VARCO-VISION-2.0 Technical Report

2025/09/12 by Cha, Young-rok, J. W. Ju, Ju, Jeongho +8 · 1 citation
Computer Science · Engineering · #Advanced Vision and Imaging #Computation and Language (cs.CL) #Computer Vision and Pattern Recognition (cs.CV) #FOS: Computer and information sciences #Industrial Vision Systems and Defect Detection

paper · pdf · doi:10.48550/arxiv.2509.10105

openalex publication_date 2025/09/12 · openalex created_date 2025/10/18 · openalex updated_date 2026/07/28

Abstract

We introduce VARCO-VISION-2.0, an open-weight bilingual vision-language model (VLM) for Korean and English with improved capabilities compared to the previous model VARCO-VISION-14B. The model supports multi-image understanding for complex inputs such as documents, charts, and tables, and delivers layoutaware OCR by predicting both textual content and its spatial location. Trained with a four-stage curriculum with memory-efficient techniques, the model achieves enhanced multimodal alignment, while preserving core language abilities and improving safety via preference optimization. Extensive benchmark evaluations demonstrate strong spatial grounding and competitive results for both languages, with the 14B model achieving 8th place on the OpenCompass VLM leaderboard among models of comparable scale. Alongside the 14B-scale model, we release a 1.7B version optimized for on-device deployment. We believe these models advance the development of bilingual VLMs and their practical applications. Two variants of VARCO-VISION-2.0 are available at Hugging Face: a full-scale 14B model and a lightweight 1.7B model.

Citations

• Expanding Performance Boundaries of Open-Source Multimodal Models with Model, Data, and Test-Time Scaling
• VARCO-VISION: Expanding Frontiers in Korean Vision-Language Models
• Enhancing the Reasoning Ability of Multimodal Large Language Models via Mixed Preference Optimization
• Pangea: A Fully Open Multilingual Multimodal LLM for 39 Languages
• Liger Kernel: Efficient Triton Kernels for LLM Training
• Evaluating Visual and Cultural Interpretation: The K-Viscuit Benchmark with Human-VLM Collaboration
• A-Bench: Are LMMs Masters at Evaluating AI-generated Images?
• Ovis: Structural Embedding Alignment for Multimodal Large Language Model
• SEED-Bench-2-Plus: Benchmarking Multimodal Large Language Models with Text-Rich Visual Comprehension
• Are We on the Right Way for Evaluating Large Vision-Language Models?
• HallusionBench: An Advanced Diagnostic Suite for Entangled Language Hallucination and Visual Illusion in Large Vision-Language Models
• Ring Attention with Blockwise Transformers for Near-Infinite Context
• Q-Bench: A Benchmark for General-Purpose Foundation Models on Low-level Vision
• Efficient Memory Management for Large Language Model Serving with PagedAttention
• YaRN: Efficient Context Window Extension of Large Language Models
• MM-Vet: Evaluating Large Multimodal Models for Integrated Capabilities
• SEED-Bench: Benchmarking Multimodal LLMs with Generative Comprehension
• Kosmos-2: Grounding Multimodal Large Language Models to the World
• Judging LLM-as-a-Judge with MT-Bench and Chatbot Arena
• TIES-Merging: Resolving Interference When Merging Models
• Direct Preference Optimization: Your Language Model is Secretly a Reward Model
• Evaluating Object Hallucination in Large Vision-Language Models
• Large Language Models Can Self-Improve
• Learn to Explain: Multimodal Reasoning via Thought Chains for Science Question Answering
• Large Language Models are Zero-Shot Reasoners
• ChartQA: A Benchmark for Question Answering about Charts with Visual and Logical Reasoning
• Training language models to follow instructions with human feedback
• 8-bit Optimizers via Block-wise Quantization
• Compressing Visual-linguistic Model via Knowledge Distillation
• Measuring Massive Multitask Language Understanding
• DocVQA: A Dataset for VQA on Document Images
• DistilBERT, a distilled version of BERT: smaller, faster, cheaper and lighter
• Towards VQA Models That Can Read
• Averaging Weights Leads to Wider Optima and Better Generalization
• Real-Time Single Image and Video Super-Resolution Using an Efficient Sub-Pixel Convolutional Neural Network
• Training Deep Nets with Sublinear Memory Cost
• A Diagram Is Worth A Dozen Images
• Distilling the Knowledge in a Neural Network

Cited by

• Beyond Line-Level Filtering for the Pretraining Corpora of LLMs

Related

• VARCO-VISION: Expanding Frontiers in Korean Vision-Language Models
• Exploring OCR-augmented Generation for Bilingual VQA
• TowerVision: Understanding and Improving Multilinguality in Vision-Language Models
• MVL-SIB: A Massively Multilingual Vision-Language Benchmark for Cross-Modal Topical Matching
• POINTS1.5: Building a Vision-Language Model towards Real World Applications