Breaking the SFT Plateau: Multimodal Structured Reinforcement Learning for Chart-to-Code Generation

ArXi:2508.13587v2 Announce Type: replace While reinforcement learning (RL) has proven highly effective for general reasoning in vision-language models, its application to tasks requiring deep understanding of information-rich images and structured output generation remains underexplored. Chart-to-code generation exemplifies this challenge, demanding complex reasoning over visual charts to produce structured code. Supervised fine-tuning (SFT) alone is often insufficient, highlighting the need for effective RL strategies tailored to structured outputs.