Presentation

<
1. AI Score Hallucination in Vision-Language Models: Measuring Self-Assessment Calibration for Iterative 3D Scene Positioning
SessionPosters: 3D & Geometry
DescriptionStoryboardTo3D uses vision-language models to iteratively position 3D objects from hand-drawn storyboards in Unreal Engine 5.6. Across three VLMs and 36 scenarios, I characterized AI Score Hallucination: a 37.2 percentage point gap between models' self-reported scores and expert-validated success, with implications for VLM-as-judge stopping criteria in creative pipelines.
Contributor
Tyler Varacchi
Drexel University
Event Type
Poster
TimeSunday, 19 July 20269:00am - 5:30pm PDT
LocationWest Hall Lobby
Session TimeSunday, 19 July 20269:00am - 5:30pm PDTMonday, 20 July 20269:00am - 5:30pm PDTTuesday, 21 July 20269:00am - 5:30pm PDTWednesday, 22 July 20269:00am - 5:30pm PDTThursday, 23 July 20269:00am - 5:30pm PDT
LocationWest Hall Lobby
Registration Categories
Full Conference Supporter
Full Conference
Experience
Next PresentationNext Presentation2. Aurora: Real-Time Co-Sculpting of Mesh-Based Multi-Agent Geometry for Fabrication-Aware Design
Similar Presentations
Can AI Judge Its Own Work? Self‑Assessment and Trust in Creative AI Pipelines
From Human Vision to Machine Vision
19. Deploying Speech-Driven 3D Facial Animation in Unreal Engine for Production-Ready Digital Humans