Saved in:
Bibliographic Details
Main Authors: Liang, Hao, Wu, Ruitao, Zeng, Bohan, Niu, Junbo, Zhang, Wentao, Dong, Bin
Format: Preprint
Published: 2025
Subjects:
Online Access:https://arxiv.org/abs/2509.06079
Tags: Add Tag
No Tags, Be the first to tag this record!
_version_ 1866908524315934720
author Liang, Hao
Wu, Ruitao
Zeng, Bohan
Niu, Junbo
Zhang, Wentao
Dong, Bin
author_facet Liang, Hao
Wu, Ruitao
Zeng, Bohan
Niu, Junbo
Zhang, Wentao
Dong, Bin
contents Multimodal reasoning remains a fundamental challenge in artificial intelligence. Despite substantial advances in text-based reasoning, even state-of-the-art models such as GPT-o3 struggle to maintain strong performance in multimodal scenarios. To address this gap, we introduce a caption-assisted reasoning framework that effectively bridges visual and textual modalities. Our approach achieved 1st place in the ICML 2025 AI for Math Workshop \& Challenge 2: SeePhys, highlighting its effectiveness and robustness. Furthermore, we validate its generalization on the MathVerse benchmark for geometric reasoning, demonstrating the versatility of our method. Our code is publicly available at https://github.com/OpenDCAI/SciReasoner.
format Preprint
id arxiv_https___arxiv_org_abs_2509_06079
institution arXiv
publishDate 2025
record_format arxiv
spellingShingle Multimodal Reasoning for Science: Technical Report and 1st Place Solution to the ICML 2025 SeePhys Challenge
Liang, Hao
Wu, Ruitao
Zeng, Bohan
Niu, Junbo
Zhang, Wentao
Dong, Bin
Computation and Language
Computer Vision and Pattern Recognition
Multimodal reasoning remains a fundamental challenge in artificial intelligence. Despite substantial advances in text-based reasoning, even state-of-the-art models such as GPT-o3 struggle to maintain strong performance in multimodal scenarios. To address this gap, we introduce a caption-assisted reasoning framework that effectively bridges visual and textual modalities. Our approach achieved 1st place in the ICML 2025 AI for Math Workshop \& Challenge 2: SeePhys, highlighting its effectiveness and robustness. Furthermore, we validate its generalization on the MathVerse benchmark for geometric reasoning, demonstrating the versatility of our method. Our code is publicly available at https://github.com/OpenDCAI/SciReasoner.
title Multimodal Reasoning for Science: Technical Report and 1st Place Solution to the ICML 2025 SeePhys Challenge
topic Computation and Language
Computer Vision and Pattern Recognition
url https://arxiv.org/abs/2509.06079