ACL 2025 paper citation.

README.md

@@ -7,6 +7,8 @@ sdk: gradio
 app_file: app.py
 pinned: true
 license: apache-2.0
+tags:
+  - leaderboard
 short_description: Evaluation of language models on Maltese tasks
 sdk_version: 5.19.0
 ---

app.py

@@ -100,7 +100,7 @@ def init_leaderboard(dataframe):
 demo = gr.Blocks(css=custom_css)
 with demo:
     gr.HTML(TITLE)
-    gr.Markdown(INTRODUCTION_TEXT, elem_classes="markdown-text")
+    gr.HTML(INTRODUCTION_TEXT, elem_classes="markdown-text")
 
     with gr.Tabs(elem_classes="tab-buttons") as tabs:
         with gr.TabItem("🏅 LLM Benchmark", elem_id="llm-benchmark-tab-table", id=0):

src/about.py

@@ -1,3 +1,5 @@
+import re
+
 from dataclasses import dataclass
 from enum import Enum
 
@@ -18,6 +20,7 @@ class Task:
     benchmark: str
     metric: str
     col_name: str
+    url: str
     task_type: TaskType
     is_primary_metric: bool = True
 
@@ -26,22 +29,22 @@ class Task:
 # ---------------------------------------------------
 class Tasks(Enum):
     # task_key in the json file, metric_key in the json file, name to display in the leaderboard 
-    task0 = Task("sentiment_mlt", "f1", "Sentiment Analysis (F1)", TaskType.NLU)
-    task1 = Task("sib200_mlt", "f1", "SIB200 (F1)", TaskType.NLU)
-    task2 = Task("taxi1500_mlt", "f1", "Taxi1500 (F1)", TaskType.NLU)
-    task3 = Task("maltese_news_categories", "loglikelihood", "Maltese News Categories (F1)", TaskType.NLU)
-    task4 = Task("multieurlex_mlt", "loglikelihood", "MultiEURLEX (F1)", TaskType.NLU)
-    task5 = Task("belebele_mlt", "acc", "Belebele (Accuracy)", TaskType.NLU)
-    task6 = Task("opus100_eng-mlt", "bleu", "OPUS-100 EN→MT (BLEU)", TaskType.NLG, False)
-    task7 = Task("opus100_eng-mlt", "chrf", "OPUS-100 EN→MT (ChrF)", TaskType.NLG)
-    task8 = Task("flores200_eng-mlt", "bleu", "Flores-200 EN→MT (BLEU)", TaskType.NLG, False)
-    task9 = Task("flores200_eng-mlt", "chrf", "Flores-200 EN→MT (ChrF)", TaskType.NLG)
-    task10 = Task("webnlg_mlt", "chrf", "WebNLG (ChrF)", TaskType.NLG)
-    task11 = Task("webnlg_mlt", "rouge", "WebNLG (Rouge-L)", TaskType.NLG, False)
-    task12 = Task("eurlexsum_mlt", "chrf", "EUR-Lex-Sum (ChrF)", TaskType.NLG, False)
-    task13 = Task("eurlexsum_mlt", "rouge", "EUR-Lex-Sum (Rouge-L)", TaskType.NLG)
-    task14 = Task("maltese_news_headlines", "chrf", "Maltese News Headlines (ChrF)", TaskType.NLG, False)
-    task15 = Task("maltese_news_headlines", "rouge", "Maltese News Headlines (Rouge-L)", TaskType.NLG)
+    task0 = Task("sentiment_mlt", "f1", "Sentiment Analysis (F1)", "https://github.com/jerbarnes/typology_of_crosslingual/tree/master/data/sentiment/mt", TaskType.NLU)
+    task1 = Task("sib200_mlt", "f1", "SIB200 (F1)", "https://huggingface.co/datasets/Davlan/sib200/viewer/mlt_Latn", TaskType.NLU)
+    task2 = Task("taxi1500_mlt", "f1", "Taxi1500 (F1)", "https://github.com/cisnlp/Taxi1500", TaskType.NLU)
+    task3 = Task("maltese_news_categories", "loglikelihood", "Maltese News Categories (F1)", "https://huggingface.co/datasets/MLRS/maltese_news_categories", TaskType.NLU)
+    task4 = Task("multieurlex_mlt", "loglikelihood", "MultiEURLEX (F1)", "https://huggingface.co/datasets/nlpaueb/multi_eurlex", TaskType.NLU)
+    task5 = Task("belebele_mlt", "acc", "Belebele (Accuracy)", "https://huggingface.co/datasets/facebook/belebele/viewer/mlt_Latn", TaskType.NLU)
+    task6 = Task("opus100_eng-mlt", "bleu", "OPUS-100 EN→MT (BLEU)", "https://huggingface.co/datasets/MLRS/OPUS-MT-EN-Fixed", TaskType.NLG, False)
+    task7 = Task("opus100_eng-mlt", "chrf", "OPUS-100 EN→MT (ChrF)", "https://huggingface.co/datasets/MLRS/OPUS-MT-EN-Fixed", TaskType.NLG)
+    task8 = Task("flores200_eng-mlt", "bleu", "Flores-200 EN→MT (BLEU)", "https://huggingface.co/datasets/Muennighoff/flores200", TaskType.NLG, False)
+    task9 = Task("flores200_eng-mlt", "chrf", "Flores-200 EN→MT (ChrF)", "https://huggingface.co/datasets/Muennighoff/flores200", TaskType.NLG)
+    task10 = Task("webnlg_mlt", "chrf", "WebNLG (ChrF)", "https://synalp.gitlabpages.inria.fr/webnlg-challenge/challenge_2023/", TaskType.NLG)
+    task11 = Task("webnlg_mlt", "rouge", "WebNLG (Rouge-L)", "https://synalp.gitlabpages.inria.fr/webnlg-challenge/challenge_2023/", TaskType.NLG, False)
+    task12 = Task("eurlexsum_mlt", "chrf", "EUR-Lex-Sum (ChrF)", "https://huggingface.co/datasets/dennlinger/eur-lex-sum", TaskType.NLG, False)
+    task13 = Task("eurlexsum_mlt", "rouge", "EUR-Lex-Sum (Rouge-L)", "https://huggingface.co/datasets/dennlinger/eur-lex-sum", TaskType.NLG)
+    task14 = Task("maltese_news_headlines", "chrf", "Maltese News Headlines (ChrF)", "https://huggingface.co/datasets/MLRS/maltese_news_headlines", TaskType.NLG, False)
+    task15 = Task("maltese_news_headlines", "rouge", "Maltese News Headlines (Rouge-L)", "https://huggingface.co/datasets/MLRS/maltese_news_headlines", TaskType.NLG)
 
 NUM_FEWSHOT = 0 # Change with your few shot
 # ---------------------------------------------------
@@ -49,20 +52,36 @@ NUM_FEWSHOT = 0 # Change with your few shot
 
 
 # Your leaderboard name
-TITLE = """<h1 align="center" id="space-title">🇲🇹 MELABench Leaderboard</h1>"""
+TITLE = """
+<h1 align="center" id="space-title">
+<img src="https://raw.githubusercontent.com/MLRS/MELABench/refs/heads/main/logo.jpg" alt="MELABench logo" width="200px">
+Leaderboard
+</h1>
+"""
 
 # What does your leaderboard evaluate?
 INTRODUCTION_TEXT = """
-A Maltese Evaluation Language Benchmark
+<p align="center">A Maltese Evaluation Language Benchmark 🇲🇹</p>
 """
 
 # Which evaluations are you running? how can people reproduce what you have?
+tasks = {task_type.value.display_name: {} for task_type in TaskType}
+for task in Tasks:
+    tasks[task.value.task_type.value.display_name][re.sub(r" \(.*\)$", "", task.value.col_name)] = task.value.url
 LLM_BENCHMARKS_TEXT = f"""
-## How it works
-
-## Reproducibility
-To reproduce our results, here is the commands you can run:
+MELABench evaluates language model capabilities on Maltese.
+Currently, the following tasks are supported:
+""" + \
+"\n".join([
+    f"- {task_type}:\n" + "\n".join(f"  - [{task}]({url})" for task, url in sub_tasks.items()) + "\n"
+    for task_type, sub_tasks in tasks.items()
+]) + \
+"""
+The leaderboard is developed and maintained by people managing [MLRS](https://mlrs.research.um.edu.mt/).
+We plan to expand our initial work with more tasks, if you would like to contribute your data, please reach out!
+If you would like to include results for models/setups we did not include, we also accept submissions.
 
+This work was introduced in [MELABenchv1: Benchmarking Large Language Models against Smaller Fine-Tuned Models for Low-Resource Maltese NLP](https://arxiv.org/abs/2506.04385).
 """
 
 EVALUATION_QUEUE_TEXT = """
@@ -73,4 +92,22 @@ In addition, we ask for additional metadata about model training.
 
 CITATION_BUTTON_LABEL = "Copy the following snippet to cite these results"
 CITATION_BUTTON_TEXT = r"""
+@inproceedings{micallef-borg-2025-melabenchv1,
+    title = "{MELAB}enchv1: Benchmarking Large Language Models against Smaller Fine-Tuned Models for Low-Resource {M}altese {NLP}",
+    author = "Micallef, Kurt  and
+      Borg, Claudia",
+    editor = "Che, Wanxiang  and
+      Nabende, Joyce  and
+      Shutova, Ekaterina  and
+      Pilehvar, Mohammad Taher",
+    booktitle = "Findings of the Association for Computational Linguistics: ACL 2025",
+    month = jul,
+    year = "2025",
+    address = "Vienna, Austria",
+    publisher = "Association for Computational Linguistics",
+    url = "https://aclanthology.org/2025.findings-acl.1053/",
+    doi = "10.18653/v1/2025.findings-acl.1053",
+    pages = "20505--20527",
+    ISBN = "979-8-89176-256-5",
+}
 """