Needle In A Multimodal Haystack
Weiyun Wang, Shuibo Zhang, Yiming Ren, Yuchen Duan, Tiantong Li, Shuo Liu, Mengkang Hu, Zhe Chen, Kaipeng Zhang, Lewei Lu, Xizhou Zhu, Ping Luo, Yu Qiao, Jifeng Dai, Wenqi Shao, Wenhai Wang
2024-06-17
Summary
This paper introduces Needle In A Multimodal Haystack (MM-NIAH), a new benchmark created to test how well large multimodal language models (MLLMs) can understand long documents that contain both text and images. It focuses on evaluating the models' abilities to retrieve information, count items, and reason based on the content in these documents.
What's the problem?
As MLLMs become more advanced, it’s important to evaluate their understanding of complex, long documents that combine different types of information. However, existing benchmarks often do not adequately test this ability, especially when it comes to visual information. This lack of comprehensive evaluation makes it hard to know how well these models perform in real-world scenarios where they need to process and understand long texts and images together.
What's the solution?
To address this issue, the authors developed MM-NIAH, which includes three types of tasks: retrieval (finding specific information), counting (determining how many items are present), and reasoning (making inferences based on the information). The benchmark uses a variety of documents with key information scattered throughout, requiring the models to extract and process this information effectively. The authors tested several leading MLLMs on these tasks and found that there is still a lot of room for improvement in their performance.
Why it matters?
This research is significant because it provides a structured way to evaluate how well AI models can comprehend complex multimodal documents. By highlighting the challenges that current models face, MM-NIAH aims to encourage further research and development in this area. Improving these models will be crucial for applications in fields like education, healthcare, and data analysis, where understanding detailed information from multiple sources is essential.
Abstract
With the rapid advancement of multimodal large language models (MLLMs), their evaluation has become increasingly comprehensive. However, understanding long multimodal content, as a foundational ability for real-world applications, remains underexplored. In this work, we present Needle In A Multimodal Haystack (MM-NIAH), the first benchmark specifically designed to systematically evaluate the capability of existing MLLMs to comprehend long multimodal documents. Our benchmark includes three types of evaluation tasks: multimodal retrieval, counting, and reasoning. In each task, the model is required to answer the questions according to different key information scattered throughout the given multimodal document. Evaluating the leading MLLMs on MM-NIAH, we observe that existing models still have significant room for improvement on these tasks, especially on vision-centric evaluation. We hope this work can provide a platform for further research on long multimodal document comprehension and contribute to the advancement of MLLMs. Code and benchmark are released at https://github.com/OpenGVLab/MM-NIAH.