Update README.md

README.md

@@ -57,7 +57,7 @@ size_categories:
 We introduce a large semi-automatically generated dataset of ~400,000 descriptive sentences about commonsense knowledge that can be true or false in which negation is present in about 2/3 of the corpus in different forms that we use to evaluate LLMs.
 </p>
 
-- 📖 Paper: [This is not a Dataset: A Large Negation Benchmark to Challenge Large Language Models (EMNLP'23)]()
+- 📖 Paper: [This is not a Dataset: A Large Negation Benchmark to Challenge Large Language Models (EMNLP'23)](http://arxiv.org/abs/2310.15941)
 - 💻 Baseline Code and the Official Scorer: [https://github.com/hitz-zentroa/This-is-not-a-Dataset](https://github.com/hitz-zentroa/This-is-not-a-Dataset)
 
 # Data explanation