Reliable Reasoning in SVG-LLMs via Multi-Task Multi-Reward Reinforcement Learning

• URL: http://arxiv.org/abs/2603.16189v1
• Date: Tue, 17 Mar 2026 07:16:30 GMT
• Title: Reliable Reasoning in SVG-LLMs via Multi-Task Multi-Reward Reinforcement Learning
• Abstract summary: We present Chain-of-Thought-Reinforcement Learning for SVG, a unified framework to expose the model's reasoning process during SVG generation.<n>By training the SVG model to generate group-level code, our framework significantly improves structural coherence and visual fidelity.<n>Our approach systematically enhances overall generation capabilities, achieving higher task success rates, superior SVG code quality, and exceptional visual fidelity.
• Score: 15.21004381065588
• License: http://creativecommons.org/licenses/by-nc-sa/4.0/
• Abstract: With the rapid advancement of vision-language models, an increasing number of studies have explored their potential for SVG generation tasks. Although existing approaches improve performance by constructing large-scale SVG datasets and introducing SVG-specific tokens, they still suffer from limited generalization, redundant paths in code outputs, and a lack of explicit reasoning. In this work, we present CTRL-S (Chain-of-Thought Reinforcement Learning for SVG), a unified framework that introduces a chain-of-thought mechanism to explicitly expose the model's reasoning process during SVG generation. To support this structured reasoning, we construct SVG-Sophia, a high-quality dataset containing 145K samples across SVG code refinement, Text-to-SVG, and Image-to-SVG tasks. By training the model to generate group-level structured SVG code, CTRL-S significantly improves structural coherence and visual fidelity. Furthermore, we adopt the GRPO algorithm and design a multi-reward optimization framework, incorporating DINO, image-text similarity, format, and code efficiency rewards. Through joint multi-reward optimization and multi-task training, our approach systematically enhances overall generation capabilities. Extensive experiments show that CTRL-S outperforms existing methods, achieving higher task success rates, superior SVG code quality, and exceptional visual fidelity.

Related papers

• SVGEval: A Vision-Grounded Framework for Perceptual-Quality Benchmarking and Evaluation in Text-to-SVG Generation [49.00672742747506]
  Multimodal large models are increasingly used to generate scalable vector graphics (SVG)<n>Existing protocols are often code-centric or borrow-image metrics after rendering SVGs.<n>We introduce SVGEval, a vision-grounded multimodal benchmark for human-aligned SVG quality assessment.
  arXiv  Detail & Related papers  (2026-08-03T09:41:18Z)
• DuetSVG: Unified Multimodal SVG Generation with Internal Visual Guidance [48.98604326855894]
  We introduce DuetSVG, a unified multimodal model that jointly generates image tokens and corresponding SVG tokens in an end-to-end manner.<n>At inference, we apply a novel test-time scaling strategy that leverages the model's native visual predictions as guidance to improve SVG decoding quality.
  arXiv  Detail & Related papers  (2025-12-11T18:23:03Z)
• RoboSVG: A Unified Framework for Interactive SVG Generation with Multi-modal Guidance [32.59099674596894]
  RoboSVG is a unified framework for generating interactive SVGs guided by textual, visual, and numerical signals.<n>To support this framework, we construct RoboDraw, a large-scale dataset of one million examples.<n>RoboSVG achieves superior query compliance and visual fidelity across tasks, establishing a new state of the art in versatile SVG generation.
  arXiv  Detail & Related papers  (2025-10-26T13:57:08Z)
• InternSVG: Towards Unified SVG Tasks with Multimodal Large Language Models [65.49118879021016]
  We present the InternSVG family, an integrated data-benchmark-model suite.<n>At its core is SAgoge, the largest and most comprehensive multimodal dataset for SVG tasks.<n>We propose InternSVG, a unified MLLM for SVG understanding, editing, and generation with SVG-specific special tokens.
  arXiv  Detail & Related papers  (2025-10-13T12:38:04Z)
• SVGen: Interpretable Vector Graphics Generation with Large Language Models [61.62816031675714]
  We introduce SVG-1M, a large-scale dataset of high-quality SVGs paired with natural language descriptions.<n>We create well-aligned Text to SVG training pairs, including a subset with Chain of Thought annotations for enhanced semantic guidance.<n>Based on this dataset, we propose SVGen, an end-to-end model that generates SVG code from natural language inputs.
  arXiv  Detail & Related papers  (2025-08-06T15:00:24Z)
• Reason-SVG: Hybrid Reward RL for Aha-Moments in Vector Graphics Generation [29.418375886989992]
  We introduce Reason-SVG, a framework designed to enhance Large Language Models (LLMs) reasoning for SVG generation.<n>Reason-SVG pioneers the "Drawing-with-Thought" (DwT) paradigm, in which models generate both SVG code and explicit design rationales.<n>We introduce the SVGX-DwT-10k dataset, a high-quality corpus of 10,000 SVG-DwT pairs, where each SVG code is generated based on explicit DwT reasoning.
  arXiv  Detail & Related papers  (2025-05-30T11:57:58Z)
• OmniSVG: A Unified Scalable Vector Graphics Generation Model [69.59073636922287]
  We propose OmniSVG, a unified framework that leverages pre-trained Vision-Language Models for end-to-end multimodal SVG generation.<n>By parameterizing SVG commands and coordinates into discrete tokens, OmniSVG decouples structural logic from low-level geometry for efficient training while maintaining the synthesis of complex SVG structure.<n>We introduce MMSVG-2M, a multimodal dataset with two million annotated SVG assets, along with a standardized evaluation protocol for conditional SVG generation tasks.
  arXiv  Detail & Related papers  (2025-04-08T17:59:49Z)
• NeuralSVG: An Implicit Representation for Text-to-Vector Generation [54.4153300455889]
  We propose NeuralSVG, an implicit neural representation for generating vector graphics from text prompts.<n>To encourage a layered structure in the generated SVG, we introduce a dropout-based regularization technique.<n>We demonstrate that NeuralSVG outperforms existing methods in generating structured and flexible SVG.
  arXiv  Detail & Related papers  (2025-01-07T18:50:06Z)

This list is automatically generated from the titles and abstracts of the papers in this site.

This site does not guarantee the quality of this site (including all information) and is not responsible for any consequences.