2nd Workshop on
GenAI, Agents, and the Future of VIS
2nd Workshop on
IEEE VIS 2026 | November 9, 2026| Boston, MA
About the Workshop
Recent advances in agents (i.e., autonomous, goal-driven AI systems that iteratively observe, act, and learn from their environments) offer a fundamentally different approach from traditional AI models that passively respond to input. These agentic AI systems are rapidly reshaping how we approach data-intensive tasks and providing new opportunities for the VIS community. Imagine an agent autonomously generating visualizations to analyze complex data, discovering patterns collaboratively, testing hypotheses, and communicating visual insights at a speed and scale beyond human capability.
Yet, the emergence of these powerful systems raises critical questions that the VIS community must address: Could autonomous agents eventually replace human data scientists, and if not, how might they best collaborate? Are current visualization techniques and interfaces, originally designed for human analysts, suitable for agent interactions? How can VIS designers effectively integrate agents into their workflows without compromising human agency? And to what extent should agents help shape and educate the next generation of visualization researchers?
Through a mix of keynote talks, paper presentations, and an agentic VIS challenge, this workshop invites researchers and practitioners to share innovative ideas, explore these questions, and discuss strategies to transform the impact of VIS for a future where human and AI agents co-exist.
Call for Participants
We invite participation through two submission tracks: Short Paper and AgenticVIS Challenge. Both are opportunities to showcase novel ideas and engage with the growing community at the intersection of visualization, generative AI, and agentic systems.
Track A: Short Paper
Submission deadline: Aug 15, 2026
PCS submission path: TVCG → VIS 2026 → VISxGenAI
We invite short paper submissions (2-4 pages excluding references) that explore topics across theory, systems, user studies, and applications for GenAI or agentic VIS. Submissions must follow the VGTC conference two-column format, consistent with the IEEE VIS formatting guidelines. Areas of interest include, but are not limited to, the following:
- VIS for interpreting GenAI systems.
More information
Papers presenting novel visualization techniques or systems for interpretability of frontier generative models, accompanied by interactive demos that showcase practical applications for GenAI researchers. Some example papers include:
* Lee et. al. (2024). LLM Attributor: Interactive Visual Attribution for LLM Generation
* Lin (2023). Neuronpedia: Interactive Reference and Tooling for Analyzing Neural Networks.
- Human-agent collaboration in data analysis and visualization.
- Agent-augmented VIS tools.
- VIS tools for agents that agents themselves can perceive, reason over, or act upon.
- Methods and benchmarks for assessing agent performance and agent-generated visual artifacts.
- Case studies and demos of agent systems applied to real-world VIS problems.
- Position papers for VIS and AI researchers, such as agents in VIS education, VIS for embodied agents, or multi-agent coordination in visual reasoning.
More information
Brief position papers outlining research agendas, proposing benchmarks, tooling infrastructure, or suggesting collaborative projects between visualization and interpretability experts to enhance our understanding of the frontier models.
Short papers should be double-blind submissions. Initial submissions must not include author names or institutions, and authors should take care to avoid revealing identifying information in the text.
Track B: AgenticVIS Challenge
Challenge deadline: Aug 31, 2026Both the agent submission and anonymous technical report are due on this date.
PCS submission path for the technical report: TVCG → VIS 2026 → VISxGenAI Challenge
Participants will develop agents that take datasets and guiding questions, adapted from past VAST Challenges, as input and generate visual data reports as output. More details are available on the challenge website.
- Development datasets will be released before Aug 31, 2026; hidden testing datasets will be used for final evaluation.
- During development, submissions may receive automated feedback, including benchmark metrics and agent-based peer evaluation.
- By Aug 31, 2026, participants must submit their best-performing agent through vis-arena.
- Participants must also submit an anonymous technical report through PCS. The report must follow the VGTC short paper format, be no more than two pages excluding references, describe how the agent was developed, and include a link to its best generated visual report.
- Final judging will consider both the technical report and the visual report generated by the agent on the testing dataset.
See the challenge submission guide for datasets, templates, and submission instructions.
Each submission will be evaluated by at least two reviewers based on quality and topical relevance. Accepted papers will be invited for presentation as posters, demos, or lightning talks during the workshop and will be published on the workshop website. Top-rated challenge participants will receive awards and be invited to present their solutions at the workshop.
Important Dates
- Call for Participation: May 30, 2026
- Track A Short Paper Deadline: Aug 15, 2026
- Track B AgenticVIS Challenge Deadline: Aug 31, 2026
- Author Notification: Sep 10, 2026
- Camera Ready Deadline: Oct 1, 2026
- November 9, 2026: Workshop Day
Keynote Speaker
Eugene Wu
Eugene Wu is an Associate Professor of Computer Science at Columbia University and Co-Director of the Data, Agents, and Processes Lab ( DAPLab). His research explores the foundations of computing systems that enable people and AI agents to effectively interact with, analyze, and act upon data. His work spans databases, visualization, human-computer interaction, and AI systems, with a focus on building reliable and intelligent tools for human data interaction. He has developed influential systems for interactive data analysis, data cleaning, visualization, and explanation, and his recent research investigates the infrastructure needed for safe and effective agentic systems.
Schedule
November 9, 2026 · All times are local to Boston (EST, UTC−5).
| Time | Session / Talk | Presenter | Mode |
|---|---|---|---|
| 8:00–8:10 | Welcome, Opening Remarks & Challenge Introduction | ||
| 8:10–9:05 | Paper Session I: Building & Evaluating Agentic Visualization Systems | ||
| 8:10 | CHALLENGELM Agents for Visualization Generation and Vision-First Evaluation | Tahsen Islam Sajon | Remote |
| 8:15 | Toward AI VIS Co-Scientists: An Agent Harness for End-to-End Design of Scientific Visualization Applications | Haichao Miao | In person |
| 8:22 | CanvasObs: Structured Spatial Observation for Agent Editing in 2D Workspaces | Jiahao Xu | In person |
| 8:29 | AgentPulse: Visual Analytics of Communication and Activity in LLM-Based Multi-Agent Systems | Kathrin Guckes | In person |
| 8:36 | What Should the Agent Withhold? Designing Cognitive Friction Against Over-Reliance in Visualization Design Learning | Jing Lu | Remote |
| 8:43 | From Agents to Agency: Rethinking Human Agency in Agentic Visualization | Narges Mahyar | In person |
| 8:50 | Measuring the Collaboration, Not the Agent: An Evaluation Framework for Human-AI High-Dimensional Visual Analytics | Xinxiao Li | Remote |
| 8:57 | Sycamore: Grounded Synthetic Evaluation for Visualization Systems | Huyen N. Nguyen | In person |
| 9:05–9:30 | Coffee Break & Posters(25 min)View accepted posters | ||
| 9:30–10:30 | Keynote Talk(Title to be announced) | Eugene Wu | |
| 10:30–11:00 | Coffee Break & Posters(30 min)View accepted posters | ||
| 11:00–11:55 | Paper Session II: Trust, Alignment & Visual Perception of AI Models | ||
| 11:00 | CHALLENGEGenerate, Validate, Repair: Reliable Visual Data Reporting with LLM Agents | Naga Datha Saikiran Battula | In person |
| 11:07 | How Do Artificial Agents Make Decisions in Light of Visual Data? | Shayan Monadjemi | In person |
| 11:12 | Saying Uncertain, Showing Certain: Evidence–Action Mismatches in LLM Visualization Agents | Arindam Bhattacharya | Remote |
| 11:19 | Evaluating Human-LLM Alignment on Visualization Preference | Tanmay Bakshi | Remote |
| 11:26 | On Interpretability Methods for Vision Language Models in Visualization | Pranava Madhyastha | Remote |
| 11:33 | Where LLMs Fail with Visualization DSLs | Chang Han | In person |
| 11:40 | Graphical Perception Across Encoding Channels in Multimodal LLMs | Markus Wallinger | In person |
| 11:47 | A Stevens' Power Law Measurement for Multimodal Large Language Models | Kaichun Yang | In person |
| 11:55–12:00 | Closing Remarks(Organizers) | ||
Accepted Posters
Poster sessions: 9:05–9:30 and 10:30–11:00 during the coffee breaks.
| Poster | Presenter | Mode |
|---|---|---|
| Point, Revise, Review: Grounded Agentic Analysis in Reactive Notebooks with marimo-lens | Peter Gyarmati | In person |
| EduAgent: A Visual Feedback Agent for Teaching Python to Novices | Lijie Yao | Remote |
| Food Critic: An Agentic Visualization System for Auditable Restaurant-Safety Reasoning | Arleth Salinas | In person |
| From Description to Interactive VR Scene: An LLM Agent for Authoring Immersive Healthcare Experiences | Allison Andreyev | In person |
| MAADG: A Multi-Agent Framework for On-Demand Interactive Visualization Generation in Web-Based Education | Qiyuan Liu | In person |
| Verdict-Driven Human-Agent Collaboration for Autonomous-Driving Data Mining | Chen Chen | In person |
| Semantic Action Graph: A Shared Representation for Agent Grounding and Human Interpretation of Sports Highlights | Tica Lin | In person |
| TraceVis: Connecting Agent-Generated Analyses to the Queries and Evidence Behind Them | Zhaocong Yang | In person |
| Levels of AI Autonomy and Analyst Control for Bias Mitigation Interventions in Visual Analytics | Sunggyeol Oh | In person |
| Task-Aligned Layered Evaluation for Generative Visualizations | Eunhye Kim | In person |
| ChartRevive: Reconstructing Data Visualizations from Chart Images Using MLLMs | Yuki Ueno | In person |
| Multimodal LLMs and Deep Features as Perceptual Observers of Abstract Visualizations | Meng Ling | In person |
| CHALLENGECompute-Then-Narrate: A Deterministic-Analysis Agent for Dataset-Agnostic Visual Data Storytelling | Ishrat Jahan Eliza | In person |
| CHALLENGEStoryKit: A Schema-Driven Agent for Generating and Evaluating Interactive Data Stories | Aashish Panta | In person |
| CHALLENGECompiled Design, Verified Prose: An Agent That Writes Reports It Cannot See | Yuwei Xu | In person |
| CHALLENGEFrom Data Intake to Decision-Ready Evidence: An Agentic Pipeline for Auditable Visual Analytics | Kaustav Bhattacharjee | In person |
Reviewing Committee
We thank our reviewers for contributing their time and expertise to the workshop.
- Dennis BromleyTableau Research
- Furui ChengETH Zürich
- Adam CosciaStevens Institute of Technology
- Yu FuUniversity of Central Florida
- Grace GuoHarvard University
- Hyeok KimKAIST
- Fritz LekschasRidge AI
- Kylie LinGeorgia Institute of Technology
- Maxim LisnicWorcester Polytechnic Institute
- Victor MateevitsiArgonne National Laboratory
- Andrew McNuttUniversity of Utah
- Huyen N. NguyenHarvard Medical School
- Mahsan NouraniNortheastern University
- Luc RenambotUniversity of Illinois Chicago
- Carolina VeigaJames Madison University
- Yifang WangFlorida State University
- Dylan WoottonMassachusetts Institute of Technology
- Catherine YehHarvard University
- Yuheng ZhaoAalto University
Organizers
- Zhu-Tian ChenUniversity of Minnesota-Twin Cities
- Nam Wook KimBoston College
- Saeed BoorboorUniversity of Illinois Chicago
- Shivam RavalHarvard University
- Pan HaoUniversity of Minnesota-Twin Cities
- Qianwen WangUniversity of Minnesota-Twin Cities
- Vidya SetlurTableau Research
Contact
For questions, contact visxgenai@ieeevis.org.