Scientific Graphics Program Synthesis via Dual Self-Consistency Reinforcement Learning

ArXi:2604.06079v1 Announce Type: cross Graphics Program Synthesis is pivotal for interpreting and editing visual data, effectively facilitating the reverse-engineering of static visuals into editable TikZ code. While TikZ is the de facto standard for scientific schematics due to its programmatic flexibility, its requirement for rigorous spatial precision presents a significant challenge for Multimodal Large Language Models.