Summary of Autotrust: Benchmarking Trustworthiness in Large Vision Language Models For Autonomous Driving, by Shuo Xing et al.
AutoTrust: Benchmarking Trustworthiness in Large Vision Language Models for Autonomous Driving
by Shuo Xing, Hongyuan Hua, Xiangbo Gao, Shenzhe Zhu, Renjie Li, Kexin Tian, Xiaopeng Li, Heng Huang, Tianbao Yang, Zhangyang Wang, Yang Zhou, Huaxiu Yao, Zhengzhong Tu
First submitted to arxiv on: 19 Dec 2024
Categories
- Main: Computer Vision and Pattern Recognition (cs.CV)
- Secondary: Machine Learning (cs.LG); Robotics (cs.RO)
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 The paper introduces AutoTrust, a comprehensive benchmark for evaluating the trustworthiness of large vision-language models (VLMs) in autonomous driving. It addresses the critical issue of trustworthiness in DriveVLMs, which directly impacts public transportation safety. The authors constructed a dataset with over 10k unique scenes and 18k queries to investigate trustworthiness issues in driving scenarios. They evaluated six publicly available VLMs, including generalist and specialist models, and found that general VLMs like LLaVA-v1.6 and GPT-4o-mini outperform specialized models in terms of overall trustworthiness. The authors also identified vulnerabilities to sensitive information disclosure, adversarial attacks, and biased decision-making. They emphasize the importance of addressing these issues for public safety and welfare. |
| Low | GrooveSquid.com (original content) | Low Difficulty Summary The paper is about creating a way to measure how trustworthy big computer models are when it comes to driving cars automatically. These models are very good at understanding scenes and making decisions, but we need to make sure they’re reliable and safe. The researchers made a huge dataset with lots of different scenarios and questions to test the models. They looked at six popular models and found that some general ones were better than special ones designed just for driving. But all the models had problems sharing secrets, being tricked by bad data, or making unfair decisions. This is important because we need to make sure autonomous cars are safe for everyone. |
Keywords
» Artificial intelligence » Gpt
