NuplanQA-UQ

This repository contains the code, experimental results, dataset, and scripts required to reproduce the findings of the article: "Quantifying Prediction Uncertainty of Large Vision-Language Models for Autonomous Driving: An Empirical Study" The open-source dataset is available at: [NuplanQA-UQ Dataset.rar] Procedure for reproducing the experiments - [Code & Experimental Results.rar]1. Use the scripts in the [1-Data Preparation] to download the NuPlan and NuPlanQA-Eval datasets. 2. Use the scripts in the [2-Run LVLMs] to generate predictions with the 10 open-source LVLMs used in our empirical study, and obtain the raw experimental outputs. 3. Use the scripts in the [3-Calculation], along with the raw outputs, to compute the evaluation metrics reported in the paper. 4. The folder [4-Experimental Results] contains the final experimental results of this study. These data support the direct reproduction of the uncertainty quantification results and other findings reported in the paper.