Understanding, Abstracting and Checking: Evoking Complicated Multimodal Reasoning in LMMs
Keywords: Complicated Reasoning, Large multimodal model
Abstract: The recent large multimodal models (LMMs) have demonstrated their impressive capability of image understanding. However, they still struggle to make complicated reasoning for solving a challenging multimodal problem. In this paper, we present UnAC (Understanding, Abstracting, and Checking), a novel multimodal prompting method, to synergize reasoning for complicated problems in the multimodal context of LMMs, such as GPT-4o, Gemini-1.5 and GPT-4V. To improve the understanding of the image and capture more details, we propose an adaptive visual prompting method to make LMMs able to focus on certain regions. An image abstracting prompting is designed to effectively extract information from images. Further, we propose a gradual self-checking scheme for leading to better reasoning by checking each decomposed sub-question and its answer. Extensive experiments on three public benchmarks -- MathVista, MM-Vet, and MMMU -- demonstrate the effectiveness of our method.
Primary Area: foundation or frontier models, including LLMs
Code Of Ethics: I acknowledge that I and all co-authors of this work have read and commit to adhering to the ICLR Code of Ethics.
Submission Guidelines: I certify that this submission complies with the submission instructions as described on https://iclr.cc/Conferences/2025/AuthorGuide.
Reciprocal Reviewing: I understand the reciprocal reviewing requirement as described on https://iclr.cc/Conferences/2025/CallForPapers. If none of the authors are registered as a reviewer, it may result in a desk rejection at the discretion of the program chairs. To request an exception, please complete this form at https://forms.gle/Huojr6VjkFxiQsUp6.
Anonymous Url: I certify that there is no URL (e.g., github page) that could be used to find authors’ identity.
No Acknowledgement Section: I certify that there is no acknowledgement section in this submission for double blind review.
Submission Number: 7435
Loading