Critical Question Generation (CQs-Gen) improves reasoning and critical thinking skills through Critical Questions (CQs), which identify reasoning gaps and address misinformation in NLP, especially as LLM-based chat systems are widely used for learning and may encourage superficial learning habits. The Shared Task on Critical Question Generation, hosted at the 12th Workshop on Argument Mining and co-located in ACL 2025, has aimed to address these challenges. This study proposes a CQs-Gen pipeline using Llama-3-8B-Instruct-GGUF-Q8_0 with few-shot learning, integrating text simplification, NER, and argument schemes to enhance question quality. Through an extensive experiment testing without training, fine-tuning with PEFT using LoRA on 10% of the dataset, and few-shot fine-tuning (using five examples) with an 8-bit quantized model, we demonstrate that the few-shot approach outperforms others. On the validation set, 397 out of 558 generated CQs were classified as Useful, representing 71.1% of the total. In contrast, on the test set, 49 out of 102 generated CQs, accounting for 48% of the total, were classified as Useful following evaluation through semantic similarity and manual assessments.

To encourage computational argumentation through critical question generation (CQs-Gen),we propose an ACL 2025 CQs-Gen shared task system to generate critical questions (CQs) with the best effort to counter argumentative text by discovering logical fallacies, unjustified assertions, and implicit assumptions.Our system integrates a quantized language model, semantic similarity analysis, and a meta-evaluation feedback mechanism including the key stages such as data preprocessing, rationale-augmented prompting to induce specificity, diversity filtering for redundancy elimination, enriched meta-evaluation for relevance, and a feedback-reflect-refine loop for iterative refinement. Multi-metric scoring guarantees high-quality CQs. With robust error handling, our pipeline ranked 7th among 15 teams, outperforming baseline fact-checking approaches by enabling critical engagement and successfully detecting argumentative fallacies. This study presents an adaptive, scalable method that advances argument mining and critical discourse analysis.

Misogynous content on social media, especially in memes, present challenges due to the complex reciprocation of text and images that carry offensive messages. This difficulty mostly arises from the lack of direct alignment between modalities and biases in large-scale visio-linguistic models. In this paper, we present our system for the Shared Task on Misogyny Meme Detection - DravidianLangTech@NAACL 2025. We have implemented various unimodal models, such as mBERT and IndicBERT for text data, and ViT, ResNet, and EfficientNet for image data. Moreover, we have tried combining these models and finally adopted a multimodal approach that combined mBERT for text and EfficientNet for image features, both fine-tuned to better interpret subtle language and detailed visuals. The fused features are processed through a dense neural network for classification. Our approach achieved an F1 score of 0.78120, securing 4th place and demonstrating the potential of transformer-based architectures and state-of-the-art CNNs for this task.

Ensuring a safe and inclusive online environment requires effective hate speech detection on social media. While detection systems have significantly advanced for English, many regional languages, including Malayalam, Tamil and Telugu, remain underrepresented, creating challenges in identifying harmful content accurately. These languages present unique challenges due to their complex grammar, diverse dialects, and frequent code-mixing with English. The rise of multimodal content, including text and audio, adds further complexity to detection tasks. The shared task “Multimodal Hate Speech Detection in Dravidian Languages: DravidianLangTech@NAACL 2025” has aimed to address these challenges. A Youtube-sourced dataset has been provided, labeled into five categories: Gender (G), Political (P), Religious (R), Personal Defamation (C) and Non-Hate (NH). In our approach, we have used mBERT, T5 for text and Wav2Vec2 and Whisper for audio. T5 has performed poorly compared to mBERT, which has achieved the highest F1 scores on the test dataset. For audio, Wav2Vec2 has been chosen over Whisper because it processes raw audio effectively using self-supervised learning. In the hate speech detection task, we have achieved a macro F1 score of 0.2005 for Malayalam, ranking 15th in this task, 0.1356 for Tamil and 0.1465 for Telugu, with both ranking 16th in this task.

Artificial Intelligence (AI) is opening new doors of learning and interaction. However, it has its share of problems. One major issue is the ability of AI to generate text that resembles human-written text. So, how can we tell apart human-written text from AI-generated text?With this in mind, we have worked on detecting AI-generated product reviews in Dravidian languages, mainly in Malayalam and Tamil. The “Shared Task on Detecting AI-Generated Product Reviews in Dravidian Languages,” held as part of the DravidianLangTech Workshop at NAACL 2025 has provided a dataset categorized into two categories, human-written review and AI-generated review. We have implemented four machine learning models (Random Forest, Support Vector Machine, Decision Tree, and XGBoost), four deep learning models (Long Short-Term Memory, Bidirectional Long Short-Term Memory, Gated Recurrent Unit, and Recurrent Neural Network), and three transformer-based models (AI-Human-Detector, Detect-AI-Text, and E5-Small-Lora-AI-Generated-Detector). We have conducted a comparative study among all the models by training and evaluating each model on the dataset. We have discovered that the transformer, E5-Small-Lora-AI-Generated-Detector, has provided the best result with an F1 score of 0.8994 on the test set ranking 7th position in the Malayalam language. Tamil has a higher token overlap and richer morphology than Malayalam. Thus, we obtained a worse F1 score of 0.5877 ranking 28th position in the Tamil language among all participants in the shared task.

The advancement of high-speed internet and affordable bandwidth has led to a significant increase in video content and has brought challenges in content moderation due to the spread of unsafe or harmful narratives quickly. The rise of short-form videos like “Reels”, which is easy to create and consume, has intensified these challenges even more. In case of Bengali culture-specific content, the existing content moderation system struggles. To tackle these challenges within the culture-specific Bengali codemixed domain, this paper introduces “UNBER” a novel dataset of 1,111 multimodal Bengali codemixed Facebook Reels categorized into four classes: Safe, Adult, Harmful, and Suicidal. Our contribution also involves the development of a unique annotation tool “ReelAn” to enable an efficient annotation process of reels. While many existing content moderation techniques have focused on resource-rich or monolingual languages, approaches for multimodal datasets in Bengali are rare. To fill this gap, we propose a culturally aware cross-modal attention-based fusion framework to enhance the analysis of these fast-paced videos, which achieved a macro F1 score of 0.75. Our contributions aim to significantly advance multimodal content moderation and lay the groundwork for future research in this area.

Misogynistic memes spread harmful stereotypes and toxic content across social media platforms, often combining sarcastic text and offensive visuals that make them difficult to detect using traditional methods. Our research has been part of the the Shared Task on Misogyny Meme Detection - LT- EDI@LDK 2025, identifying misogynistic memes using deep learning-based multimodal approach that leverages both textual and visual information for accurate classification of such memes. We experiment with various models including CharBERT, BiLSTM, and CLIP for text and image encoding, and explore fusion strategies like early and gated fusion. Our best performing model, CharBERT + BiLSTM + CLIP with gated fusion, achieves strong results, showing the effectiveness of combining features from both modalities. To address challenges like language mixing and class imbalance, we apply preprocessing techniques (e.g., Romanizing Chinese text) and data augmentation (e.g., image transformations, text back-translation). The results demonstrate significant improvements over unimodal baselines, highlighting the value of multimodal learning in detecting subtle and harmful content online.

This paper presents our system for the LT-EDI 2025 Shared Task on Racial Hoax Detection, addressing the critical challenge of identifying racially charged misinformation in code-mixed Hindi-English (Hinglish) social media—a low-resource, linguistically complex domain with real-world impact. We adopt a two-pronged strategy, independently fine-tuning a transformer-based model and a large language model. CharBERT was optimized using Optuna, while XLM-RoBERTa and DistilBERT were fine-tuned for the classification task. FLAN-T5-base was fine-tuned with SMOTE-based oversampling, semantic-preserving back translation, and prompt engineering, whereas LLaMA was used solely for inference. Our preprocessing included Hinglish-specific normalization, noise reduction, sentiment-aware corrections and a custom weighted loss to emphasize the minority Hoax class. Despite using FLAN-T5-base due to resource limits, our models performed well. CharBERT achieved a macro F1 of 0.70 and FLAN-T5 followed at 0.69, both outperforming baselines like DistilBERT and LLaMA-3.2-1B. Our submission ranked 4th of 11 teams, underscoring the promise of our approach for scalable misinformation detection in code-switched contexts. Future work will explore larger LLMs, adversarial training and context-aware decoding.

The proliferation of caste- and migration-based hate speech on social media poses a significant challenge, particularly in low-resource languages like Tamil. This paper presents our approach to the LT-EDI@ACL 2025 shared task, addressing this issue through a hybrid transformer-based framework. We explore a range of Machine Learning (ML), Deep Learning (DL), and multilingual transformer models, culminating in a novel m-BERT+BiLSTM hybrid architecture. This model integrates contextual embeddings from m-BERT with lexical features from TF-IDF and FastText, feeding the enriched representations into a BiLSTM to capture bidirectional semantic dependencies. Empirical results demonstrate the superiority of this hybrid architecture, achieving a macro-F1 score of 0.76 on the test set and surpassing the performance of standalone models such as MuRIL and IndicBERT. These results affirm the effectiveness of hybrid multilingual models for hate speech detection in low-resource and culturally complex linguistic settings.

In recent days, propaganda has started to influence public opinion increasingly as social media usage continues to grow. Our research has been part of the first challenge, Unimodal (Text) Propagandistic Technique Detection of ArAIEval shared task at the ArabicNLP 2024 conference, co-located with ACL 2024, identifying specific Arabic text spans using twenty-three propaganda techniques. We have augmented underrepresented techniques in the provided dataset using synonym replacement and have evaluated various machine learning (RF, SVM, MNB), deep learning (BiLSTM), and transformer-based models (bert-base-arabic, Marefa-NER, AraBERT) with transfer learning. Our comparative study has shown that the transformer model “bert-base-arabic” has outperformed other models. Evaluating the test set, it has achieved the micro-F1 score of 0.2995 which is the highest. This result has secured our team “CUET_sstm” first place among all participants in task 1 of the ArAIEval.

Swahili, spoken by around 200 million people primarily in Tanzania and Kenya, has been the focus of our research for the GEM Shared Task at INLG’24 on Underrepresented Language Summarization. We have utilized the XLSUM dataset and have manually summarized 1000 texts from a Swahili news classification dataset. To achieve the desired results, we have tested abstractive summarizers (mT5_multilingual_XLSum, t5-small, mBART-50), and an extractive summarizer (based on PageRank algorithm). But our adopted model consists of an integrated extractive-abstractive model combining the Bert Extractive Summarizer with some abstractive summarizers (t5-small, mBART-50). The integrated model overcome the drawbacks of both the extractive and abstractive summarization system and utilizes the benefit from both of it. Extractive summarizer shorten the paragraphs exceeding 512 tokens, ensuring no important information has been lost before applying the abstractive models. The abstractive summarizer use its pretrained knowledge and ensure to generate context based summary.