mlsanityData to 13 Oct 2026

VectorHarness: Recovering Editable, Relation-Preserving Structure from Scientific Graphics

arXiv 2609.13294 ↗First posted 9 Sep 2026cs.CVcited by 0

Not on the map. Why?

Abstract (arXiv)

Converting scientific graphics into editable representations remains a challenging problem for image-to-code generation because of their heterogeneous elements and complex layouts. Recent multi-agent reconstruction systems have advanced this line of work, but often follow a copy-paste paradigm: the reconstructed image closely resembles the original, while complex regions remain effectively uneditable. We instead formulate a different objective, raster-to-authoring reconstruction, which aims to recover an authoring representation that supports native, customized editing rather than mere visual replication. To this end, we present VectorHarness, a multi-agent framework for raster-to-authoring reconstruction that recovers heterogeneous components using type-appropriate native representations. Text, formulas, shapes, connectors, icons, charts, and tables are reconstructed as natively editable objects, while intrinsically image-based regions remain raster content. To systematically evaluate reconstruction quality, we introduce VectorHarness-Bench, which jointly assesses rendering fidelity, raster fallback coverage, executable object edits, and relation-preserving edits. Experiments show that VectorHarness improves executable edit success and relation preservation, reduces avoidable raster fallback, and maintains high visual fidelity across heterogeneous graphics.

Key facts. No citing papers in the map yet (data through 13 Oct 2026). Builds on 34 of its 45 references that are in the map. First posted on arXiv 9 Sep 2026.

Builds on

34 of 45 references.

References · 34

  1. 2020DeepSVG: A Hierarchical Generative Network for Vector Graphics Animationcited by 158
  2. 2022DePlot: One-shot visual language reasoning by plot-to-table translationcited by 158
  3. 2021Im2Vec: Synthesizing Vector Graphics without Vector Supervisioncited by 115
  4. 2022VectorFusion: Text-to-SVG by Abstracting Pixel-Based Diffusion Modelscited by 112
  5. 2023StarVector: Generating Scalable Vector Graphics Code from Images and Textcited by 96
  6. 2025ChartCoder: Advancing Multimodal Large Language Model for Chart-to-Code Generationcited by 89
  7. 2016Image-to-Markup Generation with Coarse-to-Fine Attentioncited by 79
  8. 2022Towards Layer-wise Image Vectorizationcited by 78
  9. 2023SVGDreamer: Text Guided SVG Generation with Diffusion Modelcited by 76
  10. 2025MakeAnything: Harnessing Diffusion Transformers for Multi-Domain Procedural Sequence Generationcited by 45
  11. 2025LayerTracer: Cognitive-Aligned Layered SVG Synthesis via Diffusion Transformercited by 43
  12. 2024FonTS: Text Rendering with Typography and Style Controlscited by 40
22 more
  1. 2024Chat2SVG: Vector Graphics Generation with Large Language Models and Image Diffusion Modelscited by 37
  2. 2022CLIPVG: Text-Guided Image Manipulation Using Differentiable Vector Graphicscited by 34
  3. 2025OmniConsistency: Learning Style-Agnostic Consistency from Paired Stylization Datacited by 32
  4. 2025RelationAdapter: Learning and Transferring Visual Relation with Diffusion Transformerscited by 26
  5. 2025EasyText: Controllable Diffusion Transformer for Multilingual Text Renderingcited by 25
  6. 2024DiffSim: Taming Diffusion Models for Evaluating Visual Similaritycited by 20
  7. 2025DiffDecompose: Layer-Wise Decomposition of Alpha-Composited Images via Diffusion Transformerscited by 20
  8. 2025LayerD: Decomposing Raster Graphic Designs into Layerscited by 20
  9. 2024SuperSVG: Superpixel-based Scalable Vector Graphics Synthesiscited by 19
  10. 2025WordCon: Word-level Typography Control in Scene Text Renderingcited by 16
  11. 2025Draw with Thought: Unleashing Multimodal Reasoning for Scientific Diagram Generationcited by 14
  12. 2025From Charts to Code: A Hierarchical Benchmark for Multimodal Modelscited by 13
  13. 2026AutoFigure-Edit: Generating Editable Scientific Illustrationcited by 13
  14. 2025MCA-Bench: A Multimodal Benchmark for Evaluating CAPTCHA Robustness Against VLM-based Attackscited by 10
  15. 2026CreatiParser: Generative Image Parsing of Raster Graphic Designs into Editable Layerscited by 9
  16. 2026Crafter: A Multi-Agent Harness for Editable Scientific Figure Generation from Diverse Inputscited by 9
  17. 2026VFIG: Vectorizing Complex Figures in SVG with Vision-Language Modelscited by 4
  18. 2025Semantic Document Derendering: SVG Reconstruction via Vision-Language Modelingcited by 3
  19. 2026TextGround4M: A Prompt-Aligned Dataset for Layout-Aware Text Renderingcited by 3
  20. 2026LiveFigure: Generating Editable Scientific Illustration with VLM Agentscited by 3
  21. 2026CharTide: Data-Centric Chart-to-Code Generation via Tri-Perspective Tuning and Inquiry-Driven Evolutioncited by 2
  22. 2026Residual Decoder Adapter: ID-Preserving Tokenizer Adaption for Autoregressive Text Renderingcited by 1

Built on by

Not cited yet.