Summary of Benchmarking Large Multimodal Models Against Common Corruptions, by Jiawei Zhang et al.
Benchmarking Large Multimodal Models against Common Corruptions
by Jiawei Zhang, Tianyu Pang, Chao Du, Yi Ren, Bo Li, Min Lin
First submitted to arxiv on: 22 Jan 2024
Categories
- Main: Machine Learning (cs.LG)
- Secondary: Computation and Language (cs.CL); Cryptography and Security (cs.CR); Computer Vision and Pattern Recognition (cs.CV); Multimedia (cs.MM)
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 addresses a gap in evaluating large multimodal models (LMMs) by analyzing their self-consistency under various corruptions. The study examines cross-modal interactions between text, image, and speech for four essential tasks: text-to-image, image-to-text, text-to-speech, and speech-to-text. To facilitate this evaluation, the authors create a comprehensive benchmark called MMCBench, which encompasses over 100 popular LMMs (150 model checkpoints). The benchmark’s availability allows for evaluating the reliability of these cutting-edge models under common corruptions, making it essential for practical deployment. By doing so, this paper contributes to a better understanding of LMMs’ capabilities and limitations. |
| Low | GrooveSquid.com (original content) | Low Difficulty Summary Large multimodal models can generate text, images, or speech, but how well do they work together? This study looks at how these models perform when given tasks like turning text into an image or speech. The authors create a special testing tool called MMCBench that lets them compare many different models on the same tasks. They want to see if these models can handle common mistakes, like typos or blurry images. By doing this, they hope to help people use these models in real-life situations. |
