mlsanityData to 7 Oct 2026

Structural Evaluation Metrics for SVG Generation via Leave-One-Out Analysis

arXiv 2604.08809 ↗First posted 9 Apr 2026cs.LG stat.APcited by 1

Not on the map. Why?

Abstract (arXiv)

SVG generation is typically evaluated by comparing rendered outputs to reference images, which captures visual similarity but not the structural properties that make SVG editable, decomposable, and reusable. Inspired by the classical jackknife, we introduce element-level leave-one-out (LOO) analysis. The procedure renders the SVG with and without each element, which yields element-level signals for quality assessment and structural analysis. From this single mechanism, we derive (i) per-element quality scores that enable zero-shot artifact detection; (ii) element-concept attribution via LOO footprints crossed with VLM-grounded concept heatmaps; and (iii) four structural metrics: purity, coverage, compactness, and locality, which quantify SVG modularity from complementary angles. These metrics extend SVG evaluation from image similarity to code structure, enabling element-level diagnosis and comparison of how visual concepts are represented, partitioned, and organized within SVG code. Their practical relevance is validated on over 19,000 edits (5 types) across 5 generation systems and 3 complexity tiers.

Key facts. Cited by 1 paper in the map (data through 7 Oct 2026). Builds on 19 of its 32 references that are in the map. First posted on arXiv 9 Apr 2026.

Builds on

19 of 32 references.

References · 19

  1. 2021Learning Transferable Visual Models From Natural Language Supervisioncited by 33,456
  2. 2021Evaluating Large Language Models Trained on Codecited by 8,053
  3. 2023Let's Verify Step by Stepcited by 3,598
  4. 2017Understanding Black-box Predictions via Influence Functionscited by 1,714
  5. 2022Solving math word problems with process- and outcome-based feedbackcited by 871
  6. 2021Image Segmentation Using Text and Image Promptscited by 509
  7. 2023Instruct-NeRF2NeRF: Editing 3D Scenes with Instructionscited by 482
  8. 2017A Neural Representation of Sketch Drawingscited by 311
  9. 2020DeepSVG: A Hierarchical Generative Network for Vector Graphics Animationcited by 158
  10. 2023StarVector: Generating Scalable Vector Graphics Code from Images and Textcited by 96
  11. 2025OmniSVG: A Unified Scalable Vector Graphics Generation Modelcited by 71
  12. 2025Rendering-Aware Reinforcement Learning for Vector Graphics Generationcited by 36
7 more
  1. 2024Text-to-Vector Generation with Neural Path Representationcited by 32
  2. 2025NeuralSVG: An Implicit Representation for Text-to-Vector Generationcited by 16
  3. 2025SVGenius: Benchmarking LLMs in SVG Understanding, Editing and Generationcited by 16
  4. 2025SVGen: Interpretable Vector Graphics Generation with Large Language Modelscited by 13
  5. 2026LICA: Layered Image Composition Annotations for Graphic Design Researchcited by 6
  6. 2026Reliable Reasoning in SVG-LLMs via Multi-Task Multi-Reward Reinforcement Learningcited by 6
  7. 2026Graphic-Design-Bench: A Comprehensive Benchmark for Evaluating AI on Graphic Design Taskscited by 6

Built on by

Cited by year (2026 to 7 Oct)

Citing papers per year: 2026: 1.

1 paper.

  1. 2026GeoSVG-RL: Geometry-Aware Reinforcement Learning for Layout-Constrained Text-to-SVG Diagram Generationcited by 2

Similar, not cited

Close in wording, no citation either way

  1. 2023Beyond Pixels: Exploring Human-Readable SVG Generation for Simple Images with Vision Language Models
  2. 2025SVGauge: Towards Human-Aligned Evaluation for SVG Generation
  3. 2025SVGThinker: Instruction-Aligned and Reasoning-Driven Text-to-SVG Generation
  4. 2026SVG-Score: Human-Aligned Evaluation of Text-to-SVG Generation
  5. 2026VFIG: Vectorizing Complex Figures in SVG with Vision-Language Models
  6. 2025UniSVG: A Unified Dataset for Vector Graphic Understanding and Generation with Multimodal Large Language Models
  7. 2026VectorGym: A Multi-Task Benchmark for SVG Code Generation, Sketching and Editing
  8. 2024Chat2SVG: Vector Graphics Generation with Large Language Models and Image Diffusion Models