Crisp-Unimib/BEEP_eval

🚗 BEst DrivEr’s License Performer (BEEP) Dataset

BEEP is a challenge benchmark designed to evaluate large language models (LLMs) through a simulation of the Italian driver’s license exam. This dataset focuses on understanding traffic laws and reasoning through driving situations, replicating the complexity of the Italian licensing process.

---

📁 Dataset Structure

Column	Data Type	Description
Categorisation Structure	[String]	Hierarchical categorisation of major, minor, and subcategories for each question
Question Text	[String]	The actual content of the question
True Answer	[Boolean]	True or false answer
Figure	[String]	Reference to an accompanying figure, if present

Note: Questions are organised into a classification system that reflects the complexity of road rules and signage.

---

📊 Summary Statistics

• Total Questions: 2920
• Last Updated: 01/07/2020

---

🔍 Key Features

• Source: The dataset is derived from the publicly accessible official document "Listato A e B", provided by the Italian Ministry of Infrastructure and Transport. It includes all questions related to driver’s license categories A and B.
• Hierarchical Structure: Questions are classified into major categories, such as "Road Signage", and further subdivided into minor and subcategories for precise categorisation.
• Question Format: The dataset primarily consists of true/false questions aimed at evaluating knowledge of traffic laws, signage, and driving behavior.
• Exclusions: For the CALAMITA challenge, questions containing images are excluded, focusing solely on text-based questions.

---

🛠️ Using the Dataset

Loading Example

You can load this dataset in Python using pandas:

import pandas as pd

# Load the dataset
df = pd.read_csv('beep_data.csv')

# Display the first few rows of the dataset
print(df.head())

Citation

If you find our work interesting, please cite us:

BibTeX:

@inproceedings{mercorio2024beep,
  title={BEEP-BEst DrivEr’s License Performer: A CALAMITA Challenge},
  author={Mercorio, Fabio and Potert{\`\i}, Daniele and Serino, Antonio and Seveso, Andrea and others},
  booktitle={CEUR WORKSHOP PROCEEDINGS},
  volume={3878},
  year={2024}
}

APA:

Mercorio, F., Potertì, D., Serino, A., & Seveso, A. (2024). BEEP-BEst DrivEr’s License Performer: A CALAMITA Challenge. In CEUR WORKSHOP PROCEEDINGS (Vol. 3878).

Contact

Andrea Seveso - andrea.seveso@unimib.it