Benedikt's Notes

IEEEVIS 2025

13 min read

Workshshops

alt.VIS (Sunday Morning)

https://altvis.github.io/

Data Melodification FM

Visualization Was Here

The Fuzzy Front Ends

Very interesting talks, Melodification could be part of survey for HCDA. TLX-on-TLX by Daniel was also a good throwback to the NASA-TLX, complete with a LaTeXtemplate to emulate the old typewriter-paper feel. …

Bio+MedVIS (Sunday Morning): Redesign Challenge Introduction

  • Spectral analysis of 31P-MRS output of 9 individuals
  • Dashboards to show spectra/peaks with rations of metabolic ranges
  • Difference between subjects focus of many works, ratios/difference

SciVis (Sunday Afternoon)

  • More focus on specific solutions
  • Inverse problem analysis could be interesting?
  • A lot of dimensionality reduction, used PCA which could be flawed in the current setting
  • Filtering on parallel coordinates/coordinates was very popular too, could be interesting to replace by target function and weight to allow fuzzy optimisation and apply Pareto front for our approach.

LDAV (Monday Morning)

Opening & Keynote

  • Keynote: Kwan-Liu Ma
  • Extract data, integrate viz in flows, integration AI/ML
  • INR: implicit neural representations, for data reduction, i.e. encode dataset in model (related to PINNs, kind of as data/observations encoded in NN), new method: Gaussian Splatting, kind of related as NERF supplanted by Splatting
  • Glyph-based viz, after spatial aggregation (using Voronoi patterns) 1
  • Uncertainty: 2D uncertainty: use small multiples, limitations of direct interpretation (requires understanding) 2, ClimateSOM to encode distribution of uncertainty (instead of summary)

Extracting Complex Topology from Multivariate Functional Approximation: Contours, Jacobi Sets, and Ridge-Valley Graphs

Guanqun Ma, David Lenz, Hanqi Guo, Tom Peterka, Bei Wang

  • Extract topology without having to sample from model?
  • Reframe Ridge-Valley graph (complex topology) from contour graph
  • Core approach: use gradient of INF to trace paths, then use these as splines, extract gradients from these, then get the Ridge-Valley graphs/Jacobi Graphs
  • Very smooth lines (B-Splines…)
  • Possible weakness: NN gradients non-smooth/well learned… (especially for higher dimensions!)

Extremely Scalable Distributed Computation of Contour Trees via Pre-Simplification

Mingzhe Li, Hamish Carr, Oliver Rübel, Bei Wang, Gunther H. Weber

ChatVis

  • Use RAG to automate viz based on prompts, evaluated on 20 examples + agentic correction flow
  • RAG is quite good…

From Soup to Bricks: Fast Clustering of Fine-Grained AMR Hierarchies for Rendering on GPUs

Stefan Zellmann, Ingo Wald

  • Optimize Kd-based rendering of flow/volumetric data.

Lossy Parallel Visualization of Large-Scale Volume Data with Error-Bounded Image Compositing

Yongfeng Qiu, Yuxiao Li, Xin Liang, Yafan Huang, Guanpeng Li, Sheng Di, Franck Cappello, Hanqi Guo

  • Compression for image exchange between nodes
  • Use cross-combinations by using cross-exchange using
  • Calculate error based on alpha-blending/compression overlaying
  • Can calculate upper bound for error.

Managing Data for Scalable and Interactive Event Sequence Visualization

Sayef Azad Sakin, Katherine E. Isaacs

  • use Kd-tree across event paths, would it be better to use binary trees to not mix non-spatially related tracks.
  • SSIM as metric for visual similarity for viz

VAST Challenge 2025

Welcome & Introduction

  • Hidden story extraction from dataset (fictional island of Oceanis)
  • MC1:
  • MC2: Bias detecion in a KG, conflict of Tourism/Fishing
  • MC3: Last year illegal fishing, this year: secret tourism patter/atypical activities detection including deceptive/contradicting information
  • DC: Design challenge, propose designs for diverse teams

BAIT Dashboard (MC2)

Interactive Platform 4 VA of Suspicious patterns (MC3)

  • Build KG from radio messages
  • Discovering patterns/groups within networks, using topic modeling over messages
  • and using similarity measures of content

Intuitive Support for Query Construction (DC)

  • Related works: State of the Art in Multivariate Network Visualization
  • TreeMap-based viz of node types/edges & hierarchies

Conference

Opening & Keynote: Visualization as a Science/The path of Viz to Science

  • Early Viz: ASCII plot to a row-based line printer, very simple yet effective, next line printer (vector graphics) of MC simulations of ray scattering in clouds
  • Early start in Supercomputers, “Mental Images” on the first GPUs (create every imaginable image on the computer)
  • Medical planning: volume data to finite element representation using segmentation and organ rendering (surgery planning)
  • Feature extraction from flow modeling
  • da Vinci Studio (before Cave Viz)
  • His points: serious application, entire problem, identify core functionality, cross disciplinary;
  • The Science of Visual Insight
    • One dataset can tell different stories (e.g. log vs linear)
    • Visualizations to make sense of data!
    • The mental representation/model/concepts/semantic networks/language/knowledge gets updated (Tobias’s “aha!”, for language - viz metaphors)
    • “modern” worldview: smybolic - nature - social orders, given truths - viz is according to him in the middle, cannot capture all truths… (this seems like the Venn diagram meme), map to different “ontologies” (not in KG sense but philosophical sense…)
    • Floridi’s infosphere / related viz to various philosophical arguments about truth and understandings, Foccaults power structures
    • Critical Visualization 3
    • What makes a science a science?
      1. Epistemic clarity: measure/truths can be validated
      2. Theoretical articulation: ontology of visual objects
      3. Normative guardrails: ethics of epistemic-ness, power, fairness?
    • Visual Knowledge - definition? - transparent in provenance, reusable, uncertainty
    • Visualizing Arguments - every viz serves a claim! Make structure/provenance clear
    • Predictive Understanding - toolset to verifiably/predict Understanding
    • Power and Responsibility of Viz, Foucault’s representations coupled with power.
    • two major problems: how to gain and communicate true knowledge

Awards

  • VGTC Lifetime award: Meister
  • Best short paper: Toward a Logic of Generalization about Visualization as a Decision Aid
    • How can viz be useful for decision making / epistemic framework: atomic viz parts vs. contextualized interpretation
    • Formalize decision theory in viz (state - signal - interpretation - action)

Best Paper Awards

”They Aren’t Built For Me”: An Exploratory Study of Strategies for Measurement of Graphical Primitives in Tactile Graphics

Areen Khalaila, Lane Harrison, Nam Wook Kim, Dylan Cashman

  • Accessibility for Data using Physicalisation
  • Tasks for BLV are exponentially harder than for sighted people
  • Evaluate errors of blind people on chart reading - some measures are similar to sighted people, some are very different (i.e. area measurements)
  • Expected outcomes: inclusive design, remove tactile distraction

ReVISit 2: A Full Experiment Life Cycle User Study Framework

Zach Cutler, Jack Wilburn, Hilson Shrestha, Yiren Ding, Brian Bollen, Khandaker Abrar Nadib, Tingying He, Andrew McNutt, Lane Harrison, Alexander Lex

  • Really great tool!

Beyond Problem Solving: Framing and Problem–Solution Co-Evolution in Data Visualization Design

Parsons, Prakash Chandra Shukla

Causality-based Visual Analytics of Sentiment Contagion in Social Media Topics

  • Interesting causality maps (by assuming time progression is not just correlation…)

Test-of-Time award

Vector field visualization

  • Use templates for critical points (fingerprints?)

Uncertainty Viz

  • Use bands to visualize uncertainties (see master thesis?)

Compression algorithm for 4D vector compression

  • Compression on block-level, with simple Mutiply-Add
  • Widely used, Open-Source (funding for porting!)

Graph-Theoretic Scagnostics

Voyager

Well-known viz browsers

Reducing Snapshots to Points

  • Analyze dynamic network by flattening them and performing dimensionality reduction and drawing them as time curves.
  • Kind of similar to GNN (according to author)

Maps & Spatial Vis

”Mapping What I Feel”: Understanding Affective Geovisualization Design Through the Lens of People-Place Relationships

  • Emotional Mapping/Geography
  • Identifying Emotional Visualization through survey (remove image/historical viz) - Affective Visualization
  • Situated and personalized compared to other emotional viz.
  • For Jesú : social/humanistic viz.

Unveiling the Visual Rhetoric of Persuasive Cartography: A Case Study of the Design of Octopus Maps

  • Octopus: literally?, historically famous (they creep into the data, rhetorically speaking)
  • Survey of maps containing octopus with deep coding of maps, propaganda/misinformation maps deeply connected! Octupos

How do Data Journalists Design Maps to Tell Stories?

  • Explore design space of these specifc maps
  • Article sourrounding maps, accompying viz/supporting media
  • Interactivity/Overlays…
  • Interactive more challenging, less common (expensive!)
  • Tight deadlines force less map inclusion, AI enables more inclusion (but at what cost)?
  • A lot of users require locator maps, data literacy often a problem/readers often lost, require well-versed audience
  • Limitations: biased towards Brazilian journalists, no mobile maps

Octupos

Algorithmically-Assisted Schematic Transit Map Design: A System and Algorithmic Core for Fast Layout Iteration

  • Their major contribution: interactive design, compared to Hannah Bast’s greedy construction!
  • Fast/Global algorithms survey
  • Kind of limited user study (but only a few people)

Algorithms for Consistent Dynamic Labeling of Maps With a Time-Slider Interface

  • Time-Based labeling of images with time/geospatial info
  • Use maximal information metric for labeling
  • Query by adjusted slider
  • Optimal + Valid Activity diagram - define constriants to display optimal viz! Using ILP for optimal labeling
  • @Julian: could be interesting to display your optimal display of progressive viz on time series

Volume-Based Space-Time Cube for Large-Scale Continuous Spatial Time Series

  • Space-Time Cube/linked views/overlay for displaying time-based data on maps
  • Does not work on large-scale data!
  • Slicing of volume (effectively volume atop map)

Vis 4 Science

Paraview-MCP: An Autonomous Visualization Agent with Direct Tool Use

  • AI agents for Para View
  • Reduce Barrier of Entry
  • No user study, but present strong use case with Blender MCP

Uncertain Mode Surfaces in 3D Symmetric Second-Order Tensor Field Ensembles

  • Prior work: Topological Tensor field
  • Tensor Mode can be extracted to get contours/limits/extremes
  • Uncertainty in Tensor Ensemble: Calculate Mean - then Mode (removes information), therefore - mode of each field, then calculate variance/…over these, viz these
  • Current limitations: only for 1D-lines, current work: add more extension

Virtual Ray Sampling for Direct Volume Rendering using Hermite Interpolation

SEG-RobustEye: Understanding medical image segmentation models

Scope Meets Screen: Lessons Learned in Designing Composite Visualizations for Marksmanship Training Across Skill Levels

Interactive Visual Analytics of Carbon Cycle Science

  • Usually: box-flow diagrams (boxes represent storage, arrows movement)

Analyzing Time-Varying Scalar Fields using Piecewise-Linear Morse-Cerf Theory

Abstract Physical Spaces

Don’t Stop Me Now: Visualizing Disruptions in Railroad Networks

  • Goal: exploring train disruption dataset
  • Disrupted segments/temporal heatmap
  • Use Case: Spreading of weather related delays and repair “cooldown”
  • https://trainviz.github.io/

Posterity: Balancing historical context and visual dynamism while visualizing a collection of American labor posters

  • Nice demo with dimensionality reduction/embedding similarity to custom posters/clustering

Chronotome: Real-Time Topic Modeling for Streaming Embedding Spaces

  • Progressive Viz of Clusters
  • Could be interesting for the RIS time viz of topics in laws?
  • Cluster evolution over time in 3D!, time constant cluster by doing 2D clustering first then expanding to 3D
  • Applied to 3 different models, and has the progressive aspects which sets it apart…

Data-Driven Compute Overlays for Interactive Geographic Simulation and Visualization

Embedding Atlas: Low-Friction, Interactive Embedding Visualization

  • Main selling point: easy setup/sane defaults/fast
  • Uses webGPU for fast rendering
  • Easy viz config, reducing friction

CFTree: Exploring Paths Through Counterfactuals

  • Uses DECE for Counterfactuals
  • Parallel coordinates over nodes

Visualizing Climate Model Outputs with CliMAScope

  • Full-On Design study with experts

MC-INR: Efficient Encoding of Multivariate Scientific Simulation Data using Meta-Learning and Clustered Implicit Neural Representations

  • Usually/Related: INR for compression
  • Their approach: not only use a single INR, globally, but use multiple INRs clustered by k-Means

Graphs and Networks

Envisage: Towards Expressive Visual Graph Querying

Dimensionality Reduction and Parameter Space Analysis

ClimateSOM: A Visual Analysis Workflow for Climate Ensemble Datasets

  • Partition SOM into clusters
  • Study with experts (this seems to be the deciding factor for many accepted papers!)

SEAL: Spatially-resolved Embedding Analysis with Linked Imaging Data

  • Gigapixel size with 100+ channels, millions of cells!
  • Bridge spatial image / embedding
  • Cells extracted and rearranged in embedding view - hybrid embedding view, render cells progressively
  • Linked view dashboard, with lasso tools to select embedding view and show in overview.
  • Calculate SHAP features based on model learned on dimensionality reduction (similar to our SciVis submission - we just use SG and use it as guidance)
  • Eval: Use Case Study with three experts in three domain
  • http://sealvis.org/ using anywidet

A Critical Analysis of the Usage of Dimensionality Reduction in Four Domains

  • Survey of DR usage in domains

Interactive Visual Analysis of Spatial Sensitivities

RSVP for VPSA : A Meta Design Study on Rapid Suggestive Visualization Prototyping for Visual Parameter Space Analysis

IEEE VIS Reviewing — On a Path to Self-Destruction?

Petra Isenberg, Gunther Weber, Narge Mahyar, Niklas Elmqvist, Han-Wei Shen, Michael Sedlmair, Melanie Tory, Helwig Hauser, Bei Wang, Tamara Munzner

  • Reviews are on a bumpy path - LMs, virtual conference, good (?): student reviews
  • What we value in reviews (related to psychology panel)
  • Survey results
    • Gatekeeping in Viz – less novelty, subjective quality, career systems; efficiency is going down (hundreds of review cycles) – beer garden theory (spend your time effectively)
    • Intransparent discussion – reviewers change score, but not underlying review; one person (critical one) leads discussion,
    • Perceived randomness – because it is a noisy sampling process; be in a good mood for reviewing ;), one (very convinced) person could lead discussion; recommendation: noise - a flaw in the human judgement
    • Unqualified reviewers – student reviewers (no weigh in review scores - that makes it feel valued for me…), unqualified PCS reviewers (how did that happen?), do not just average; also look for the good in the work, do not expect everything
    • Unnecessarily negative reviews – review fatigue, harsh gatekeeping, self-check expertise - positive framing (e.g. lacks justification instead of missing parameters)
    • Low range – people do not use the full range, update review after being convinced!
  • Discussion

From data to meaning

Stitching Meaning: Practices of Data Textile Creators

  • Survey of Creators
  • Looking at (meta-)structure, meaning, data, color, motivation

Story Ribbons: Reimagining Storyline Visualizations with Large Language Models

  • Structural timelines from text(-stories)
  • Character arcs …
  • Using LMs and correction loops, interactive explanations and provenance to the texts
  • Added a lot more fine-grained characters and themes compared to other methods

Stories: A Modular Framework for Multi-Objective Optimization of Storylines with a Focus on Fairness

  • Fairness in Viz?
  • Tradeoffs between global and local fairness (minorities)
  • Networks: MILP used instead of global layouting
  • Optimize crossings in story lines

Transportation, Buildings, and Urban Vis

StressDiffVis: Visual Analytics for Multi-Model Stress Comparison

  • Alternative Visualization for stress visualization for construction views

StreetWeave: A Declarative Grammar for Street-Overlaid Visualization of Multivariate Data

  • Maps, Urban analysis - Walkability score
  • Design space survey of walkable/accesible neighbourhood street analysis.

TraSculptor: Visual Analytics for Enhanced Decision-Making in Road Traffic Planning

  • Model traffic demand on graph, explore options on graph
  • Editor for the graph (including street width …)
  • Interesting related work for Jesús

Interaction & Provenance

DashSpace: A Live Collaborative Platform for Immersive and Ubiquitous Analytics

  • Interactive, collabortive environment that can be shared online and accross many devices (iPhone to Occolus)

Stories and the Journalists Who Tell Them

SceneLoom: Communicating Data with Scene Context

  • Dat with real-world imagery connection
  • Automated visualization creation using semantic possibilities for in-image viz

Data Speaks, But Who Gives It a Voice? Understanding Persuasive Strategies in Data-Driven News Articles

  • Emotional, Persuasive strategies
  • Implement MLLM to predict these strategies, estimate effects of strategies

Deconstructing Implicit Beliefs in Visual Data Journalism: Unstable Meanings Behind Data as Truth & Design for Insight

  • Objective/subjective truth vs. humanistic/mechanistic insights at odds
  • Instead of data which beliefs of data/design influences their decisions to truth seeking
  • Interviews revealed tension/implicit ideas in participants
  • Link to historical events underlying data insights/constructs
  • Objective/mechanism are strategic rituals; skepticism to subjective framing

Immersive Data-Driven Storytelling: Scoping an Emerging Field Through the Lenses of Research, Journalism, and Games

  • Survey over immersive data stories
  • Hard to find/a lot of methods - looked at everything from articles to Steam games
  • Nice illustrations of the stories to give a consistent view

Narrative Player: Reviving Data Narratives with Visuals

  • Text-driven storytelling with visuals from raw data (sometimes just data+text)
  • Generative visualization
  • Narrative analysis, fact analysis (extraction)
  • Automatic Viz generation
  • Q: how truthful/how to guard against hallucinations?
  • 12 students for user study: dataset/ablation studies

Best Paper Awards according to me

  • Smoothest Maps: Algorithmically-Assisted Schematic Transit Map Design: A System and Algorithmic Core for Fast Layout Iteration
  • Longest Tentacles: Unveiling the Visual Rhetoric of Persuasive Cartography: A Case Study of the Design of Octopus Maps
  • Fastest Avalanche: Data-Driven Compute Overlays for Interactive Geographic Simulation and Visualization

Footnotes

  1. https://arxiv.org/pdf/2506.23092v1 ↩

  2. https://arxiv.org/pdf/2012.11109 ↩

  3. https://academic.oup.com/cybersecurity/article/1/1/93/2366512?login=false ↩

Graph View

Backlinks