from dataclasses import dataclass from enum import Enum import json @dataclass class Task: benchmark: str metric: str col_name: str # Init: to update with your specific keys def create_task_list(): # task_key in the json file, metric_key in the json file, name to display in the leaderboard with open("src/datasets.json") as f: data = json.load(f) groups = [] names = [] for d in data: groups.append(d['group']) names.append(d['name']) groups = list(set(groups)) tasks = [] grouped_tasks = [] for name in names: tasks.append(Task(name, "metric_name", name)) for group in groups: grouped_tasks.append(Task(group, "metric_name", group)) return tasks, grouped_tasks # Your leaderboard name TITLE = """

Azerbaijani LLM Leaderboard

""" # What does your leaderboard evaluate? INTRODUCTION_TEXT = """ ## Azerbaijani Open LLM sponsored by Kapital Bank The Azerbaijani Open LLM Leaderboard is sponsored by Kapital Bank to support and develop Azerbaijani language NLP. This leaderboard offers a clear and fair ranking of open-source Azerbaijani LLMs, helping researchers, developers, and the AI community work together to improve the quality and use of Azerbaijani language models. Through this platform, we hope to bring useful AI technology to the Azerbaijani language and encourage models that are both locally relevant and internationally competitive. Contributors: Special thanks to Bugra Sen, Emir Rahimov, Kavsar Huseynova, Mirakram Aghalarov and Reshad Veliyev for their valuable contributions. ## Partners This leaderboard is supported by Kapital Bank, LocalDocs, PRODATA LLC, and the R&D Center of Baku Higher Oil School. """ LLM_BENCHMARKS_TEXT = f""" ## Azerbaijani Open LLM sponsored by Kapital Bank The Azerbaijani Open LLM Leaderboard is sponsored by Kapital Bank to support and develop Azerbaijani language NLP. This leaderboard offers a clear and fair ranking of open-source Azerbaijani LLMs, helping researchers, developers, and the AI community work together to improve the quality and use of Azerbaijani language models. Through this platform, we hope to bring useful AI technology to the Azerbaijani language and encourage models that are both locally relevant and internationally competitive.PartnersThis leaderboard is supported by Kapital Bank, LocalDocs, PRODATA LLC, and the R&D Center of Baku Higher Oil School. """ LLM_DATASET_TEXT = f""" ## Banking Call Classification MC: 192 entries; multiple-choice classification for bank-client requests. ## Banking Exam MCQ: 200–300 multiple-choice questions based on university banking exam materials. ## Banking QA: 97 entries: Question-answer pairs on Azerbaijani banking topics. ## Wiki CQA: 97 entries from Azerbaijani Wikipedia, with context, questions, and answers. ## GSM8K: 44 grade-school math problems to test multi-step reasoning. ## ARC: Elementary science questions in Azerbaijani, testing knowledge and reasoning. ## Subject-Specific MCQs: Questions across topics like informatics, history, physics, and more, each with 100 multiple-choice questions for specific subject knowledge. """ EVALUATION_QUEUE_TEXT = """ ## Some good practices before submitting a model ### 1) Make sure your model exists on hub. ### 2) Make sure your model is public. ## In case of model failure If your model is displayed in the `FAILED` category, its execution stopped. Make sure you have followed the above steps first. Please contact us if you are facing any trouble! """