Vision-G1: Towards General Reasoning Vision-Language Models via Reinforcement Learning
Proceedings of the AAAI Conference on Artificial Intelligence · 14.03.2026 · 40 (33) · 28131-28139
DOI: 10.1609/aaai.v40i33.40039 ↗Kurz gefasst
Fachartikel in der Zeitschrift Proceedings of the AAAI Conference on Artificial Intelligence, online erschienen am 14.03.2026. Verfasst von Yuheng Zha und 9 weiteren Autor:innen. Die Arbeit wurde bisher 1-mal zitiert (Crossref).
Abstract (Originalsprache)
Recent vision-language models (VLMs) show strong reasoning capabilities through training with reinforcement learning from verifiable rewards (RLVR). Despite their impressive capabilities, current VLMs focus on a limited range of reasoning tasks, such as mathematical and logical reasoning, due to the lack of readily available verifiable reward data in broader domains. As a result, these models struggle to generalize their reasoning abilities to the wide variety of challenges encountered in real-world environments. To address this limitation, we collect and assemble a comprehensive RL-ready visual reasoning training dataset encompassing 46 datasets across 13 dimensions of 5 domains, covering a wide range of realistic scenarios such as infographic reasoning, mathematical reasoning, spatial reasoning, and general science reasoning. Based on this dataset, we propose an influence function-based data filtering strategy and a multi-round data curriculum method to iteratively strengthen general visual reasoning abilities. Using this approach, we train a general reasoning VLM, namely Vision-G1. Our 7B model achieves state-of-the-art performance across nine visual reasoning benchmarks, surpassing previous similar-sized VLMs and even GPT-4o and Gemini-1.5 Flash.
Zitieren
APA
Zha, Y., Zhou, K., Wu, Y., Wang, Y., Feng, J., Xu, Z., Hao, S., Liu, Z., Xing, E. P., & Hu, Z. (2026). Vision-G1: Towards General Reasoning Vision-Language Models via Reinforcement Learning. Proceedings of the AAAI Conference on Artificial Intelligence, 40(33), 28131-28139. https://doi.org/10.1609/aaai.v40i33.40039
Vancouver
Zha Y, Zhou K, Wu Y, Wang Y, Feng J, Xu Z, et al. Vision-G1: Towards General Reasoning Vision-Language Models via Reinforcement Learning. Proceedings of the AAAI Conference on Artificial Intelligence. 2026;40(33):28131-28139. doi:10.1609/aaai.v40i33.40039
BibTeX
@article{Zha2026,
author = {Zha, Yuheng and Zhou, Kun and Wu, Yujia and Wang, Yushu and Feng, Jie and Xu, Zhi and Hao, Shibo and Liu, Zhengzhong and Xing, Eric P. and Hu, Zhiting},
title = {Vision-G1: Towards General Reasoning Vision-Language Models via Reinforcement Learning},
journal = {Proceedings of the AAAI Conference on Artificial Intelligence},
year = {2026},
volume = {40},
number = {33},
pages = {28131--28139},
doi = {10.1609/aaai.v40i33.40039}
}
EUVIMED