) to ensure the generated code matches the visual intent [11].
: Qwen2.5-VL-72B-Instruct is used as the judge model for calculating visual rewards during training [11]. 4. Experimental Results 2.8M GMAIL.txt
To break the plateau, the authors implement a two-stage Reinforcement Learning (RL) process [11]. ) to ensure the generated code matches the
: Increasing data from 2M to 2.8M results in no further performance gains, confirming the plateau [22]. Multimodal Structured Reinforcement Learning (MSRL) : 2.8M GMAIL.txt