Exaone 4.0 VL: Vision-Language Foundation Model for Enterprise AI Agent

Abstract

🚀 Excited to share our latest achievement at LG AI Research! Our team has released Exaone-4.0-VL, our next-generation Vision-Language Foundation Model, at the LG AI Research Talk Concert 2025. The 32B model achieves leading performance across multi-modal LLM benchmarks, excelling in both perception (DocVQA, InfoVQA, TextVQA) and reasoning (MMMU, AI2D, ChartQA). All training stages—from large-scale pre-training to post-training (SFT, DPO)—were conducted on our in-house Exaone platform with web-scale multi-modal data. We believe Exaone-4.0-VL will accelerate transfer learning and fine-tuning for domain-specific enterprise AI agents, enabling impactful real-world applications. Tech report will be released soon!

Jongmin Lee

Assistant Professor of Computer Science Engineering

My research focuses on computer vision and machine learning, with interests in visual geometry, 3D vision, and spatial reasoning with multi-modal LLMs. I explore applications in autonomous systems, AR/VR, robotics, and physical AI.