SVG-Sophia: Multimodal Reasoning Data for SVG Generation via Reinforcement Learning

A dataset for training SVG-LLMs using multi-task, multi-reward reinforcement learning, created by InternSVG. It includes Chain-of-Thought training data for image-to-SVG and text-to-SVG tasks. The dataset page was last updated on 2026-03-18.

Use Cases

• Fine-tuning models for image-to-SVG conversion based on the described CoT training data.
• Training models for text-to-SVG generation using the provided Chain-of-Thought reasoning data.
• Developing reinforcement learning agents for SVG refinement tasks as indicated by the dataset's purpose.

Strengths

• Dataset is explicitly designed for multi-task, multi-reward reinforcement learning, a specific training methodology.
• Includes separate training data for distinct tasks: image-to-SVG and text-to-SVG.

Limitations

• Column-level documentation is absent; field semantics must be inferred after download.
• Row count, file size, and specific data formats are unknown, which may limit suitability assessment.

Provenance

Source

InternSVG

Collection Method

Likely created for training SVG-LLMs, but specific collection method is not detailed.

Freshness

Last updated 2026-03-18 02:47:41; freshness should be verified.