SimpleOCR: Rendering Visualized Questions to Teach MLLMs to Read

TL;DR

SimpleOCR is a training strategy that enhances multimodal models' ability to read embedded text in images by structurally forcing visual engagement, leading to significant performance improvements across benchmarks.

Contribution

It introduces SimpleOCR, a plug-and-play training method that mitigates modality laziness in MLLMs by structurally constraining training with visualized questions.

Findings

SimpleOCR improves model performance by up to 5.4% on OOD benchmarks.

Models show a 12.7% performance drop in the VQ setting without SimpleOCR.

SimpleOCR achieves high data efficiency, outperforming RL-based methods with fewer samples.

Abstract

Despite the rapid advancements in Multimodal Large Language Models (MLLMs), a critical question regarding their visual grounding mechanism remains unanswered: do these models genuinely ``read'' text embedded in images, or do they merely rely on parametric shortcuts in the text prompt? In this work, we diagnose this issue by introducing the Visualized-Question (VQ) setting, where text queries are rendered directly onto images to structurally mandate visual engagement. Our diagnostic experiments on Qwen2.5-VL reveal a startling capability-utilization gap: despite possessing strong OCR capabilities, models suffer a performance degradation of up to 12.7% in the VQ setting, exposing a deep-seated ``modality laziness.'' To bridge this gap, we propose SimpleOCR, a plug-and-play training strategy that imposes a structural constraint on the learning process. By transforming training samples into the VQ format with randomized styles, SimpleOCR effectively invalidates text-based shortcuts, compelling the model to activate and optimize its visual text extraction pathways. Empirically, SimpleOCR yields robust gains without architectural modifications. On four representative OOD benchmarks, it surpasses the base model by 5.4% and GRPO based on original images by 2.7%, while exhibiting extreme data efficiency, achieving superior performance with 30x fewer samples (8.5K) than recent RL-based methods. Furthermore, its plug-and-play nature allows seamless integration with advanced RL strategies like NoisyRollout to yield complementary improvements. Code is available at https://github.com/aiming-lab/SimpleOCR.