivation of the role-workflow taxonomy (PLANNER, CREATOR, REVIEWER, CONTEXT MANAGER) and its mapping to VA pipeline stages lacks any reported validation process or quantitative assessment of coverage across the surveyed systems. Without this, it is unclear whether the taxonomy is exhaustive or if unaccounted roles/workflows exist that would necessitate revisions to the co-evolutionary framing.

    Authors: We agree that a more formal account of derivation and coverage would strengthen the presentation. The taxonomy emerged from iterative qualitative coding of agent behaviors observed across the 55 systems, aligned with standard VA pipeline stages. In revision, we will expand the relevant section to include: (1) a step-by-step description of the coding and mapping process with concrete system examples; (2) quantitative coverage metrics, such as the proportion of systems employing each role and a summary table of role-to-pipeline mappings; and (3) an explicit discussion of coverage, noting that every system in the corpus maps to at least one role while highlighting any edge cases or potential additional roles for future extension. This will provide evidence of the taxonomy's applicability within the surveyed set. revision: yes

Circularity Check

0 steps flagged

No circularity: framework derived inductively from external survey data

full rationale

The paper presents a survey of 55 external agentic VA systems followed by an inductive definition of its co-evolutionary framework and four-role taxonomy (PLANNER, CREATOR, REVIEWER, CONTEXT MANAGER) mapped to VA pipeline stages. No equations, fitted parameters, or predictions appear; the taxonomy is constructed from observed patterns in the surveyed corpus rather than by self-definition or renaming of inputs. No load-bearing self-citations or uniqueness theorems imported from prior author work are invoked to force the framework. The derivation remains self-contained against the external systems analyzed, with no reduction of claims to tautologies by construction.

Axiom & Free-Parameter Ledger

0 free parameters · 1 axioms · 2 invented entities

The paper rests on the domain assumption that established visual analytics pipeline stages remain a stable reference point for mapping new agent roles, and it introduces new conceptual entities (the framework and taxonomy) without independent falsifiable evidence beyond the survey itself.

axioms (1)
  • domain assumption Established VA pipeline stages provide a valid and complete foundation for mapping agentic roles.
    The framework explicitly maps roles onto these stages without re-deriving or validating the stages themselves.
invented entities (2)
  • Co-evolutionary framework no independent evidence
    purpose: Joint analysis of agent autonomy progression and human role shifts
    New conceptual structure proposed to address open problems in agentic VA design.
  • Role-workflow taxonomy (PLANNER, CREATOR, REVIEWER, CONTEXT MANAGER) no independent evidence
    purpose: Alignment of agent roles with VA pipeline stages
    Defined by authors from survey analysis.

pith-pipeline@v0.9.0 · 5561 in / 1429 out tokens · 85039 ms · 2026-05-10T07:30:48.635143+00:00 · methodology

discussion (0)

Sign in with ORCID, Apple, or X to comment. Anyone can read and Pith papers without signing in.

Reference graph

Works this paper leans on

74 extracted references · 60 canonical work pages · 2 internal anchors

  1. [2]

    Agrawal and M

    A. Agrawal and M. Scolnick. marimo: An open-source reactive notebook for Python. https://github.com/marimo-team/marimo, aug 2023. Apache-2.0 License. 8

    2023

  2. [3]

    Bromley and V

    D. Bromley and V . Setlur. DASH: A Bimodal Data Exploration Tool for Interactive Text and Visualizations. 3, 7

  3. [4]

    J. Chen, J. Wu, J. Guo, V . Mohanty, X. Li, J. P. Ono et al. InterChat: Enhancing Generative Visual Analytics using Multimodal Interactions. 44(3):e70112. doi: 10.1111/cgf.70112 2, 3, 4, 6, 7, 8, 9

    work page doi:10.1111/cgf.70112

  4. [5]

    N. Chen, Y . Zhang, J. Xu, K. Ren, and Y . Yang. VisEval: A Benchmark for Data Visualization in the Era of Large Language Models. 31(1):1301–

  5. [6]

    doi: 10.1109/TVCG.2024.3456320 3, 6

    work page doi:10.1109/tvcg.2024.3456320 2024

  6. [7]

    Chen and E

    Y . Chen and E. Wu. PI2: End-to-end Interactive Visualization Inter- face Generation from Queries. InProceedings of the 2022 International Conference on Management of Data, pp. 1711–1725. ACM. doi: 10. 1145/3514221.3526166 3, 4

    work page arXiv 2022

  7. [8]

    Z. Chen, J. Chen, S. Ö. Arik, M. Sra, T. Pfister, and J. Yoon. Coda: Agentic systems for collaborative data visualization, 2025. 3, 4, 5, 6, 7, 9

    2025

  8. [9]

    J. Choi, J. Lee, and J. Jo. Bavisitter: Integrating Design Guidelines into Large Language Models for Visualization Authoring. pp. 121–125. IEEE Computer Society. doi: 10.1109/VIS55277.2024.00032 3

    work page doi:10.1109/vis55277.2024.00032 2024

  9. [10]

    W. Cui. Visual analytics: A comprehensive overview.IEEE Access, 7:81555–81573, 2019. doi: 10.1109/ACCESS.2019.2923736 1

    work page doi:10.1109/access.2019.2923736 2019

  10. [11]

    A. K. Das and K. Mueller. Misvisfix: An interactive dashboard for detecting, explaining, and correcting misleading visualizations using large language models.IEEE Transactions on Visualization and Computer Graphics, 32(1):134–144, 2026. doi: 10.1109/TVCG.2025.3633884 4, 5, 8, 9

    work page doi:10.1109/tvcg.2025.3633884 2026

  11. [12]

    V . Dibia. LIDA: A Tool for Automatic Generation of Grammar-Agnostic Visualizations and Infographics using Large Language Models. In D. Bol- legala, R. Huang, and A. Ritter, eds.,Proceedings of the 61st Annual Meet- ing of the Association for Computational Linguistics (Volume 3: System Demonstrations), pp. 113–126. Association for Computational Linguistic...

    work page doi:10.18653/v1/2023.acl-demo.11 2023

  12. [13]

    Y . Fu, D. Bromley, and V . Setlur. DATAWEA VER: Authoring Data-Driven Narratives through the Integrated Composition of Visualization and Text. 44(3):e70098. doi: 10.1111/cgf.70098 3, 4, 7

    work page doi:10.1111/cgf.70098

  13. [14]

    T. Guo, X. Chen, Y . Wang, R. Chang, S. Pei, N. V . Chawla et al. Large Lan- guage Model based Multi-Agents: A Survey of Progress and Challenges. doi: 10.48550/arXiv.2402.01680 1

    work page internal anchor Pith review doi:10.48550/arxiv.2402.01680

  14. [15]

    P. F. Gyarmati, D. Moritz, T. Möller, and L. Koesten. A composable agentic system for automated visual data reporting, 2025. 3, 4, 6, 7, 8

    2025

  15. [16]

    P. Hao, D. Kang, N. Hinds, and Q. Wang. Flowforge: Guiding the creation of multi-agent workflows with design space visualization as a thinking scaffold.IEEE Transactions on Visualization and Computer Graphics, 32(1):1032–1042, 2026. doi: 10.1109/TVCG.2025.3634627 3, 4, 7

    work page doi:10.1109/tvcg.2025.3634627 2026

  16. [17]

    Hong and A

    M.-H. Hong and A. Crisan. Conversational AI Threads for Visualizing Multidimensional Datasets. doi: 10.48550/arXiv.2311.05590 3, 6

    work page doi:10.48550/arxiv.2311.05590

  17. [18]

    Huang, J

    H. Huang, J. Chen, C. Wang, and C. Li. SUPQA: LLM-based Geo- Visualization for Subjective Urban Performance Question-Answering. 44(3):e70106. doi: 10.1111/cgf.70106 3

    work page doi:10.1111/cgf.70106

  18. [19]

    Hutchinson, R

    M. Hutchinson, R. Jianu, A. Slingsby, and P. Madhyastha. Llm- assisted visual analytics: Opportunities and challenges.arXiv preprint arXiv:2409.02691, 2024. 1

    work page arXiv 2024

  19. [20]

    M. S. Islam, M. T. R. Laskar, M. R. Parvez, E. Hoque, and S. Joty. Data- Narrative: Automated Data-Driven Storytelling with Visualizations and Texts. doi: 10.48550/arXiv.2408.05346 2, 3, 4, 7

    work page doi:10.48550/arxiv.2408.05346

  20. [21]

    Ji and A

    T. Ji and A. A. Narechania. NL4DV-Stylist: Styling Data Visualiza- tions Using Natural Language and Example Charts. doi: 10.31219/osf. io/fs4en_v1 3

    work page doi:10.31219/osf

  21. [22]

    D. A. Keim, F. Mansmann, and J. Thomas. Visual analytics: how much visualization and how much analytics?SIGKDD Explor. Newsl., 11(2):5–8, May 2010. doi: 10.1145/1809400.1809403 1

    work page doi:10.1145/1809400.1809403 2010

  22. [23]

    Tjahjadi, Jiho Kim, Junpu Yu, Minji Park, Jiawen Zhang, Jon E

    N. Khanal, C. M. Yu, J.-C. Chiu, A. Chaudhary, Z. Zhang, K. Katija et al. FathomGPT: A natural language interface for interactively exploring ocean science data. InProceedings of the 37th Annual ACM Symposium on User Interface Software and Technology, pp. 1–15. ACM. doi: 10. 1145/3654777.3676462 3, 4, 5, 7

    work page arXiv

  23. [24]

    J. Kim, S. Lee, H. Jeon, K.-J. Lee, H.-J. Bae, B. Kim et al. PhenoFlow: A Human-LLM Driven Visual Analytics System for Exploring Large and Complex Stroke Datasets. 2, 3, 4, 6, 7, 8, 9

  24. [25]

    W. Koh, J. Yoon, M. Lee, Y . Song, J. Cho, J. Kang et al. C^2: Scalable Auto-Feedback for LLM-based Chart Generation. In L. Chiruzzo, A. Ritter, and L. Wang, eds.,Proceedings of the 2025 Conference of the Nations of the Americas Chapter of the Association for Computational Linguistics: Human Language Technologies (Volume 1: Long Papers), pp. 4525–4566. As...

    work page doi:10.18653/v1/2025.naacl 2025

  25. [26]

    G. Li, X. Wang, G. Aodeng, S. Zheng, Y . Zhang, C. Ou et al. Visualization Generation with Large Language Models: An Evaluation. doi: 10.48550/ arXiv.2401.11255 3

    work page arXiv

  26. [27]

    S. Li, X. Chen, Y . Song, Y . Song, and C. Zhang. Prompt4Vis: Prompting Large Language Models with Example Mining and Schema Filtering for Tabular Data Visualization. doi: 10.48550/arXiv.2402.07909 3, 4, 7

    work page doi:10.48550/arxiv.2402.07909

  27. [29]

    C. Liu, J. Yu, Y . Guo, J. Zhuang, Y . Luo, and X. Yuan. Breathing New Life into Existing Visualizations:A Natural Language-Driven Manipulation Framework. 3

  28. [30]

    S. Liu, H. Miao, Z. Li, M. Olson, V . Pascucci, and P.-T. Bremer. A V A: Towards Autonomous Visualization Agents through Visual Perception- Driven Decision-Making.Computer Graphics Forum, 2024. doi: 10. 1111/cgf.15093 3, 4, 5

    2024

  29. [31]

    T. Luo, C. Huang, L. Shen, B. Li, S. Shen, W. Zeng et al. nvBench 2.0: Re- solving Ambiguity in Text-to-Visualization through Stepwise Reasoning. 1, 3

  30. [32]

    Y . Luo, N. Tang, G. Li, J. Tang, C. Chai, and X. Qin. Natural Language to Visualization by Neural Machine Translation. 28(1):217–226. doi: 10. 1109/TVCG.2021.3114848 1, 3, 4

    work page arXiv 2021

  31. [33]

    S. L’Yi, A. v. d. Brandt, E. Adams, H. N. Nguyen, and N. Gehlenborg. Learnable and Expressive Visualization Authoring Through Blended Inter- faces. doi: 10.31219/osf.io/pjcn4 3

    work page doi:10.31219/osf.io/pjcn4

  32. [34]

    Maddigan and T

    P. Maddigan and T. Susnjak. Chat2VIS: Generating Data Visualizations via Natural Language Using ChatGPT, Codex and GPT-3 Large Language Models. 11:45181–45193. doi: 10.1109/ACCESS.2023.3274199 3, 4, 5

    work page doi:10.1109/access.2023.3274199 2023

  33. [35]

    Mitra, A

    R. Mitra, A. Narechania, A. Endert, and J. Stasko. Facilitating Conversa- tional Interaction in Natural Language Interfaces for Visualization. doi: 10.48550/arXiv.2207.00189 3

    work page doi:10.48550/arxiv.2207.00189

  34. [36]

    Moritz, C

    D. Moritz, C. Wang, G. Nelson, H. Lin, A. Smith, B. Howe et al. For- malizing visualization design knowledge as constraints: Actionable and extensible models in draco.IEEE Trans. Visualization & Comp. Graphics (Proc. InfoVis), 2019. doi: 10.1109/TVCG.2018.2865240 6

    work page doi:10.1109/tvcg.2018.2865240 2019

  35. [37]

    Narechania, A

    A. Narechania, A. Srinivasan, and J. Stasko. NL4DV: A Toolkit for Generating Analytic Specifications for Data Visualization from Natural Language Queries. 27(2):369–379. doi: 10.1109/TVCG.2020.3030378 1, 5

    work page doi:10.1109/tvcg.2020.3030378 2020

  36. [38]

    Ouyang, J

    G. Ouyang, J. Chen, Z. Nie, Y . Gui, Y . Wan, H. Zhang et al. nvAgent: Automated Data Visualization from Natural Language via Collaborative Agent Workflow. In W. Che, J. Nabende, E. Shutova, and M. T. Pilehvar, eds.,Proceedings of the 63rd Annual Meeting of the Association for Computational Linguistics (Volume 1: Long Papers), pp. 19534–19567. Association...

    work page doi:10.18653/v1/2025.acl 2025

  37. [39]

    M. J. Page, J. E. McKenzie, P. M. Bossuyt, I. Boutron, T. C. Hoffmann, C. D. Mulrow et al. The prisma 2020 statement: an updated guideline for reporting systematic reviews.BMJ, 372:n71, 2021. 2

    2020

  38. [40]

    Pesaran Zadeh, J

    F. Pesaran Zadeh, J. Kim, J.-H. Kim, and G. Kim. Text2Chart31: In- struction Tuning for Chart Generation with Automatic Feedback. In Y . Al-Onaizan, M. Bansal, and Y .-N. Chen, eds.,Proceedings of the 2024 Conference on Empirical Methods in Natural Language Processing, pp. 11459–11480. Association for Computational Linguistics. doi: 10.18653/ v1/2024.emnl...

    2024

  39. [41]

    Rahman, M

    M. Rahman, M. T. R. Laskar, S. Joty, and E. Hoque. Text2Vis: A Chal- lenging and Diverse Benchmark for Generating Multimodal Visualizations from Text. doi: 10.48550/arXiv.2507.19969 3

    work page doi:10.48550/arxiv.2507.19969

  40. [42]

    S. Sah, R. Mitra, A. Narechania, A. Endert, J. Stasko, and W. Dou. Generat- ing Analytic Specifications for Data Visualization from Natural Language Queries using Large Language Models. doi: 10.48550/arXiv.2408.13391 3

    work page doi:10.48550/arxiv.2408.13391

  41. [43]

    Setlur and S

    V . Setlur and S. Ridet. RÉCITKIT: A Spatial Toolkit for Designing and Evaluating Human-Centered Immersive Data Narratives. doi: 10.48550/ arXiv.2508.18670 3

    work page arXiv

  42. [44]

    Z. Shao, Y . Shan, Y . He, Y . Yao, J. Wang, X. Zhang et al. Do language model agents align with humans in rating visualizations? an empirical study.IEEE Computer Graphics and Applications, 45(6):14–28, 2025. doi: 10.1109/MCG.2025.3586461 3, 4, 6, 8

    work page doi:10.1109/mcg.2025.3586461 2025

  43. [45]

    L. Shen, E. Shen, Y . Luo, X. Yang, X. Hu, X. Zhang et al. Towards Natural Language Interfaces for Data Visualization: A Survey. 29(6):3121–3144. doi: 10.1109/TVCG.2022.3148007 2, 9

    work page doi:10.1109/tvcg.2022.3148007 2022

  44. [46]

    L. Shen, E. Shen, Z. Tai, Y . Song, and J. Wang. TaskVis: Task-oriented Visualization Recommendation. In M. Agus, C. Garth, and A. Kerren, eds.,EuroVis 2021 - Short Papers. The Eurographics Association, 2021. doi: 10.2312/evs.20211061 5

    work page doi:10.2312/evs.20211061 2021

  45. [47]

    Y . Shi, B. Li, Y . Luo, L. Chen, and N. Tang. Augmenting Realistic Charts with Virtual Overlays. InProceedings of the 2025 CHI Conference on Human Factors in Computing Systems, pp. 1–23. ACM. doi: 10.1145/ 3706598.3714320 3

    work page arXiv 2025

  46. [48]

    Shuai, B

    Z. Shuai, B. Li, S. Yan, Y . Luo, and W. Yang. DeepVIS: Bridging Natural Language and Data Visualization Through Step-wise Reasoning. doi: 10. 48550/arXiv.2508.01700 3, 7, 8

    work page arXiv

  47. [49]

    Y . Song, X. Zhao, and R. C.-W. Wong. Marrying Dialogue Systems with Data Visualization: Interactive Data Visualization Generation from Natural Language Conversations. InProceedings of the 30th ACM SIGKDD Conference on Knowledge Discovery and Data Mining, pp. 2733–2744. ACM. doi: 10.1145/3637528.3671935 3, 7

    work page doi:10.1145/3637528.3671935

  48. [50]

    Y . Song, X. Zhao, R. C.-W. Wong, and D. Jiang. RGVisNet: A Hybrid Retrieval-Generation Neural Framework Towards Automatic Data Visual- ization Generation. InProceedings of the 28th ACM SIGKDD Conference on Knowledge Discovery and Data Mining, pp. 1646–1655. ACM. doi: 10 .1145/3534678.3539330 3, 6

    work page arXiv

  49. [51]

    InProceedings of the 30th International Conference on Intelligent User Interfaces

    A. Srinivasan, V . Setlur, and A. Satyanarayan. Pluto: Authoring Se- mantically Aligned Text and Charts for Data-Driven Communication. In Proceedings of the 30th International Conference on Intelligent User Interfaces, pp. 1123–1140. ACM. doi: 10.1145/3708359.3712122 3

    work page doi:10.1145/3708359.3712122

  50. [52]

    two heads are better than one

    N. Sultanum and A. Srinivasan. DATATALES: Investigating the use of Large Language Models for Authoring Data-Driven Articles. In2023 IEEE Visualization and Visual Analytics (VIS), pp. 231–235. doi: 10. 1109/VIS54172.2023.00055 3, 6

    work page arXiv 2023

  51. [53]

    Tabalba, N

    R. Tabalba, N. Kirshenbaum, J. Leigh, A. Bhatacharya, A. Johnson, V . Grosso et al. Articulate+ : An Always-Listening Natural Language Interface for Creating Data Visualizations. InProceedings of the 4th Conference on Conversational User Interfaces, pp. 1–6. ACM. doi: 10. 1145/3543829.3544534 4

    work page arXiv

  52. [54]

    J. Tang, Y . Luo, M. Ouzzani, G. Li, and H. Chen. Sevi: Speech-to- Visualization through Neural Machine Translation. InProceedings of the 2022 International Conference on Management of Data, pp. 2353–2356. ACM. doi: 10.1145/3514221.3520150 3

    work page doi:10.1145/3514221.3520150 2022

  53. [55]

    Y . Tian, W. Cui, D. Deng, X. Yi, Y . Yang, H. Zhang et al. ChartGPT: Leveraging LLMs to Generate Charts from Abstract Natural Language. 31(3):1731–1745. doi: 10.1109/TVCG.2024.3368621 3

    work page doi:10.1109/tvcg.2024.3368621 2024

  54. [56]

    Glassman, Jeevana Priya Inala, and Chenglong Wang

    P. Vaithilingam, E. L. Glassman, J. P. Inala, and C. Wang. DynaVis: Dynamically Synthesized UI Widgets for Visualization Editing. InPro- ceedings of the CHI Conference on Human Factors in Computing Systems, pp. 1–17. ACM. doi: 10.1145/3613904.3642639 3, 5

    work page doi:10.1145/3613904.3642639

  55. [57]

    V oigt, N

    H. V oigt, N. Carvalhais, M. Meuschke, M. Reichstein, S. Zarrieß, and K. Lawonn. VIST5: An Adaptive, Retrieval-Augmented Language Model for Visualization-oriented Dialog. In Y . Feng and E. Lefever, eds.,Proceed- ings of the 2023 Conference on Empirical Methods in Natural Language Processing: System Demonstrations, pp. 70–81. Association for Computa- tion...

    work page doi:10.18653/v1/2023.emnlp-demo.5 2023

  56. [58]

    Z. Wan, Y . Song, S. Li, C. J. Zhang, and R. C.-W. Wong. DataVisT5: A Pre-trained Language Model for Jointly Understanding Text and Data Visualization. In2025 IEEE 41st International Conference on Data Engineering (ICDE), pp. 1704–1717. doi: 10.1109/ICDE65448.2025. 00131 3

    work page doi:10.1109/icde65448.2025 2025

  57. [59]

    C. Wang, B. Lee, S. M. Drucker, D. Marshall, and J. Gao. Data Formulator 2: Iterative Creation of Data Visualizations, with AI Transforming Data Along the Way. InProceedings of the 2025 CHI Conference on Human Factors in Computing Systems, pp. 1–17. ACM. doi: 10.1145/3706598. 3713296 3, 4, 7

    work page doi:10.1145/3706598 2025

  58. [60]

    C. Wang, J. Thompson, and B. Lee. Data Formulator: AI-powered Concept-driven Visualization Authoring. pp. 1–11. doi: 10.1109/TVCG. 2023.3326585 3, 5

    work page doi:10.1109/tvcg 2023

  59. [61]

    H. W. Wang, L. Birnbaum, and V . Setlur. Jupybara: Operationalizing a Design Space for Actionable Data Analysis and Storytelling with LLMs. In Proceedings of the 2025 CHI Conference on Human Factors in Computing Systems, pp. 1–24. ACM. doi: 10.1145/3706598.3713913 3, 7, 8

    work page doi:10.1145/3706598.3713913 2025

  60. [62]

    L. Wang, C. Ma, X. Feng, Z. Zhang, H. Yang, J. Zhang et al. A survey on large language model based autonomous agents.Frontiers of Computer Science, 18(6):186345, 2024. 1, 2

    2024

  61. [63]

    L. Wang, Z. Wang, S. Xiao, L. Liu, F. Tsung, and W. Zeng. Vizta: En- hancing comprehension of distributional visualization with visual-lexical fused conversational interface.Computer Graphics Forum, 44(3):e70110,

  62. [64]

    doi: 10.1111/cgf.70110 3, 4, 7, 8

    work page doi:10.1111/cgf.70110

  63. [65]

    Q. Wang, Z. Chen, Y . Wang, and H. Qu. A Survey on ML4VIS: Applying Machine Learning Advances to Data Visualization. 28(12):5134–5153. doi: 10.1109/TVCG.2021.3106142 2

    work page doi:10.1109/tvcg.2021.3106142 2021

  64. [66]

    A. Wu, Y . Wang, X. Shu, D. Moritz, W. Cui, H. Zhang et al. AI4VIS: Survey on Artificial Intelligence Approaches for Data Visualization. 28(12):5049–5070. doi: 10.1109/TVCG.2021.3099002 2

    work page doi:10.1109/tvcg.2021.3099002 2021

  65. [67]

    Y . Wu, Y . Wan, H. Zhang, Y . Sui, W. Wei, W. Zhao et al. Automated Data Visualization from Natural Language via Large Language Models: An Exploratory Study. 2(3):1–28. doi: 10.1145/3654992 3, 4, 5

    work page doi:10.1145/3654992

  66. [68]

    Z. Yang, Z. Zhou, S. Wang, X. Cong, X. Han, Y . Yan et al. MatPlotA- gent: Method and Evaluation for LLM-Based Agentic Scientific Data Visualization. In L.-W. Ku, A. Martins, and V . Srikumar, eds.,Findings of the Association for Computational Linguistics: ACL 2024, pp. 11789– 11804. Association for Computational Linguistics. doi: 10.18653/v1/2024 .findin...

    work page doi:10.18653/v1/2024 2024

  67. [69]

    S. Yao, J. Zhao, D. Yu, N. Du, I. Shafran, K. Narasimhan et al. ReAct: Synergizing Reasoning and Acting in Language Models. doi: 10.48550/ arXiv.2210.03629 1

    work page internal anchor Pith review Pith/arXiv arXiv

  68. [70]

    Zhang, X

    H. Zhang, X. Zhang, J. Zhou, K. Wu, Y . Song, and R. C.-W. Wong. Speech-to-Visualization: Toward End-to-End Speech-Driven Data Visu- alization Generation from Natural Language Questions. In R. P. Ribeiro, B. Pfahringer, N. Japkowicz, P. Larrañaga, A. M. Jorge, C. Soares et al., eds.,Machine Learning and Knowledge Discovery in Databases. Research Track, pp...

    work page doi:10.1007/978-3-032

  69. [71]

    Zhang, Y

    W. Zhang, Y . Wang, Y . Song, V . J. Wei, Y . Tian, Y . Qi et al. Natural Language Interfaces for Tabular Data Querying and Visualization: A Survey. doi: 10.48550/arXiv.2310.17894 2, 9

    work page doi:10.48550/arxiv.2310.17894

  70. [72]

    X. Zhao, X. Liu, Y . Haoyue, X. Luo, F. Zeng, J. Li et al. ChartEdit: How far are MLLMs from automating chart analysis? evaluating MLLMs’ capability via chart editing. In W. Che, J. Nabende, E. Shutova, and M. T. Pilehvar, eds.,Findings of the Association for Computational Linguistics: ACL 2025, pp. 3616–3630. Association for Computational Linguistics, Vi...

    work page doi:10.18653/v1/2025.findings-acl.185 2025

  71. [73]

    Y . Zhao, X. Shu, L. Fan, L. Gao, Y . Zhang, and S. Chen. Proactiveva: Proactive visual analytics with llm-based ui agent.IEEE Transactions on Visualization and Computer Graphics, 32(1):451–461, 2026. doi: 10. 1109/TVCG.2025.3642628 3, 4, 5, 6, 7, 8, 9

    work page arXiv 2026

  72. [74]

    Y . Zhao, J. Wang, L. Xiang, X. Zhang, Z. Guo, C. Turkay et al. LightV A: Lightweight Visual Analytics with LLM Agent-Based Task Planning and Execution. 31(9):6162–6177. doi: 10.1109/TVCG.2024.3496112 3, 4, 5, 6, 7, 8, 9

    work page doi:10.1109/tvcg.2024.3496112 2024

  73. [75]

    Y . Zhao, Y . Zhang, Y . Zhang, X. Zhao, J. Wang, Z. Shao et al. Leva: Using large language models to enhance visual analytics.IEEE Transactions on Visualization and Computer Graphics, 31(3):1830–1847, 2025. doi: 10. 1109/TVCG.2024.3368060 3, 4, 7, 8, 9

    work page arXiv 2025

  74. [76]

    Y . Zhu, L. Wang, C. Yang, X. Lin, B. Li, W. Zhou et al. A survey of data agents: Emerging paradigm or overstated hype?, 2026. 2

    2026