Back to feed
arXiv cs.AI·

Self-Distillation Policy Optimization via Visual Feedback: Bridging Code and Visual Artifacts

Signal
78
Hype
25
In three linesVisual-SDPO, a self-distillation framework using visual feedback, improves code generation for visual artifacts (charts, web pages, slides). The method traces detected visual defects back to responsible code statements and amplifies distillation signals on those statements. On ChartMimic, Design2Code, and AeSlides benchmarks, it gains +10 absolute points over zero-shot and +2.4 over GRPO.
Read source
Your take?
Code generationVisionReinforcement learningBenchmarks

Summary generated by Claude — human-verified