CursorTalk: Multimodal, Spatio-Temporal Inputs for AI-assisted Web UI Editing
IUI 2027 Conditionally accepted
A co-speech and gesture-based multimodal tool for AI-assisted web UI editing.
Aditi Mishra
Senior Researcher · Fujitsu Research of America
I build HCI tools that make AI more accessible to people, with a focus on interactive systems for generative AI, video understanding, and visual analytics.
I received my PhD from Arizona State University, advised by Dr. Chris Bryan.
I'm always open to chatting about new ideas. Feel free to email me or send a DM on LinkedIn!
Publications
Map2Video: Street View Imagery Driven AI Video Generation
UIST 2026
Diffusion-based video generation driven by interactive map annotations and camera framing on street view imagery.
FuzzySeek: Multimodal Refinement of Imprecise Video Queries for Moment Retrieval
ACM IUI 2026
An interactive video moment retrieval tool for building and refining multimodal queries with AI.
WhatIF: Branched Narrative Fiction Visualization for Authoring Emergent Narratives using Large Language Models
ACM Creativity & Cognition 2025 🏅 Honorable Mention
A visual analytics interface for interactive fiction authors to explore parallel emergent stories.
VECnA: Visual Exploration, Comparison and Analysis of Reconstructed Spatiotemporal Scientific Simulation Data
EuroVis Short Papers 2025
A visualization interface for comparing reconstructions of spatiotemporal datasets.
PromptAid: Prompt Exploration, Perturbation, Testing and Iteration using Visual Analytics for Large Language Models
IEEE TVCG 2025
A visual analytics interface that helps people iterate on LLM prompts using linguistic and contextual recommendations.
Filling the Void: Deep Learning-based Reconstruction of Sampled Spatiotemporal Scientific Simulation Data
DRBSD @ SC 2024
A deep learning method for reconstructing large unstructured scientific datasets.
Characterizing Large Language Models as Rationalizers of Knowledge-intensive Tasks
ACL Findings 2024
Evaluating how well LLMs generate rationales grounded in external knowledge for knowledge-intensive tasks.
On CLIP's Ability of Analyzing Fake Images at Large Scale: Why are they fake?
GenAICHI Workshop 2024
A visual analytics system to explore, understand and summarize how vision-language models differ on real and fake images.
ConceptExplainer: Interactive Explanation for Deep Neural Networks from a Concept Perspective
IEEE TVCG 2022
Explaining deep neural networks using image-based concepts extracted with ACE and TCAV.
Why? Why not? When? Visual Explanations of Agent Behavior in Reinforcement Learning
IEEE PacificVis 2022
A visual analytics interface for questioning, and so building trust in, an autonomous agent's decisions.
News Kaleidoscope: Visual Investigation of Coverage Diversity in News Event Reporting
IEEE PacificVis 2022
A visual analytics system that helps journalism experts identify underlying polarities across news organizations.
ChartStory: Automated Partitioning, Layout, and Captioning of Charts into Comic-Style Narratives
IEEE TVCG 2021
Crafting comic-style data stories from a collection of user-created charts.
Analyzing Gaze Behavior for Text-Embellished Narrative Visualizations under Different Task Scenarios
Visual Informatics 2020
A study of perception in text-embellished narrative visualizations across observation, search and recall tasks.
TotemFinder: A Visual Analytics Approach for Image-based Key Players Identification
VAST Challenge @ IEEE VIS 2020 🏅 Honorable Mention
An interactive visualization system for analyzing the VAST 2020 Mini-Challenge 2 dataset.