Summary of Visual Question Decomposition on Multimodal Large Language Models, by Haowei Zhang et al.
Visual Question Decomposition on Multimodal Large Language Models
by Haowei Zhang, Jianzhe Liu, Zhen Han, Shuo Chen, Bailan He, Volker Tresp, Zhiqiang Xu, Jindong Gu
First submitted to arxiv on: 28 Sep 2024
Categories
- Main: Computation and Language (cs.CL)
- Secondary: Artificial Intelligence (cs.AI); Computer Vision and Pattern Recognition (cs.CV); Machine Learning (cs.LG)
GrooveSquid.com Paper Summaries
GrooveSquid.com’s goal is to make artificial intelligence research accessible by summarizing AI papers in simpler terms. Each summary below covers the same AI paper, written at different levels of difficulty. The medium difficulty and low difficulty versions are original summaries written by GrooveSquid.com, while the high difficulty version is the paper’s original abstract. Feel free to learn from the version that suits you best!
| Summary difficulty | Written by | Summary |
|---|---|---|
| High | Paper authors | High Difficulty Summary Read the original abstract here |
| Medium | GrooveSquid.com (original content) | Medium Difficulty Summary This paper explores the ability of Multimodal Large Language Models (MLLMs) to decompose complex questions into smaller, more manageable parts, a process known as visual question decomposition. To evaluate this capability, the authors propose a systematic framework including a dataset and several evaluation criteria. The results show that existing MLLMs struggle to produce high-quality sub-questions. To address this limitation, the authors introduce a finetuning dataset called DecoVQA+ and an efficient pipeline for enhancing the model’s question decomposition capability. The finetuned models demonstrate significant improvements in sub-question quality and selective decomposition policy, also achieving higher accuracy on VQA benchmark datasets. |
| Low | GrooveSquid.com (original content) | Low Difficulty Summary This paper looks at how well computers can break down complicated questions into simpler ones. This is important because it helps us understand what we’re asking the computer to do. The authors want to see if a special kind of computer program called Multimodal Large Language Models (MLLMs) are good at doing this. They make a test to see how well these computers do, and find that they don’t do very well. To help them get better, the authors create a new dataset and a way to train the computers. When they use this training method, the computers get much better at breaking down questions and answering them correctly. |
