Simple o3: Towards Interleaved Vision-Language Reasoning

Fuente: arXiv
Guardado en:
Detalles Bibliográficos
Autores principales: Wang, Ye, Chen, Qianglong, Li, Zejun, Wang, Siyuan, Guo, Shijie, Zhang, Zhirui, Wei, Zhongyu
Formato: Preprint
Publicado: 2025
Materias:
Acceso en línea:
Etiquetas: Agregar Etiqueta
Sin Etiquetas, Sea el primero en etiquetar este registro!
_version_ 1866916902921568256
author Wang, Ye
Chen, Qianglong
Li, Zejun
Wang, Siyuan
Guo, Shijie
Zhang, Zhirui
Wei, Zhongyu
author_facet Wang, Ye
Chen, Qianglong
Li, Zejun
Wang, Siyuan
Guo, Shijie
Zhang, Zhirui
Wei, Zhongyu
contents Multimodal Large Language Models (MLLMs) have shown impressive performance on vision-language tasks, but their long Chain-of-Thought (CoT) capabilities in multimodal scenarios remain underexplored. Inspired by OpenAI's o3 model, which emulates human-like ''thinking with image'' through iterative visual transformations and linguistic reasoning, we propose Simple o3, an end-to-end framework that integrates dynamic tool interactions (e.g., cropping, zooming, and reusing) into interleaved vision-language reasoning via supervised fine-tuning (SFT). Our approach features a scalable data synthesis pipeline that generates high-quality interleaved vision-language reasoning chains via an ''observe-reason-act'' cycle, complete with executable visual operations and rigorous verification, yielding the open-source TWI-Tools-146K dataset. Experimental results demonstrate Simple o3's superior performance on diverse benchmarks, outperforming existing approaches. By combining enhanced reasoning capabilities, Simple o3 establishes a powerful yet computationally affordable paradigm for advancing multimodal reasoning. Remarkably, we provide the first in-depth analysis of different interleaved reasoning strategies, offering insights into their impact on model performance. We found that by introducing additional visual tokens for interleaved vision-language reasoning, reusing and magnifying the original image significantly improves the model's visual reasoning and fine-grained perception, while image cropping based on precise visual grounding allows the model to effectively focus on key entities or regions, further enhancing its capabilities.
format Preprint
id arxiv_https___arxiv_org_abs_2508_12109
institution arXiv
publishDate 2025
record_format arxiv
spellingShingle Simple o3: Towards Interleaved Vision-Language Reasoning
Wang, Ye
Chen, Qianglong
Li, Zejun
Wang, Siyuan
Guo, Shijie
Zhang, Zhirui
Wei, Zhongyu
Computer Vision and Pattern Recognition
Artificial Intelligence
Multimodal Large Language Models (MLLMs) have shown impressive performance on vision-language tasks, but their long Chain-of-Thought (CoT) capabilities in multimodal scenarios remain underexplored. Inspired by OpenAI's o3 model, which emulates human-like ''thinking with image'' through iterative visual transformations and linguistic reasoning, we propose Simple o3, an end-to-end framework that integrates dynamic tool interactions (e.g., cropping, zooming, and reusing) into interleaved vision-language reasoning via supervised fine-tuning (SFT). Our approach features a scalable data synthesis pipeline that generates high-quality interleaved vision-language reasoning chains via an ''observe-reason-act'' cycle, complete with executable visual operations and rigorous verification, yielding the open-source TWI-Tools-146K dataset. Experimental results demonstrate Simple o3's superior performance on diverse benchmarks, outperforming existing approaches. By combining enhanced reasoning capabilities, Simple o3 establishes a powerful yet computationally affordable paradigm for advancing multimodal reasoning. Remarkably, we provide the first in-depth analysis of different interleaved reasoning strategies, offering insights into their impact on model performance. We found that by introducing additional visual tokens for interleaved vision-language reasoning, reusing and magnifying the original image significantly improves the model's visual reasoning and fine-grained perception, while image cropping based on precise visual grounding allows the model to effectively focus on key entities or regions, further enhancing its capabilities.
title Simple o3: Towards Interleaved Vision-Language Reasoning
topic Computer Vision and Pattern Recognition
Artificial Intelligence
url https://arxiv.org/abs/2508.12109