# Vector-Bench: Can Models Surgically Edit SVG Code?

arXiv: https://arxiv.org/abs/2607.19056 · HTML: https://mlsanity.com/paper/2607.19056/

## Key facts

- Cited by 1 papers in the map (data through 14 Oct 2026).
- Builds on 5 of its 9 references that are in the map.
- First posted on arXiv 21 Jul 2026.

## Builds on (most cited first, 5 of 5)

- [StarVector: Generating Scalable Vector Graphics Code from Images and Text](https://mlsanity.com/paper/2312.11556/) (2023)
- [IconShop: Text-Guided Vector Icon Synthesis with Autoregressive Transformers](https://mlsanity.com/paper/2304.14400/) (2023)
- [Empowering LLMs to Understand and Generate Complex Vector Graphics](https://mlsanity.com/paper/2412.11102/) (2024)
- [SVGEditBench: A Benchmark Dataset for Quantitative Assessment of LLM's SVG Editing Capabilities](https://mlsanity.com/paper/2404.13710/) (2024)
- [VectorEdits: A Dataset and Benchmark for Instruction-Based Editing of Vector Graphics](https://mlsanity.com/paper/2506.15903/) (2025)

## Built on by (most cited first, 1 of 1)

- [Beyond Plausibility: Verifiable Fine-Grained Image Editing on Structured Assets](https://mlsanity.com/paper/2610.04381/) (2026)

## Abstract (arXiv)

Instruction-based vector editing requires two capabilities: making a requested change and leaving everything else alone. The second is easy to miss when an output is judged only as a raster image. We introduce Vector-Bench, a compact, difficult benchmark of 40 SVG repair tasks. Each task pairs a corrupted SVG program with an author-written visual instruction, a hidden target program, 5.05 annotated repairs on average, and an average of 60.55 protected objects. Instructions describe visible defects without exposing element identifiers, coordinates, color codes, or path data. We define a deterministic binary specification reward: requested repairs use attribute-aware perceptual tolerances, while unrequested rendering- or application-relevant structure must remain semantically unchanged and the result must be a valid SVG. Canonical target equality and stricter source fidelity are retained as diagnostics. Validity-gated repair progress, a near-complete tier, and valid-output Unintended Change Rate (UCR) explain partial outcomes. We evaluate 34 model endpoints (25 listed as open-weight, 5 inexpensive controls, and 4 frontier closed endpoints) over 1360 requests. The strongest endpoint reaches only 15.0% full specification success, despite 43.7% mean repair progress, showing that apparent repair progress and specification-faithful editing remain substantially different. All prompts, outputs, scoring code, costs, and per-task reports are released.