from dataclasses import dataclass from enum import Enum @dataclass class Task: benchmark: str metric: str metric_type: str col_name: str # Select your tasks here # --------------------------------------------------- class Tasks(Enum): # task_key in the json file, metric_key in the json file, name to display in the leaderboard #task1 = Task("text-entailment_1", "acc", "CPS", "TE") #task2 = Task("text-entailment_2", "acc", "average_accuracy", "TE Prompt Average") #task3 = Task("text-entailment_3", "acc", "std_accuracy", "TE Prompt Std") #task4 = Task("text-entailment_4", "acc", "best_prompt", "TE Best Prompt") #task5 = Task("text-entailment_5", "acc", "prompt_id", "TE Best Prompt Id") #task6 = Task("sentiment-analysis_1", "acc", "CPS", "SA") #task7 = Task("sentiment-analysis_2", "acc", "average_accuracy", "SA Prompt Average") #task8 = Task("sentiment-analysis_3", "acc", "std_accuracy", "SA STD Accuracy") #task9 = Task("sentiment-analysis_4", "acc", "best_prompt", "SA Best Prompt") #task10 = Task("sentiment-analysis_5", "acc", "prompt_id", "SA Best Prompt Id") #task11 = Task("hate-speech-detection_1", "acc", "CPS", "HS") #task12 = Task("hate-speech-detection_2", "acc", "average_accuracy", "HS Prompt Average") #task13 = Task("hate-speech-detection_3", "acc", "std_accuracy", "HS Prompt Std") #task14 = Task("hate-speech-detection_4", "acc", "best_prompt", "HS Best Prompt") #task15 = Task("hate-speech-detection_5", "acc", "prompt_id", "HS Best Prompt Id") #task16 = Task("admission-test_1", "acc", "CPS", "AT") #task17 = Task("admission-test_2", "acc", "average_accuracy", "AT Prompt Average") #task18 = Task("admission-test_3", "acc", "std_accuracy", "AT Prompt Std") #task19 = Task("admission-test_4", "acc", "best_prompt", "AT Best Prompt") #task20 = Task("admission-test_5", "acc", "prompt_id", "AT Best Prompt Id") #task21 = Task("word-in-context_1", "acc", "CPS", "WIC") #task22 = Task("word-in-context_2", "acc", "average_accuracy", "WIC Prompt Average") #task23 = Task("word-in-context_3", "acc", "std_accuracy", "WIC Prompt Std") #task24 = Task("word-in-context_4", "acc", "best_prompt", "WIC Best Prompt") #task25 = Task("word-in-context_5", "acc", "prompt_id", "WIC Best Prompt Id") #task26 = Task("faq_1", "acc", "CPS", "FAQ") #task27 = Task("faq_2", "acc", "average_accuracy", "FAQ Prompt Average") #task28 = Task("faq_3", "acc", "std_accuracy", "FAQ Prompt Std") #task29 = Task("faq_4", "acc", "best_prompt", "FAQ Best Prompt") #task30 = Task("faq_5", "acc", "prompt_id", "FAQ Best Prompt Id") #task31 = Task("lexical-substitution_1", "acc", "CPS", "LS") #task32 = Task("lexical-substitution_2", "acc", "average_accuracy", "LS Prompt Average") #task33 = Task("lexical-substitution_3", "acc", "std_accuracy", "LS Prompt Std") #task34 = Task("lexical-substitution_4", "acc", "best_prompt", "LS Best Prompt") #task35 = Task("lexical-substitution_5", "acc", "prompt_id", "LS Best Prompt Id") #task36 = Task("summarization-fanpage_1", "acc", "CPS", "SU") #task37 = Task("summarization-fanpage_2", "acc", "average_accuracy", "SU Prompt Average") #task38 = Task("summarization-fanpage_3", "acc", "std_accuracy", "SU Prompt Std") #task39 = Task("summarization-fanpage_4", "acc", "best_prompt", "SU Best Prompt") #task40 = Task("summarization-fanpage_5", "acc", "prompt_id", "SU Best Prompt Id") #task41 = Task("evalita NER_1", "acc", "CPS", "NER") #task42 = Task("evalita NER_2", "acc", "average_accuracy", "NER Prompt Average") #task43 = Task("evalita NER_3", "acc", "std_accuracy", "NER Prompt Std") #task44 = Task("evalita NER_4", "acc", "best_prompt", "NER Best Prompt") #task45 = Task("evalita NER_5", "acc", "prompt_id", "NER Best Prompt Id") #task46 = Task("relation-extraction_1", "acc", "CPS", "REL") #task47 = Task("relation-extraction_2", "acc", "average_accuracy", "REL Prompt Average") #task48 = Task("relation-extraction_5", "acc", "std_accuracy", "REL Prompt Std") #task49 = Task("relation-extraction_3", "acc", "best_prompt", "REL Best Prompt") #task50 = Task("relation-extraction_4", "acc", "prompt_id", "REL Best Prompt Id") task1 = Task("RE_1", "acc", "CPS", "REL-E3C") task2 = Task("RE_2", "acc", "average_accuracy", "REL-E3C Prompt Average") task3 = Task("RE_5", "acc", "std_accuracy", "REL-E3C Prompt Std") task4 = Task("RE_3", "acc", "best_prompt", "REL-E3C Best Prompt") task5 = Task("RE_4", "acc", "prompt_id", "REL-E3C Best Prompt Id") task6 = Task("NER_1", "acc", "CPS", "NER-E3C") task7 = Task("NER_2", "acc", "average_accuracy", "NER-E3C Prompt Average") task8 = Task("NER_3", "acc", "std_accuracy", "NER-E3C Prompt Std") task9 = Task("NER_4", "acc", "best_prompt", "NER-E3C Best Prompt") task10 = Task("NER_5", "acc", "prompt_id", "NER-E3C Best Prompt Id") task11 = Task("RML-CRF_1", "acc", "CPS", "CRF-RML") task12 = Task("RML-CRF_2", "acc", "average_accuracy", "CRF-RML Prompt Average") task13 = Task("RML-CRF_3", "acc", "std_accuracy", "CRF-RML Prompt Std") task14 = Task("RML-CRF_4", "acc", "best_prompt", "CRF-RML Best Prompt") task15 = Task("RML-CRF_5", "acc", "prompt_id", "CRF-RML Best Prompt Id") task16 = Task("DIA-CRF_1", "acc", "CPS", "CRF-DIA") task17 = Task("DIA-CRF_2", "acc", "average_accuracy", "CRF-DIA Prompt Average") task18 = Task("DIA-CRF_3", "acc", "std_accuracy", "CRF-DIA Prompt Std") task19 = Task("DIA-CRF_4", "acc", "best_prompt", "CRF-DIA Best Prompt") task20 = Task("DIA-CRF_5", "acc", "prompt_id", "CRF-DIA Best Prompt Id") task21 = Task("HIS-CRF_1", "acc", "CPS", "CRF-HIS") task22 = Task("HIS-CRF_2", "acc", "average_accuracy", "CRF-HIS Prompt Average") task23 = Task("HIS-CRF_3", "acc", "std_accuracy", "CRF-HIS Prompt Std") task24 = Task("HIS-CRF_4", "acc", "best_prompt", "CRF-HIS Best Prompt") task25 = Task("HIS-CRF_5", "acc", "prompt_id", "CRF-HIS Best Prompt Id") task26 = Task("NER-PHARMAER_1", "acc", "CPS", "NER-PHA") task27 = Task("NER-PHARMAER_2", "acc", "average_accuracy", "NER-PHA Prompt Average") task28 = Task("NER-PHARMAER_3", "acc", "std_accuracy", "NER-PHA Prompt Std") task29 = Task("NER-PHARMAER_4", "acc", "best_prompt", "NER-PHA Best Prompt") task30 = Task("NER-PHARMAER_5", "acc", "prompt_id", "NER-PHA Best Prompt Id") ''' task0 = Task("TextualEntailment", "acc", "Textual Entailment") task1 = Task("TextualEntailment_best", "acc", "TextualEntailment Best") task2 = Task("Sentiment Analysis", "acc", "Sentiment Analysis") task3 = Task("Sentiment Analysis_best", "acc", "Sentiment Analysis_best") task4 = Task("Hate Speech", "acc", "Hate Speech") task5 = Task("Hate Speech_best", "acc", "Hate Speech_best") task6 = Task("Admission Test", "acc", "Admission Test") task7 = Task("Admission Test_best", "acc", "Admission Test_best") task8 = Task("Word in Context", "acc", "Word in Context") task9 = Task("Word in Context_best", "acc", "Word in Context_best") task10 = Task("FAQ", "acc", "FAQ") task11 = Task("FAQ_best", "acc", "FAQ_best") task12 = Task("Lexical Substitution", "acc", "Lexical Substitution") task13 = Task("Lexical Substitution_best", "acc", "Lexical Substitution_best") task14 = Task("Summarization", "acc", "Summarization") task15 = Task("Summarization_best", "acc", "Summarization_best") task16 = Task("NER", "acc", "NER") task17 = Task("NER_best", "acc", "NER_best") task18 = Task("REL", "acc", "REL") task19 = Task("REL_best", "acc", "REL_best") ''' # Your leaderboard name TITLE = """

🚀 ECREAM-LLM Leaderboard 🚀

""" # What does your leaderboard evaluate? INTRODUCTION_TEXT = """

The eCream-LLM leaderboard , developed within the eCream Project (enabling Clinical Research in Emergency and Acute care Medicine), is designed to evaluate Large Language Models (LLMs) on several tasks pertaining to the medical domain. Its distinguishing features are:
(i) all tasks are implemented for six languages including English, Italian, Slovak, Slovenian, Polish and Greek;
(ii) the leaderboard includes generative tasks, allowing for a more natural interaction with LLMs;
(iii) all tasks are evaluated against multiple prompts, this way mitigating the model sensitivity to specific prompts and allowing a fairer evaluation.

**Generative tasks:** 🏷️NER-E3C (Named Entity Recognition - E3C), 🔗REL-E3C (Relation Extraction -E3C), 😃CRF-RML(CRF RML), NER-PHA ( Named Entity Recognition - PharamaER.IT)
**Multiple-choice task:** 🏥CRF-DIA (CRF Diagnosis), 📝CRF-HIS (CRF History) """ # Which evaluations are you running? how can people reproduce what you have? LLM_BENCHMARKS_TEXT = f""" ### Groups - `evalita-mp`: All tasks (perplexity and non-perplexity based). - `evalita-mp_gen`: Only generative tasks. #### Tasks The following Evalita-LLM tasks can also be evaluated in isolation: - `evalita-mp_ner_group`: Named Entity Recognition (NER) - `evalita-mp_re`: Relation Extraction (REL) ### Usage ```bash lm_eval --model hf --model_args pretrained=meta-llama/Llama-2-7b-hf --tasks evalita-mp_re --device cuda:0 --batch_size 1 ``` """ EVALUATION_QUEUE_TEXT = """ ## Some good practices before submitting a model ### 1) Make sure you can load your model and tokenizer using AutoClasses: ```python from transformers import AutoConfig, AutoModel, AutoTokenizer config = AutoConfig.from_pretrained("your model name", revision=revision) model = AutoModel.from_pretrained("your model name", revision=revision) tokenizer = AutoTokenizer.from_pretrained("your model name", revision=revision) ``` If this step fails, follow the error messages to debug your model before submitting it. It's likely your model has been improperly uploaded. Note: make sure your model is public! Note: if your model needs `use_remote_code=True`, we do not support this option yet but we are working on adding it, stay posted! ### 2) Convert your model weights to [safetensors](https://huggingface.co/docs/safetensors/index) It's a new format for storing weights which is safer and faster to load and use. It will also allow us to add the number of parameters of your model to the `Extended Viewer`! ### 3) Make sure your model has an open license! This is a leaderboard for Open LLMs, and we'd love for as many people as possible to know they can use your model 🤗 ### 4) Fill up your model card When we add extra information about models to the leaderboard, it will be automatically taken from the model card ## 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. If everything is done, check you can launch the EleutherAIHarness on your model locally, using the above command without modifications (you can add `--limit` to limit the number of examples per task). """ CITATION_BUTTON_LABEL = "Copy the following snippet to cite these results" CITATION_BUTTON_TEXT = r""" @article{magnini2025cost, title={A cost-effective approach to counterbalance the scarcity of medical datasets}, author={Magnini, Bernardo and Farzi, Saeed and Ferrazzi, Pietro and Ghosh, Soumitra and Lavelli, Alberto and Mezzanotte, Giulia and Speranza, Manuela}, journal={Frontiers in Disaster and Emergency Medicine}, volume={3}, pages={1558200}, year={2025}, publisher={Frontiers Media SA}, url={https://www.frontiersin.org/journals/disaster-and-emergency-medicine/articles/10.3389/femer.2025.1558200/full} } """