pith. sign in

arxiv: 2604.17002 · v1 · submitted 2026-04-18 · 💻 cs.HC

Intelligent Drill-Down: Large Language Model-Driven Drill-Down Technique for Human-AI Collaborative Visual Exploration

Zhijun Zheng , Tian Qiu , Yuheng Zhao , Siming Chen This is my paper

Pith reviewed 2026-05-10 06:34 UTC · model grok-4.3

classification 💻 cs.HC
keywords visual analyticsdrill-downlarge language modelspath recommendationhuman-AI collaborationmultidimensional datainteractive explorationinsight generation
0
0 comments X

The pith

Large language models recommend drill-down paths in visual analytics by being trained to approximate a validated greedy algorithm.

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

This paper sets out to show that an LLM can assist analysts exploring large multidimensional datasets by recommending sequences of filters that lead to higher-value insights. The approach trains the model to mimic a greedy algorithm already shown to select good paths, then adds layers that read user interaction history to infer intent and manage multiple parallel exploration branches at once. A full system implements this with hierarchical navigation controls, an interactive visualization panel, and an insight panel that surfaces both findings and new path suggestions. The goal is to keep users oriented when the space of possible drill-downs grows too large for manual search. Demonstrations and a user study are offered as evidence that the method lowers cognitive load while supporting collaborative human-AI analysis.

Core claim

In the Intelligent Drill-Down Framework a large language model is used to generate visual insights, interpret user intent from interaction data, and produce suitable drill-down paths. The key technique trains the LLM to approximate a validated greedy algorithm for path recommendation. Additional components construct charts that reflect detected user intent and maintain a set of parallel exploration branches. These elements are realized in a hybrid interface that offers hierarchical navigation, a visualization panel for direct data interaction, and an insight panel that displays analytical results together with the generated recommendations. Effectiveness is illustrated through a concrete use

What carries the argument

The LLM-based drill-down path recommendation method, which is trained to approximate a validated greedy algorithm while also incorporating user-interaction data to infer intent and manage parallel branches.

Load-bearing premise

An LLM trained to approximate the greedy algorithm will reliably generate appropriate drill-down paths that match user intent and produce higher-value insights in real multidimensional data.

What would settle it

A side-by-side comparison on held-out multidimensional datasets in which the LLM-generated paths are measured against the paths the original greedy algorithm would have chosen, or a user study in which insight quality and task completion time show no reliable improvement over a non-LLM baseline interface.

Figures

Figures reproduced from arXiv: 2604.17002 by Siming Chen, Tian Qiu, Yuheng Zhao, Zhijun Zheng.

Figure 1
Figure 1. Figure 1: Mapping between drill-down patterns (interaction modes) and associated challenges. The lines identify which interaction behav￾iors lead to specific challenges (line styles alternate solely for visual clarity). 3.1 Challenges Through a literature review, we categorize common drill-down pat￾terns (i.e., existing interaction modes) and analyze the specific chal￾lenges they induce, as mapped in [PITH_FULL_IMA… view at source ↗
Figure 2
Figure 2. Figure 2: The workflow of the Intelligent Drill-Down. It mainly includes four stages: user intent understanding (a), visualization generation (b), [PITH_FULL_IMAGE:figures/full_fig_p004_2.png] view at source ↗
Figure 3
Figure 3. Figure 3: The Intelligent Drill-Down system mainly comprises 4 components. (a) Users upload datasets and instructions via (a1) to receive visualizations in (a2); (a3) illustrates the drill-down process. (b) Path navigation (breadcrumb trail) and parallel controller for branch management. (c) Configuration for model selection (c1) and interaction tracking (c2). (d) The insight panel for data-driven decision support. … view at source ↗
Figure 4
Figure 4. Figure 4: A usage scenario to demonstrate how Intelligent Drill-Down operates. [PITH_FULL_IMAGE:figures/full_fig_p007_4.png] view at source ↗
Figure 5
Figure 5. Figure 5: Subjective ratings on a 7-point Likert scale (1 = Strongly Disagree, 7 = Strongly Agree). The p value stands for significance level with *,**,*** stands for p<.1, .05 and .01, respectively. We employed the Wilcoxon signed-rank test to analyze the paired results. The significance levels were defined as follows: (*) for p < 0.1, (**) for p < 0.05, and (***) for p < 0.01. In the comparative experiment, we fou… view at source ↗
Figure 6
Figure 6. Figure 6: Results from the subjective questionnaires. The stack bars [PITH_FULL_IMAGE:figures/full_fig_p009_6.png] view at source ↗
read the original abstract

In visual analytics, applying filters to drill-down and extract higher-value insights is a common and important data analysis method. When the drill-down space becomes excessively large, analysts may lose orientation, leading to decreased efficiency in the drill-down process. To tackle these challenges, we propose the Intelligent Drill-Down Framework, in which a large language model (LLM) facilitates the generation of visual insights, leverages user interaction data to interpret user intent, and generates appropriate drill-down paths. Our method is designed to assist users in identifying valuable drill-down paths when exploring multidimensional data, thereby reducing the cognitive burden of data interpretation and facilitating the generation of insights. Specifically, we propose a drill-down path recommendation method, in which the LLM is trained to approximate a validated greedy algorithm. Secondly, we analyze the user's intent to construct a drill-down chart. Finally, we design a branch management method. Building upon this framework, we designed a system that includes a hybrid interface providing hierarchical navigation to monitor users and manage parallel branches, a visualization panel for interactive data exploration, and an insight panel to present analytical findings and generate drill-down recommendations. We evaluated the effectiveness of our method through a demonstrative use case and a user study.

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

1 major / 2 minor

Summary. The manuscript presents the Intelligent Drill-Down Framework for human-AI collaborative visual exploration of multidimensional data. It uses a large language model trained via supervised fine-tuning to approximate a validated greedy algorithm for recommending drill-down paths, while also analyzing user intent to construct charts and managing exploration branches. The system includes a hybrid interface with hierarchical navigation, a visualization panel, and an insight panel. Evaluation consists of a demonstrative use case and a within-subjects user study with 12 participants that reports reduced task completion times and higher self-reported insight quality compared to unaided exploration.

Significance. If the central claims hold, this work offers a practical method to mitigate disorientation in large drill-down spaces, enhancing efficiency and insight generation in visual analytics. The explicit training of the LLM to mimic a greedy baseline, achieving over 85% path agreement on held-out synthetic cubes, provides a solid foundation. The inclusion of a user study adds empirical support for improved human-AI collaboration. This could influence the design of future intelligent visual exploration tools.

major comments (1)
  1. [User study section] User study section: The within-subjects study with n=12 reports reduced task time and higher self-reported insight quality, but provides no statistical significance tests, effect sizes, or breakdown of quantitative metrics (e.g., exact time savings or insight counts). This makes it difficult to assess whether the results robustly support the claim that the LLM-driven paths yield higher-value insights than unaided exploration.
minor comments (2)
  1. [Abstract] The abstract mentions the user study and use case but omits key quantitative findings such as the >85% path-agreement rate; adding these would improve the summary of contributions.
  2. [Method description] The description of the supervised fine-tuning procedure (greedy paths augmented with interaction logs) would benefit from explicit details on training set size and validation splits to support reproducibility.

Simulated Author's Rebuttal

1 responses · 0 unresolved

We thank the referee for the constructive feedback and the recommendation of minor revision. We agree that the user study section would benefit from additional statistical details to strengthen the claims, and we will incorporate these in the revised manuscript.

read point-by-point responses

  1. Referee: [User study section] User study section: The within-subjects study with n=12 reports reduced task time and higher self-reported insight quality, but provides no statistical significance tests, effect sizes, or breakdown of quantitative metrics (e.g., exact time savings or insight counts). This makes it difficult to assess whether the results robustly support the claim that the LLM-driven paths yield higher-value insights than unaided exploration.

    Authors: We acknowledge this limitation in the current draft. The manuscript reports aggregate observations of reduced task completion times and higher self-reported insight quality, but does not include inferential statistics. In the revision, we will add: (1) appropriate statistical tests (e.g., paired t-tests or Wilcoxon signed-rank tests given the within-subjects design and small n), (2) effect sizes (Cohen's d or rank-biserial correlation), (3) descriptive breakdowns including means, standard deviations, exact time savings per participant or condition, and counts of insights generated. These additions will allow readers to evaluate the robustness of the results more rigorously. We note that n=12 is modest and the study is primarily demonstrative, but the requested quantitative details are feasible to include from the collected data. revision: yes

Circularity Check

0 steps flagged

No significant circularity detected

full rationale

The paper's core proposal is a system design in which an LLM is trained via supervised fine-tuning to approximate a separately validated greedy algorithm for drill-down path recommendation. This approximation is then embedded in a hybrid interface and evaluated via path-agreement metrics on held-out data plus a within-subjects user study. No equations, definitions, or self-citations reduce the claimed performance gains to a fitted parameter, a self-referential loop, or an unverified uniqueness theorem imported from the authors' prior work. The derivation chain remains self-contained against external benchmarks (the greedy baseline and user-study outcomes) and does not collapse any prediction back into its own training inputs by construction.

Axiom & Free-Parameter Ledger

0 free parameters · 2 axioms · 0 invented entities

The framework rests on the assumption that LLMs can be effectively trained to mimic a validated greedy algorithm and to infer intent from interaction logs without introducing new free parameters or invented entities beyond standard LLM usage.

axioms (2)
  • domain assumption LLMs can be trained to approximate a validated greedy algorithm for drill-down path recommendation
    Explicitly stated as the core of the path recommendation method.
  • domain assumption User interaction data can be used by the LLM to accurately interpret intent for constructing drill-down charts
    Described in the second component of the framework.

pith-pipeline@v0.9.0 · 5520 in / 1316 out tokens · 43936 ms · 2026-05-10T06:34:25.257094+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

69 extracted references · 69 canonical work pages · 1 internal anchor

  1. [1]

    Abras, D

    C. Abras, D. Maloney-Krichmar, J. Preece, et al. User-centered de- sign.Bainbridge, W. Encyclopedia of Human-Computer Interaction. Thousand Oaks: Sage Publications, 37(4):445–456, 2004

    work page 2004

  2. [2]

    GPT-4 Technical Report

    J. Achiam, S. Adler, S. Agarwal, L. Ahmad, I. Akkaya, F. L. Aleman, D. Almeida, J. Altenschmidt, S. Altman, S. Anadkat, et al. Gpt-4 technical report.arXiv preprint arXiv:2303.08774, 2023

    work page internal anchor Pith review Pith/arXiv arXiv 2023

  3. [3]

    P. A. Boncz, T. Rühl, and F. Kwakkel. The drill down benchmark. In VLDB, pp. 628–632, 1998

    work page 1998

  4. [4]

    Brehmer and T

    M. Brehmer and T. Munzner. A multi-level typology of abstract vi- sualization tasks.IEEE transactions on visualization and computer graphics, 19(12):2376–2385, 2013

    work page 2013

  5. [5]

    Ceneda, T

    D. Ceneda, T. Gschwandtner, T. May, S. Miksch, H.-J. Schulz, M. Streit, and C. Tominski. Characterizing guidance in visual analytics.IEEE transactions on visualization and computer graphics, 23(1):111–120, 2016

    work page 2016

  6. [6]

    Chaudhuri and U

    S. Chaudhuri and U. Dayal. An overview of data warehousing and olap technology.ACM Sigmod record, 26(1):65–74, 1997

    work page 1997

  7. [7]

    W. S. Cleveland and R. McGill. Graphical perception: Theory, ex- perimentation, and application to the development of graphical meth- ods.Journal of the American statistical association, 79(387):531–554, 1984

    work page 1984

  8. [8]

    Conklin, S

    N. Conklin, S. Prabhakar, and C. North. Multiple foci drill-down through tuple and attribute aggregation polyarchies in tabular data. In IEEE Symposium on Information Visualization, 2002. INFOVIS 2002., pp. 131–134. IEEE, 2002

    work page 2002

  9. [9]

    Demiralp, P

    Ç. Demiralp, P. J. Haas, S. Parthasarathy, and T. Pedapati. Foresight: Recommending visual insights.arXiv preprint arXiv:1707.03877, 2017

    work page arXiv 2017

  10. [10]

    Derthick and S

    M. Derthick and S. F. Roth. Data exploration across temporal contexts. InProceedings of the 5th international conference on Intelligent user interfaces, pp. 60–67, 2000

    work page 2000

  11. [11]

    V . Dibia. Lida: A tool for automatic generation of grammar-agnostic visualizations and infographics using large language models.arXiv preprint arXiv:2303.02927, 2023

    work page arXiv 2023

  12. [12]

    Dimara and C

    E. Dimara and C. Perin. What is interaction for data visualization? IEEE transactions on visualization and computer graphics, 26(1):119– 129, 2019

    work page 2019

  13. [13]

    Elmqvist and J.-D

    N. Elmqvist and J.-D. Fekete. Hierarchical aggregation for information visualization: Overview, techniques, and design guidelines.IEEE transactions on visualization and computer graphics, 16(3):439–454, 2009

    work page 2009

  14. [14]

    Endert, P

    A. Endert, P. Fiaux, and C. North. Semantic interaction for sensemak- ing: inferring analytical reasoning for model steering.IEEE Trans- actions on Visualization and Computer Graphics, 18(12):2879–2888, 2012

    work page 2012

  15. [15]

    Favero, L

    A. Favero, L. Zancato, M. Trager, S. Choudhary, P. Perera, A. Achille, A. Swaminathan, and S. Soatto. Multi-modal hallucination control by visual information grounding. InProceedings of the IEEE/CVF Conference on Computer Vision and Pattern Recognition, pp. 14303– 14312, 2024

    work page 2024

  16. [16]

    J. Fu, S. Huangfu, H. Yan, S.-K. Ng, and X. Qiu. Hint-before-solving prompting: Guiding llms to effectively utilize encoded knowledge. arXiv preprint arXiv:2402.14310, 2024

    work page arXiv 2024

  17. [17]

    D. Gao, H. Wang, Y . Li, X. Sun, Y . Qian, B. Ding, and J. Zhou. Text- to-sql empowered by large language models: A benchmark evaluation. arXiv preprint arXiv:2308.15363, 2023

    work page arXiv 2023

  18. [18]

    L. Gao, A. Madaan, S. Zhou, U. Alon, P. Liu, Y . Yang, J. Callan, and G. Neubig. Pal: Program-aided language models. InInternational Conference on Machine Learning, pp. 10764–10799. PMLR, 2023

    work page 2023

  19. [19]

    T. Gao, M. Dontcheva, E. Adar, Z. Liu, and K. G. Karahalios. Data- tone: Managing ambiguity in natural language interfaces for data visualization. InProceedings of the 28th annual acm symposium on user interface software & technology, pp. 489–500, 2015

    work page 2015

  20. [20]

    Gotz and M

    D. Gotz and M. X. Zhou. Characterizing users’ visual analytic activity for insight provenance.Information Visualization, 8(1):42–55, 2009

    work page 2009

  21. [21]

    J. P. Hansen, A. S. Johansen, D. W. Hansen, K. Itoh, and S. Mashino. Command without a click: Dwell time typing by mouse and gaze selections. In10th International Conference on Human-Computer Interaction, pp. 121–128. IOS Press, 2003

    work page 2003

  22. [22]

    J. Heer, J. Mackinlay, C. Stolte, and M. Agrawala. Graphical histories for visualization: Supporting analysis, communication, and evaluation. IEEE transactions on visualization and computer graphics, 14(6):1189– 1196, 2008

    work page 2008

  23. [23]

    Ikeda, J

    R. Ikeda, J. Cho, C. Fang, S. Salihoglu, S. Torikai, and J. Widom. Provenance-based debugging and drill-down in data-oriented work- flows. In2012 IEEE 28th International Conference on Data Engineer- ing, pp. 1249–1252. IEEE, 2012

    work page 2012

  24. [24]

    Joglekar, H

    M. Joglekar, H. Garcia-Molina, and A. Parameswaran. Interactive data exploration with smart drill-down.IEEE Transactions on Knowledge and Data Engineering, 31(1):46–60, 2017

    work page 2017

  25. [25]

    Kehrer, P

    J. Kehrer, P. Muigg, H. Doleisch, and H. Hauser. Interactive visual analysis of heterogeneous scientific data across an interface.IEEE Transactions on Visualization and Computer Graphics, 17(7):934–946, 2010

    work page 2010

  26. [26]

    Khosravi, S

    H. Khosravi, S. Shabaninejad, A. Bakharia, S. Sadiq, M. Indulska, and D. Gasevic. Intelligent learning analytics dashboards: Automated drill- down recommendations to support teacher data exploration.Journal of Learning Analytics, 8(3):133–154, 2021

    work page 2021

  27. [27]

    H. Lam. A framework of interaction costs in information visualization. IEEE transactions on visualization and computer graphics, 14(6):1149– 1156, 2008

    work page 2008

  28. [28]

    A. Lee, E. Che, and T. Peng. How well do llms compress their own chain-of-thought? a token complexity approach.arXiv preprint arXiv:2503.01141, 2025

    work page arXiv 2025

  29. [29]

    B. Lee, P. Isenberg, N. H. Riche, and S. Carpendale. Beyond mouse and keyboard: Expanding design considerations for information visual- ization interactions.IEEE Transactions on Visualization and Computer Graphics, 18(12):2689–2698, 2012

    work page 2012

  30. [30]

    D. J.-L. Lee, H. Dev, H. Hu, H. Elmeleegy, and A. Parameswaran. Avoiding drill-down fallacies with vispilot: Assisted exploration of data subsets. InProceedings of the 24th International Conference on Intelligent User Interfaces, pp. 186–196, 2019

    work page 2019

  31. [31]

    G. Li, X. Wang, G. Aodeng, S. Zheng, Y . Zhang, C. Ou, S. Wang, and C. H. Liu. Visualization generation with large language models: An evaluation.arXiv preprint arXiv:2401.11255, 2024

    work page arXiv 2024

  32. [32]

    H. Li, G. Appleby, K. Alperin, S. R. Gomez, and A. Suh. The role of visualization in llm-assisted knowledge graph systems: Effects on user trust, exploration, and workflows.arXiv preprint arXiv:2505.21512, 2025

    work page arXiv 2025

  33. [33]

    Y . Li, Y . Qi, Y . Shi, Q. Chen, N. Cao, and S. Chen. Diverse interaction recommendation for public users exploring multi-view visualization using deep learning.IEEE transactions on visualization and computer graphics, 29(1):95–105, 2022

    work page 2022

  34. [34]

    L. Lins, J. T. Klosowski, and C. Scheidegger. Nanocubes for real- time exploration of spatiotemporal datasets.IEEE Transactions on Visualization and Computer Graphics, 19(12):2456–2465, 2013

    work page 2013

  35. [35]

    C. Liu, L. Xie, Y . Han, D. Wei, and X. Yuan. Autocaption: An approach to generate natural language description from visualization automati- cally. In2020 IEEE Pacific visualization symposium (PacificVis), pp. 191–195. IEEE, 2020

    work page 2020

  36. [36]

    R. Luo, L. Sun, Y . Xia, T. Qin, S. Zhang, H. Poon, and T.-Y . Liu. Biogpt: generative pre-trained transformer for biomedical text generation and mining.Briefings in bioinformatics, 23(6):bbac409, 2022

    work page 2022

  37. [37]

    Mackinlay

    J. Mackinlay. Automating the design of graphical presentations of relational information.Acm Transactions On Graphics (Tog), 5(2):110– 141, 1986

    work page 1986

  38. [38]

    Maddigan and T

    P. Maddigan and T. Susnjak. Chat2vis: Generating data visualizations via natural language using chatgpt, codex and gpt-3 large language models.Ieee Access, 11:45181–45193, 2023

    work page 2023

  39. [39]

    Mafrur, M

    R. Mafrur, M. A. Sharaf, and H. A. Khan. Dive: Diversifying view recommendation for visual data exploration. InProceedings of the 27th ACM International Conference on Information and Knowledge Management, pp. 1123–1132, 2018

    work page 2018

  40. [40]

    Nemeth, D

    M. Nemeth, D. Borkin, A. Nemethova, and G. Michalconok. Deep drill-down analysis for failures detection in the production line. In2021 23rd International Conference on Process Control (PC), pp. 325–330. IEEE, 2021

    work page 2021

  41. [41]

    Y . Qin, K. Song, Y . Hu, W. Yao, S. Cho, X. Wang, X. Wu, F. Liu, P. Liu, and D. Yu. Infobench: Evaluating instruction following ability in large language models.arXiv preprint arXiv:2401.03601, 2024

    work page arXiv 2024

  42. [42]

    T. Qiu, F. Wang, S. Huang, M. Guo, Y . Zhao, J. Li, and S. Chen. Smartmlvs: Llm-enabled multiple linked views generation for interac- tive visualization. In2025 IEEE 18th Pacific Visualization Conference (PacificVis), pp. 58–68. IEEE, 2025

    work page 2025

  43. [43]

    E. D. Ragan, A. Endert, J. Sanyal, and J. Chen. Characterizing prove- nance in visualization and data analysis: an organizational framework of provenance types and purposes.IEEE transactions on visualization and computer graphics, 22(1):31–40, 2015

    work page 2015

  44. [44]

    Sarawagi

    S. Sarawagi. User-adaptive exploration of multidimensional data. In VLDB, pp. 307–316. ResearchGate GmbH, 2000

    work page 2000

  45. [45]

    Sarawagi, R

    S. Sarawagi, R. Agrawal, and N. Megiddo. Discovery-driven explo- ration of olap data cubes. InInternational Conference on Extending Database Technology, pp. 168–182. Springer, 1998

    work page 1998

  46. [46]

    Satyanarayan, D

    A. Satyanarayan, D. Moritz, K. Wongsuphasawat, and J. Heer. Vega- lite: A grammar of interactive graphics.IEEE transactions on visual- ization and computer graphics, 23(1):341–350, 2016

    work page 2016

  47. [47]

    Savva, N

    M. Savva, N. Kong, A. Chhajta, L. Fei-Fei, M. Agrawala, and J. Heer. Revision: Automated classification, analysis and redesign of chart images. InProceedings of the 24th annual ACM symposium on User interface software and technology, pp. 393–402, 2011

    work page 2011

  48. [48]

    Shanmugasundaram, U

    J. Shanmugasundaram, U. Fayyad, and P. S. Bradley. Compressed data cubes for olap aggregate query approximation on continuous dimensions. InProceedings of the fifth ACM SIGKDD international conference on Knowledge discovery and data mining, pp. 223–232, 1999

    work page 1999

  49. [49]

    Shneiderman

    B. Shneiderman. Human-centered artificial intelligence: Reliable, safe & trustworthy.International Journal of Human–Computer Interaction, 36(6):495–504, 2020

    work page 2020

  50. [50]

    Siddiqui, A

    T. Siddiqui, A. Kim, J. Lee, K. Karahalios, and A. Parameswaran. Ef- fortless data exploration with zenvisage: an expressive and interactive visual analytics system.arXiv preprint arXiv:1604.03583, 2016

    work page arXiv 2016

  51. [51]

    Singhal, T

    K. Singhal, T. Tu, J. Gottweis, R. Sayres, E. Wulczyn, M. Amin, L. Hou, K. Clark, S. R. Pfohl, H. Cole-Lewis, et al. Toward expert- level medical question answering with large language models.Nature Medicine, 31(3):943–950, 2025

    work page 2025

  52. [52]

    Stolte, D

    C. Stolte, D. Tang, and P. Hanrahan. Polaris: A system for query, analysis, and visualization of multidimensional relational databases. IEEE Transactions on visualization and computer graphics, 8(1):52–65, 2002

    work page 2002

  53. [53]

    F. Tan, P. Cascante-Bonilla, X. Guo, H. Wu, S. Feng, and V . Ordonez. Drill-down: Interactive retrieval of complex scenes using natural lan- guage queries.Advances in neural information processing systems, 32, 2019

    work page 2019

  54. [54]

    Terry and E

    M. Terry and E. D. Mynatt. Side views: persistent, on-demand pre- views for open-ended tasks. InProceedings of the 15th annual ACM symposium on User interface software and technology, pp. 71–80, 2002

    work page 2002

  55. [55]

    Y . Tian, W. Cui, D. Deng, X. Yi, Y . Yang, H. Zhang, and Y . Wu. Chartgpt: Leveraging llms to generate charts from abstract natural language.IEEE Transactions on Visualization and Computer Graphics, 31(3):1731–1745, 2024

    work page 2024

  56. [56]

    Vartak, S

    M. Vartak, S. Madden, A. Parameswaran, and N. Polyzotis. Seedb: automatically generating query visualizations. 2014

    work page 2014

  57. [57]

    Vartak, S

    M. Vartak, S. Rahman, S. Madden, A. Parameswaran, and N. Polyzotis. Seedb: Efficient data-driven visualization recommendations to support visual analytics. InProceedings of the VLDB Endowment International Conference on Very Large Data Bases, vol. 8, p. 2182, 2015

    work page 2015

  58. [58]

    F. Wang, B. Wang, X. Shu, Z. Liu, Z. Shao, C. Liu, and S. Chen. Chartinsighter: An approach for mitigating hallucination in time-series chart summary generation with a benchmark dataset.IEEE Transac- tions on Visualization and Computer Graphics, 2025

    work page 2025

  59. [59]

    S. Won, H. Kwak, J. Shin, J. Han, and K. Jung. Break: Breaking the dialogue state tracking barrier with beam search and re-ranking. InProceedings of the 61st Annual Meeting of the Association for Computational Linguistics (Volume 1: Long Papers), pp. 2832–2846, 2023

    work page 2023

  60. [60]

    Wongsuphasawat, D

    K. Wongsuphasawat, D. Moritz, A. Anand, J. Mackinlay, B. Howe, and J. Heer. V oyager: Exploratory analysis via faceted browsing of visualization recommendations.IEEE transactions on visualization and computer graphics, 22(1):649–658, 2015

    work page 2015

  61. [61]

    Wongsuphasawat, D

    K. Wongsuphasawat, D. Moritz, A. Anand, J. Mackinlay, B. Howe, and J. Heer. Towards a general-purpose query language for visualization recommendation. InProceedings of the workshop on human-in-the- loop data analytics, pp. 1–6, 2016

    work page 2016

  62. [62]

    Wongsuphasawat, Z

    K. Wongsuphasawat, Z. Qu, D. Moritz, R. Chang, F. Ouk, A. Anand, J. Mackinlay, B. Howe, and J. Heer. V oyager 2: Augmenting visual analysis with partial view specifications. InProceedings of the 2017 chi conference on human factors in computing systems, pp. 2648–2659, 2017

    work page 2017

  63. [63]

    Y . Wu, Y . Wan, H. Zhang, Y . Sui, W. Wei, W. Zhao, G. Xu, and H. Jin. Automated data visualization from natural language via large language models: An exploratory study.Proceedings of the ACM on Management of Data, 2(3):1–28, 2024

    work page 2024

  64. [64]

    J. Xu, Z. Li, W. Chen, Q. Wang, X. Gao, Q. Cai, and Z. Ling. On- device language models: A comprehensive review.arXiv preprint arXiv:2409.00088, 2024

    work page arXiv 2024

  65. [65]

    S. Yao, J. Zhao, D. Yu, N. Du, I. Shafran, K. R. Narasimhan, and Y . Cao. React: Synergizing reasoning and acting in language models. InThe eleventh international conference on learning representations, 2022

    work page 2022

  66. [66]

    Zhang, S

    D. Zhang, S. Tang, D. Yang, and L. Jiang. An effective drill-down paths pruning method in olap. InFourth International Conference on Fuzzy Systems and Knowledge Discovery (FSKD 2007), vol. 4, pp. 649–653. IEEE, 2007

    work page 2007

  67. [67]

    Y . Zhao, X. Shu, L. Fan, L. Gao, Y . Zhang, and S. Chen. Proactiveva: Proactive visual analytics with llm-based ui agent.arXiv preprint arXiv:2507.18165, 2025

    work page arXiv 2025

  68. [68]

    Y . Zhao, J. Wang, L. Xiang, X. Zhang, Z. Guo, C. Turkay, Y . Zhang, and S. Chen. Lightva: Lightweight visual analytics with llm agent- based task planning and execution.IEEE Transactions on Visualization and Computer Graphics, 2024

    work page 2024

  69. [69]

    Y . Zhao, Y . Zhang, Y . Zhang, X. Zhao, J. Wang, Z. Shao, C. Turkay, and S. Chen. Leva: Using large language models to enhance visual analytics.IEEE transactions on visualization and computer graphics, 31(3):1830–1847, 2024

    work page 2024