Research map / Vision-language and multimodal / Multimodal LLMs
SVGEval: A Vision-Grounded Framework for Perceptual-Quality Benchmarking and Evaluation in Text-to-SVG Generation
Yiming Wang, Ye Chen, Hanqi Chen, Bingbing Ni
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 |
|---|---|---|---|
| add rationale (reasoning) supervision (ablation) | SVGEval-Scorer · SVGEval Part II | MAE: 0.8330 -> 0.5772 | Improved |
| add rationale (reasoning) supervision (ablation) | SVGEval-Scorer · SVGEval Part II | adjacent accuracy: 60.30 -> 71.95 | Improved |
| add SVG code input alongside image (ablation) | SVGEval-Scorer · SVGEval Part II | MAE: 0.6220 -> 0.5772 | Improved |
| add SVG code input alongside image (ablation) | SVGEval-Scorer · SVGEval Part II Spatial | MAE: 0.6400 -> 0.5167 | Improved |
| add SVG code input alongside image (ablation) | SVGEval-Scorer · SVGEval Part II | adjacent accuracy: 70.40 -> 71.95 | Improved |
| add rendered image input alongside SVG code (ablation) | SVGEval-Scorer · SVGEval Part II | MAE: 0.9557 -> 0.5772 | Improved |
| add rendered image input alongside SVG code (ablation) | SVGEval-Scorer · SVGEval Part II | adjacent accuracy: 57.75 -> 71.95 | Improved |
| fine-tune Qwen3-VL 30B A3B scorer vs best general MLLM (Gemini-3 Pro) | SVGEval-Scorer · SVGEval Part II | MAE: 0.7068 -> 0.5772 | Improved |
| fine-tune scorer vs best general MLLM on Spatial dimension | SVGEval-Scorer · SVGEval Part II Spatial | MAE: 0.8400 -> 0.5167 | Improved |
| fine-tune scorer vs Gemini-3 Pro on Semantic dimension | SVGEval-Scorer · SVGEval Part II Semantic | MAE: 0.6160 -> 0.6129 | No clear effect |
| compare models on Spatial versus Semantic dimension | GPT-5 · SVGEval Part II | MAE: 0.7840 -> 1.2258 | Hurt |
Closest papers by meaning
- VFIG: Vectorizing Complex Figures in SVG with Vision-Language Models (NeurIPS 2026 · similarity 0.90)
- InternSVG: Towards Unified SVG Tasks with Multimodal Large Language Models (ICLR 2026 · similarity 0.90)
- OmniSVG: A Unified Scalable Vector Graphics Generation Model (NeurIPS 2025 · similarity 0.89)
- Chat2SVG: Vector Graphics Generation with Large Language Models and Image Diffusion Models (CVPR 2025 · similarity 0.89)
- R4-CGQA: Retrieval-based Vision Language Models for Computer Graphics Image Quality Assessment (CVPR 2026 · similarity 0.87)
- VisJudge-Bench: Aesthetics and Quality Assessment of Visualizations (ICLR 2026 · 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.
