_version_ 1866912336973922304
author Zhu, Jinguo
Wang, Weiyun
Chen, Zhe
Liu, Zhaoyang
Ye, Shenglong
Gu, Lixin
Tian, Hao
Duan, Yuchen
Su, Weijie
Shao, Jie
Gao, Zhangwei
Cui, Erfei
Wang, Xuehui
Cao, Yue
Liu, Yangzhou
Wei, Xingguang
Zhang, Hongjie
Wang, Haomin
Xu, Weiye
Li, Hao
Wang, Jiahao
Deng, Nianchen
Li, Songze
He, Yinan
Jiang, Tan
Luo, Jiapeng
Wang, Yi
He, Conghui
Shi, Botian
Zhang, Xingcheng
Shao, Wenqi
He, Junjun
Xiong, Yingtong
Qu, Wenwen
Sun, Peng
Jiao, Penglong
Lv, Han
Wu, Lijun
Zhang, Kaipeng
Deng, Huipeng
Ge, Jiaye
Chen, Kai
Wang, Limin
Dou, Min
Lu, Lewei
Zhu, Xizhou
Lu, Tong
Lin, Dahua
Qiao, Yu
Dai, Jifeng
Wang, Wenhai
author_facet Zhu, Jinguo
Wang, Weiyun
Chen, Zhe
Liu, Zhaoyang
Ye, Shenglong
Gu, Lixin
Tian, Hao
Duan, Yuchen
Su, Weijie
Shao, Jie
Gao, Zhangwei
Cui, Erfei
Wang, Xuehui
Cao, Yue
Liu, Yangzhou
Wei, Xingguang
Zhang, Hongjie
Wang, Haomin
Xu, Weiye
Li, Hao
Wang, Jiahao
Deng, Nianchen
Li, Songze
He, Yinan
Jiang, Tan
Luo, Jiapeng
Wang, Yi
He, Conghui
Shi, Botian
Zhang, Xingcheng
Shao, Wenqi
He, Junjun
Xiong, Yingtong
Qu, Wenwen
Sun, Peng
Jiao, Penglong
Lv, Han
Wu, Lijun
Zhang, Kaipeng
Deng, Huipeng
Ge, Jiaye
Chen, Kai
Wang, Limin
Dou, Min
Lu, Lewei
Zhu, Xizhou
Lu, Tong
Lin, Dahua
Qiao, Yu
Dai, Jifeng
Wang, Wenhai
contents We introduce InternVL3, a significant advancement in the InternVL series featuring a native multimodal pre-training paradigm. Rather than adapting a text-only large language model (LLM) into a multimodal large language model (MLLM) that supports visual inputs, InternVL3 jointly acquires multimodal and linguistic capabilities from both diverse multimodal data and pure-text corpora during a single pre-training stage. This unified training paradigm effectively addresses the complexities and alignment challenges commonly encountered in conventional post-hoc training pipelines for MLLMs. To further improve performance and scalability, InternVL3 incorporates variable visual position encoding (V2PE) to support extended multimodal contexts, employs advanced post-training techniques such as supervised fine-tuning (SFT) and mixed preference optimization (MPO), and adopts test-time scaling strategies alongside an optimized training infrastructure. Extensive empirical evaluations demonstrate that InternVL3 delivers superior performance across a wide range of multi-modal tasks. In particular, InternVL3-78B achieves a score of 72.2 on the MMMU benchmark, setting a new state-of-the-art among open-source MLLMs. Its capabilities remain highly competitive with leading proprietary models, including ChatGPT-4o, Claude 3.5 Sonnet, and Gemini 2.5 Pro, while also maintaining strong pure-language proficiency. In pursuit of open-science principles, we will publicly release both the training data and model weights to foster further research and development in next-generation MLLMs.
format Preprint
id arxiv_https___arxiv_org_abs_2504_10479
institution arXiv
publishDate 2025
record_format arxiv
spellingShingle InternVL3: Exploring Advanced Training and Test-Time Recipes for Open-Source Multimodal Models
Zhu, Jinguo
Wang, Weiyun
Chen, Zhe
Liu, Zhaoyang
Ye, Shenglong
Gu, Lixin
Tian, Hao
Duan, Yuchen
Su, Weijie
Shao, Jie
Gao, Zhangwei
Cui, Erfei
Wang, Xuehui
Cao, Yue
Liu, Yangzhou
Wei, Xingguang
Zhang, Hongjie
Wang, Haomin
Xu, Weiye
Li, Hao
Wang, Jiahao
Deng, Nianchen
Li, Songze
He, Yinan
Jiang, Tan
Luo, Jiapeng
Wang, Yi
He, Conghui
Shi, Botian
Zhang, Xingcheng
Shao, Wenqi
He, Junjun
Xiong, Yingtong
Qu, Wenwen
Sun, Peng
Jiao, Penglong
Lv, Han
Wu, Lijun
Zhang, Kaipeng
Deng, Huipeng
Ge, Jiaye
Chen, Kai
Wang, Limin
Dou, Min
Lu, Lewei
Zhu, Xizhou
Lu, Tong
Lin, Dahua
Qiao, Yu
Dai, Jifeng
Wang, Wenhai
Computer Vision and Pattern Recognition
We introduce InternVL3, a significant advancement in the InternVL series featuring a native multimodal pre-training paradigm. Rather than adapting a text-only large language model (LLM) into a multimodal large language model (MLLM) that supports visual inputs, InternVL3 jointly acquires multimodal and linguistic capabilities from both diverse multimodal data and pure-text corpora during a single pre-training stage. This unified training paradigm effectively addresses the complexities and alignment challenges commonly encountered in conventional post-hoc training pipelines for MLLMs. To further improve performance and scalability, InternVL3 incorporates variable visual position encoding (V2PE) to support extended multimodal contexts, employs advanced post-training techniques such as supervised fine-tuning (SFT) and mixed preference optimization (MPO), and adopts test-time scaling strategies alongside an optimized training infrastructure. Extensive empirical evaluations demonstrate that InternVL3 delivers superior performance across a wide range of multi-modal tasks. In particular, InternVL3-78B achieves a score of 72.2 on the MMMU benchmark, setting a new state-of-the-art among open-source MLLMs. Its capabilities remain highly competitive with leading proprietary models, including ChatGPT-4o, Claude 3.5 Sonnet, and Gemini 2.5 Pro, while also maintaining strong pure-language proficiency. In pursuit of open-science principles, we will publicly release both the training data and model weights to foster further research and development in next-generation MLLMs.
title InternVL3: Exploring Advanced Training and Test-Time Recipes for Open-Source Multimodal Models
topic Computer Vision and Pattern Recognition
url https://arxiv.org/abs/2504.10479