2.8m Gmail.txt -

: Uses 11k pairs with a balance of textual and visual rewards (

: Uses 22k data pairs focusing on textual accuracy ( 2.8M GMAIL.txt

) used in the RL stages or the used to measure the success of the 2.8M dataset? : Uses 11k pairs with a balance of

: The model is tested on subsets ranging from 200k to 2.8 million samples. The authors use a specific of chart-to-code data

The paper addresses the "SFT plateau," a phenomenon where Supervised Fine-Tuning (SFT) performance on Large Language Models (LLMs) stops improving even as the dataset size increases [11, 22]. The authors use a specific of chart-to-code data to demonstrate this limitation and propose Multimodal Structured Reinforcement Learning (MSRL) as a solution [11, 22]. 2. Methodology Supervised Fine-Tuning (SFT) Phase : Baseline Model : Qwen2.5-VL-7B-Instruct [11, 22].

: Increasing data from 2M to 2.8M results in no further performance gains, confirming the plateau [22]. Multimodal Structured Reinforcement Learning (MSRL) :

) to ensure the generated code matches the visual intent [11].