DeepVision-103K: A Visually Diverse, Broad-Coverage, and Verifiable Mathematical Dataset for Multimodal Reasoning

TL;DR

DeepVision-103K is a large, diverse dataset designed to improve multimodal reasoning in models by covering extensive mathematical topics and visual elements, leading to better performance and generalization.

Contribution

The paper introduces DeepVision-103K, a comprehensive dataset that enhances data diversity and coverage for reinforcement learning with verifiable rewards in multimodal reasoning.

Findings

Models trained on DeepVision-103K perform well on mathematical benchmarks.

DeepVision-103K improves models' visual perception and reasoning capabilities.

The dataset enables better generalization to multimodal reasoning tasks.

Abstract

Reinforcement Learning with Verifiable Rewards (RLVR) has been shown effective in enhancing the visual reflection and reasoning capabilities of Large Multimodal Models (LMMs). However, existing datasets are predominantly derived from either small-scale manual construction or recombination of prior resources, which limits data diversity and coverage, thereby constraining further gains in model performance. To this end, we introduce \textbf{DeepVision-103K}, a comprehensive dataset for RLVR training that covers diverse K12 mathematical topics, extensive knowledge points, and rich visual elements. Models trained on DeepVision achieve strong performance on multimodal mathematical benchmarks, and generalize effectively to general multimodal reasoning tasks. Further analysis reveals enhanced visual perception, reflection and reasoning capabilities in trained models, validating DeepVision's effectiveness for advancing multimodal reasoning. Data: \href{https://huggingface.co/datasets/skylenage/DeepVision-103K}{this url}.