Research map / Vision-language and multimodal / Robot action and navigation
Vector Prism: Animating Vector Graphics by Stratifying Semantic Structure
Jooyeol Yun, Jaegul Choo
Read the paperCheck your idea against 137,217 papers
AI-extracted. The measured effects below were read by AI from the paper's full text (no summary yet). Numbers and claims can be wrong or missing context, so check the paper before you cite or build on it.
Measured effects reported in the paper
| What was changed | Where | Measured | Effect |
|---|---|---|---|
| restructure SVG into semantic groups via Dawid-Skene weighted voting before animating with an LLM | Vector Prism (GPT-5 animation) · SVG animation, 114 pairs | GPT-T2V: 40.92 (GPT-5 pipeline) -> 76.14 | Improved |
| use reliability-weighted Bayes decision rule instead of majority voting over rendering views (ablation) | Vector Prism semantic labeling · Semantic grouping, DBI | Davies-Bouldin index: 12.6 -> 0.82 | Improved |
| aggregate multi-view VLM labels rather than keep the original rendering-oriented SVG groupings (ablation) | SVG semantic grouping · Semantic grouping, DBI | Davies-Bouldin index: 33.8 -> 12.6 (majority voting) | Improved |
| generate vector CSS animations instead of raster video | Vector Prism vs Sora 2 · Encoding efficiency | file size: 54.8x compression vs Sora 2 | Improved |
| use vector-based animation instead of video generation models for instruction following | Vector Prism vs Sora 2 / Wan 2.2 · User study | preference: 63.3% vs Sora 2; 76.5% vs Wan 2.2 | Improved |
| compare video models against Vector Prism on perceptual quality | Wan 2.2 and Sora 2 · DOVER | DOVER: Wan 2.2 3.72, Sora 2 4.19 vs Ours 4.97 | Hurt |
| limit motion to raise perceptual quality | GPT-5 pipeline without semantic structure · DOVER vs GPT-T2V | DOVER / GPT-T2V: DOVER 4.92 but GPT-T2V 40.92 | Mixed |
| treat SVG primitives as atomic units without subdividing | Vector Prism · Failure case | instruction following | Hurt |
Closest papers by meaning
- Rendering-Aware Reinforcement Learning for Vector Graphics Generation (NeurIPS 2025 · similarity 0.90)
- Chat2SVG: Vector Graphics Generation with Large Language Models and Image Diffusion Models (CVPR 2025 · similarity 0.89)
- Animation2Code: Evaluating Temporal Visual Reasoning in Video-to-Code Generation (NeurIPS 2026 · similarity 0.88)
- InternSVG: Towards Unified SVG Tasks with Multimodal Large Language Models (ICLR 2026 · similarity 0.88)
- StarVector: Generating Scalable Vector Graphics Code from Images and Text (AAAI 2025 · similarity 0.88)
- Empowering LLMs to Understand and Generate Complex Vector Graphics (CVPR 2025 · similarity 0.87)
Writing a paper in this area? scime draft is a free LaTeX editor with AI included, real-time co-authors and a check of your references and venue rules before you submit. It also imports OpenAI Prism projects.
