Throughout this paper we present our system developed to solve SemEval-2025 Task 11: Bridging the Gap in Text-based Emotion Detection Track A. To participate in this contest, we use an architecture based on a pretrained encoder model as the shared part of the model and then add specific head to adapt the shared part for each language. In the first part of this report, we will introduce the task and the specific track in which we participated and then elaborate on the dataset and the system we developed to handle the task. Finally, we will analyze our results and discuss limitations and potential strength point of our solution that could be leveraged in future work to improve results on similar tasks.

This paper presents our WordWiz system for SemEval-2025 Task 10: Narrative Extraction. We employed a combination of targeted preprocessing techniques and instruction-tuned language models to generate concise, accurate narrative explanations across five languages. Our approach leverages an evidence refinement strategy that removes irrelevant sentences, improving signal-to-noise ratio in training examples. We fine-tuned Microsoft’s Phi-3.5 model using both Supervised Fine-Tuning (SFT). During inference, we implemented a multi-temperature sampling strategy that generates multiple candidate explanations and selects the optimal response using narrative relevance scoring. Notably, our smaller Phi-3.5 model consistently outperformed larger alternatives like Llama-3.1-8B across most languages. Our system achieved significant improvements over the baseline across all languages, with F1 scores ranging from 0.7486 (Portuguese) to 0.6839 (Bulgarian), demonstrating the effectiveness of evidence-guided instruction tuning for narrative extraction.

Recent advancements in natural language processing (NLP) have prompted the development of sophisticated reasoning benchmarks. This paper presents our system for the SemEval 2024 Task 9 competition and also investigates the efficacy of fine-tuning language models (LMs) on BrainTeaser—a benchmark designed to evaluate NLP models’ lateral thinking and creative reasoning abilities. Our experiments focus on two prominent families of pre-trained models, BERT and T5. Additionally, we explore the potential benefits of multi-task fine-tuning on commonsense reasoning datasets to enhance performance. Our top-performing model, DeBERTa-v3-large, achieves an impressive overall accuracy of 93.33%, surpassing human performance.

Based on recent progress in image-text retrieval techniques, this paper presents a fine-tuned model for the Visual Word Sense Disambiguation (VWSD) task. The proposed system fine-tunes a pre-trained model using ITC and ITM losses and employs a candidate selection approach for faster inference. The system was trained on the VWSD task dataset and evaluated on a separate test set using Mean Reciprocal Rank (MRR) metric. Additionally, the system was tested on the provided test set which contained Persian and Italian languages, and the results were evaluated on each language separately. Our proposed system demonstrates the potential of fine-tuning pre-trained models for complex language tasks and provides insights for further research in the field of image text retrieval.

Depression is a common and serious mental illness that early detection can improve the patient’s symptoms and make depression easier to treat. This paper mainly introduces the relevant content of the task “Detecting Signs of Depression from Social Media Text at DepSign-LT-EDI@ACL-2022”. The goal of DepSign is to classify the signs of depression into three labels namely “not depressed”, “moderately depressed”, and “severely depressed” based on social media’s posts. In this paper, we propose a predictive ensemble model that utilizes the fine-tuned contextualized word embedding, ALBERT, DistilBERT, RoBERTa, and BERT base model. We show that our model outperforms the baseline models in all considered metrics and achieves an F1 score of 54% and accuracy of 61%, ranking 5th on the leader-board for the DepSign task.

In this paper, we describe our proposed methods for the multilingual word-in-Context disambiguation task in SemEval-2021. In this task, systems should determine whether a word that occurs in two different sentences is used with the same meaning or not. We proposed several methods using a pre-trained BERT model. In two of them, we paraphrased sentences and add them as input to the BERT, and in one of them, we used WordNet to add some extra lexical information. We evaluated our proposed methods on test data in SemEval- 2021 task 2.