from enum import Enum from dataclasses import dataclass # Define TaskInfo and Tasks as before @dataclass class TaskInfo: benchmark: str col_name: str metric: str # Replace these with actual subjects from your dataset class Tasks(Enum): History = TaskInfo(benchmark='History', col_name='History', metric='accuracy') Mathematics = TaskInfo(benchmark='Mathematics', col_name='Mathematics', metric='accuracy') Science = TaskInfo(benchmark='Science', col_name='Science', metric='accuracy') Geography = TaskInfo(benchmark='Geography', col_name='Geography', metric='accuracy') Literature = TaskInfo(benchmark='Literature', col_name='Literature', metric='accuracy') Art = TaskInfo(benchmark='Art', col_name='Art', metric='accuracy') Physics = TaskInfo(benchmark='Physics', col_name='Physics', metric='accuracy') Chemistry = TaskInfo(benchmark='Chemistry', col_name='Chemistry', metric='accuracy') Biology = TaskInfo(benchmark='Biology', col_name='Biology', metric='accuracy') ComputerScience = TaskInfo(benchmark='Computer Science', col_name='Computer Science', metric='accuracy') # Now include the variables expected by app.py TITLE = """

🌐 LLM Leaderboard for MMMLU Evaluation 🌐

""" INTRODUCTION_TEXT = """ Welcome to the LLM Leaderboard for the MMMLU dataset evaluation. This leaderboard displays the performance of various language models on the MMMLU dataset across different subjects. """ LLM_BENCHMARKS_TEXT = """ ## About the MMMLU Benchmark The Massive Multitask Multilingual Language Understanding (MMMLU) benchmark is designed to evaluate models on a wide range of subjects. ## How to Interpret the Leaderboard - **Model**: The name of the model evaluated. - **Average ⬆️**: The average accuracy across all subjects. - **Subject Columns**: The accuracy (%) for each individual subject. ## How to Submit Your Model Go to the **Submit here!** tab and provide your model details to have it evaluated and appear on the leaderboard. """ EVALUATION_QUEUE_TEXT = """ Below are the lists of models that have been evaluated, are currently being evaluated, or are pending evaluation. """ CITATION_BUTTON_LABEL = "Citation" CITATION_BUTTON_TEXT = """ If you use this leaderboard or the MMMLU dataset in your research, please cite: @article{your_citation_here, title={Your Title}, author={Your Name}, journal={Your Journal}, year={2024} } """