Self-Distillation Policy Optimization via Visual Feedback: Bridging Code and Visual Artifacts
Quick Answer
The Visual-SDPO framework enhances code-generated visual artifacts by utilizing visual feedback for self-distillation, improving performance by over 10 points on benchmarks like ChartMimic and Design2Code, with fewer training steps and no added inference costs.
Key Points
- Visual-SDPO improves visual artifact generation by treating rendered feedback as privileged context.
- Introduces Visual-Grounded Code Credit Weighting to target specific code statements for defect correction.
- Achieves over 10 absolute points improvement on benchmarks with fewer training steps.
- Utilizes a unified Qwen3-VL-8B-Instruct backbone for various visual generation tasks.
- Maintains learnability of failed executions through self-distillation paths.
Paper Resources
Source Excerpt
arXiv:2606. 10334v1 Announce Type: new Abstract: Code-generating (LLMs) increasingly produce visual artifacts such as charts, web pages, and slides by writing programs that are executed by non-differentiable renderers, committing to code before observing the render. As a result, otherwise executable code often yields artifacts with visually salient defects, including overlapping elements, clipped text, broken alignment, low contrast, and overflow.
We study visual-feedback self-distillation for code-generated visual artifacts. …
Want this in your inbox every morning?
Daily brief at your local 8am — bilingual EN/中文, free.
More from arXiv cs.AI
See more →AINTMA: Agentic AI Architecture for Autonomous Test Management with Generative Intelligence, Secure Cloud Communication and Adaptive Quality Analytics
AINTMA, an autonomous test management architecture utilizing six specialized AI agents, achieves 88.4% test prioritization accuracy and reduces defect escape rates from 8.3% to 2.1%. The system demonstrates a 340% ROI within nine months, showcasing the potential of agentic AI in enhancing software quality management in cloud environments.