Austria

This paper describes the approach of the
Unibuc - NLP team in tackling the SemEval
2025 Workshop, Task 11: Bridging the Gap
in Text-Based Emotion Detection. We mainly
focused on experiments using large language
models (Gemini, Qwen, DeepSeek) with ei-
ther few-shot prompting or fine-tuning. With
our final system, for the multi-label emotion
detection track (track A), we got an F1-macro
of 0.7546 (26/96 teams) for the English sub-
set, 0.1727 (35/36 teams) for the Portuguese
(Mozambican) subset and 0.325 (1/31 teams)
for the Emakhuwa subset.

ACL 2025

Team Unibuc - NLP at SemEval 2025 Task 11: Few-shot text-based emotion detection

### Welcome to The 63rd Annual Meeting of the Association for Computational Linguistics (ACL 2025)

Message from the General Chair: 
*It is my great pleasure and honor to welcome you to the 63rd Annual Meeting of the Association for Computational Linguistics (ACL 2025), held in beautiful Vienna, Austria, from July 27 to August 1, 2025. ACL2025continues our field’s tradition of excellence in scholarship, innovation, and inclusivity, and I am deeply grateful to the many volunteers who have worked tirelessly to bring this event to life.* 
[Read more](https://drive.google.com/file/d/1GI_hvOpjswAuYdUTromfeDiPpCcqidwg/view?usp=sharing)

To access this event page, you need to log in with the **email address you registered with**. Access credentials will be sent to your email from Underline - subject line "Welcome to ACL 2025". Please be sure to check your spam email folder if you do not see an email confirmation right away.

Please log in

To access this event page, you need to log in with the **email address you registered with**. 

Welcome to The 63rd Annual Meeting of the Association for Computational Linguistics (ACL 2025)

workshop paper

To address the growing need for ESG verification, we propose an
NLP framework combining RAG and Structured Prompts for
multilingual ESG analysis. In SemEval-2025, our system ranked
4th in English (0.5611) and 1st in French (0.5747), outperforming
traditional methods and proving its scalability and reliability.

CYUT at SemEval-2025 Task 6: Prompting with Precision â€“ ESG Analysis via Structured Prompts

We showcase our approach for multilingual fact-checked claim retrieval, developed for SemEval 2025 Task 7. By adapting large language models with contrastive fine-tuning and techniques like LoRA and Prompt Tuning, we achieved strong results in both monolingual and crosslingual tracks. In this talk, we’ll share key insights, limitations, and future directions.

RACAI at SemEval-2025 Task 7: Efficient adaptation of Large Language Models for Multilingual and Crosslingual Fact-Checked Claim Retrieval

This paper presents the ZJUKLAB team's submission for SemEval-2025 Task 4: Unlearning Sensitive Content from Large Language Models. This task aims to selectively erase sensitive knowledge from large language models, avoiding both over-forgetting and under-forgetting issues. We propose an unlearning system that leverages Model Merging (specifically TIES-Merging), combining two specialized models into a more balanced unlearned model. Our system achieves competitive results, ranking second among 26 teams, with an online score of 0.944 for Task Aggregate and 0.487 for overall Aggregate. In this paper, we also conduct local experiments and perform a comprehensive analysis of the unlearning process, examining performance trajectories, loss dynamics, and weight perspectives, along with several supplementary experiments, to understand the effectiveness of our method. Furthermore, we analyze the shortcomings of our method and evaluation metrics, emphasizing that MIA scores and ROUGE-based metrics alone are insufficient to fully evaluate successful unlearning. Finally, we emphasize the need for more comprehensive evaluation methodologies and rethinking of unlearning objectives in future research.

ZJUKLAB at SemEval-2025 Task 4: Unlearning via Model Merging.

We tackle the challenge of multi-label emotion detection in short texts, focusing on SemEval2025 Task 11 Track A. Our approach, RoEmo, combines generative and discriminative models in an ensemble strategy to classify texts into five emotions: anger, fear, joy, sadness, and surprise. The generative model, instructionfinetuned on emotion detection datasets, undergoes additional fine-tuning on the SemEval2025 Task 11 Track A dataset to enhance its performance for this specific task. Meanwhile, the discriminative model, based on binary classification, offers a straightforward yet effective approach to classification. We review recent advancements in multi-label emotion detection and analyze the task dataset. Our results show that RoEmo ranks among the top-performing systems, demonstrating high accuracy and reliability.

MRS at SemEval-2025 Task 11: A Hybrid Approach for Bridging the Gap in Text-Based Emotion Detection

This work describes our system submitted to SemEval-2025 Task 8 “Question Answering over Tabular Data.” The shared task focuses on tackling real-life table question answering (TQA) involving extremely large tables with the additional challenges of interpreting complex questions. To address these issues, we leverage a framework of Multi-Agent Collaboration with Tool use (MACT), a method that combines planning and tool use. The planning module breaks down a complex question by designing a step-by-step plan. This plan is translated into Python code by a coding model, and a Python interpreter executes the code to generate an answer. Our system demonstrates competitive performance in the shared task and is ranked 5th out of 38 in the open-source model category. We provide a detailed analysis of our model, evaluating the effectiveness and the efficiency of each component, and identify common error patterns. Our work offers essential insights and recommendations for future advancements in developing TQA systems.

G-MACT at SemEval-2025 Task 8: Exploring Planning and Tool Use in Question Answering over Tabular Data

The Food Hazard Detection (SemEval-2025 Task 9) advances explainable classification of food-incident reports collected from web sources, including social media and regulatory agency websites, to support timely risk mitigation for public health and the economy. This task is complicated by a highly imbalanced, long-tail label distribution and the need for transparent, reliable AI. We present a robust Knowledge-Augmented Data approach that integrates Retrieval-Augmented Generation (RAG) with domain-specific knowledge from the PubMed API to enrich and balance the training data. Our method leverages domain-specific knowledge to expand datasets and curate high-quality data that enhances overall data integrity. We hypothesize that Knowledge-Augmented Data improves Macro-F1 scores, the primary evaluation metric. Our approach achieved a top-2 ranking across both subtasks, demonstrating its effectiveness in advancing NLP applications for food safety and contributing to more reliable food hazard detection.

MyMy at SemEval-2025 Task 9: A Robust Knowledge-Augmented Data Approach for Reliable Food Hazard Detection

This paper presents our system developed for SemEval-2025 Task 7, which focuses on retrieving verified fact-checked claims related to social media posts, both in monolingual and crosslingual settings.
To address challenges such as data noise, language resource imbalance, and translation inconsistency, we design a three-stage retrieval framework:
Retrieval: Select candidate claims using strong multilingual models.
Re-ranking: Use fine-tuned rerankers to improve relevance.
Voting: Combine multiple model outputs for robust final predictions.
Our system achieves competitive performance, ranking 5th in the monolingual and 7th in the crosslingual track. The results highlight the effectiveness of ensemble strategies and multilingual model selection in combating misinformation.

QUST_NLP at SemEval-2025 Task 7: A Three-Stage Retrieval Framework for Monolingual and Crosslingual Fact-Checked Claim Retrieval

We present SALT, a parameter-efficient approach to entity-aware machine translation that challenges the trend toward ever-larger language models. Our system combines simple SQL-based entity retrieval with a 600M parameter NLLB model, enhanced through logit biasing for knowledge integration. Despite its simplicity, SALT achieves first place in SemEval-2025 Task 2 among systems not using gold-standard data, outperforming approaches with over 100 times more parameters.
Through extensive ablations, we demonstrate that straightforward string matching rivals complex neural retrievers, single-entity candidates outperform multiple candidates, and targeted knowledge integration via logit biasing can match the performance of multi-billion-parameter LLMs. Our findings suggest that for specialized translation tasks with structured knowledge, clever integration strategies can be more effective than raw parameter scaling, achieving competitive results while remaining computationally efficient enough to run on consumer hardware.

SALT at SemEval-2025 Task 2: A SQL-based Approach for LLM-Free Entity-Aware-Translation

This paper presents Trans-Sent, our system developed for Track A: Multi-label Emotion Detection of SemEval-2025 Task 11: Bridging the Gap in Text-based Emotion Detection. The aim of this task is to predict the emotions that a speaker may convey i.e. perceived emotions. A target text snippet is given and our goal is to label the text by 1 and 0 as ‘yes’ and ‘no’ for different emotional classes like joy, sadness, fear, anger, surprise and/or disgust. Trans-Sent is a Transformer-based sentiment extraction model. It uses various pre-trained Bidirectional Encoder Representations from Transformers or BERT models for individual language to generate emotions and classify the emotion-labels. We have participated in the task for 7 different languages and achieve competitive results. Our system gives best result for Russian language, as it is ranked ninth among all ranked teams.

Trans-Sent at SemEval-2025 Task 11: Text-based Multi-label Emotion Detection using Pre-Trained BERT Transformer Models

We develop a system in the form of a pipeline consisting of two parts: 
Candidate Recall Module, which selects the most probable correct labels from numerous of labels based on the BERT model
LLM Prediction Module, which is used to generate the final prediction based on Large Language Models(LLM) with External Attention Module
Furthermore, with the help of data augmentation we achieve a good result


Downloads

Next from ACL 2025

CYUT at SemEval-2025 Task 6: Prompting with Precision â€“ ESG Analysis via Structured Prompts

Similar lecture

Trans-Sent at SemEval-2025 Task 11: Text-based Multi-label Emotion Detection using Pre-Trained BERT Transformer Models

.css-70qvj9{display:-webkit-box;display:-webkit-flex;display:-ms-flexbox;display:flex;-webkit-align-items:center;-webkit-box-align:center;-ms-flex-align:center;align-items:center;}Downloads

Next from ACL 2025

CYUT at SemEval-2025 Task 6: Prompting with Precision â€“ ESG Analysis via Structured Prompts

Similar lecture

Trans-Sent at SemEval-2025 Task 11: Text-based Multi-label Emotion Detection using Pre-Trained BERT Transformer Models

Downloads