Is this you? Claim this profile to correct it, add a bio and choose the work people see first.
Claim this profileAcademic lineage
Possible advisorsa guess from early papers, not confirmed
- Jinman KimSuggested from co-authorship
- Adam G. DunnSuggested from co-authorship
Is this you? Claim this profile to confirm or dismiss it.
Works279 from public data
- COVIDSenti: A Large-Scale Benchmark Twitter Data Set for COVID-19 Sentiment Analysis345
This study presents a new large-scale sentiment data set COVIDSENTI, which consists of 90 000 COVID-19-related tweets collected in the early stages of the pandemic, from February to March 2020 and supports the view that there is a need to develop a proactive and agile public health presence to combat the spread of negative sentiment on social media following a pandemic.
- Transformer based Deep Intelligent Contextual Embedding for Twitter sentiment analysis298
This paper presents D I C E T, a transformer-based method for sentiment analysis that encodes representation from a transformer and applies deep intelligent contextual embedding to enhance the quality of tweets by removing noise while taking word sentiments, polysemy, syntax, and semantic knowledge into account.
- A Comprehensive Survey on Word Representation Models: From Classical to State-of-the-Art Word Representation Language Models203
A variety of text representation methods, and model designs have blossomed in the context of NLP, including SOTA LMs are described, which can transform large volumes of text into effective vector representations capturing the same semantic information.
- Conventional to online education during COVID-19 pandemic: Do develop and underdeveloped nations cope alike173
This study compared access & use of online learning among Bruneians and Pakistanis and found Bruneian are more satisfied with online learning as compared to Pakistanis.
- An Automatic Detection of Breast Cancer Diagnosis and Prognosis Based on Machine Learning Using Ensemble of Classifiers129
The results showed that the ensemble method outperformed other state-of-the-art methods and achieved 98.83% accuracy, which is of great importance to the medical industry and relevant research community.
- A survey of pre-processing techniques to improve short-text quality: a case study on hate speech detection on twitter120
This paper analyzes twelve different pre- processing techniques on three pre-classified Twitter datasets on hate speech and proposes a systematic approach to text pre-processing to apply differentPre-processing techniques in order to retain features without information loss.
- Early Identification of Depression Severity Levels on Reddit Using Ordinal Classification107
This paper reformulates the early identification of depression as an ordinal classification task over clinical depression standards such as Beck’s Depression Inventory and the Depressive Disorder Annotation scheme to identify depression severity levels and proposes a hierarchical attention method optimized to factor in the increasing depression severity Levels through a soft probability distribution.
- Breast cancer risk prediction using machine learning: a systematic review90
This review provides a comprehensive guide for understanding the current status of breast cancer risk assessment using AI, and offers investigators a different perspective on the use of AI for breast cancer risk prediction, incorporating numerous imaging and non-imaging features.
- A Recommendation Engine for Predicting Movie Ratings Using a Big Data Approach88
This study implemented a movie recommendation system based on a collaborative filtering approach using the alternating least squared (ALS) model to predict the best-rated movies and showed an accuracy of 97% and predicted the top 1000 ratings for movies.
- Large language models (LLM) in computational social science: prospects, current state, and challenges84
A solid background on the potential of LLMs in CSS, their past applications, current problems, and how they can pave the way for revolutionizing CSS is provided.
- BioALBERT: A Simple and Effective Pre-trained Language Model for Biomedical Named Entity Recognition82
The proposed biomedical ALBERT (A Lite Bidirectional Encoder Representations from Transformers for Biomedical Text Mining) is an effective domain-specific pre-trained language model trained on a huge biomedical corpus designed to capture biomedical context-dependent NER.
- Prompting Large Language Models for Topic Modeling72
This paper proposes PromptTopic, a novel topic modeling approach that harnesses the advanced language understanding of large language models (LLMs) to address challenges of short text datasets that lack co-occurring words.
- The role of unmanned aerial vehicles and mmWave in 5G: Recent advances and challenges70
A consolidated synthesis on the role of UAVs and mmWave in 5G, emphasis on recent developments and challenges, and the scope of artificial intelligence and machine learning techniques as an efficient solution for combating the dynamic and complex nature of U AV‐based cellular communication networks are discussed.
- Adaption of distance learning to continue the academic year amid COVID-19 lockdown69
The findings indicate that trust in information sources such as institute and media information or interpersonal communication related to distance learning programs is correlated with awareness and readiness, and that readiness strongly influences the adoption of distance learning amid the COVID-19 pandemic.
- Benchmarking for biomedical natural language processing tasks with a domain specific ALBERT67
8 variants of BioALberT, a domain-specific adaptation of a lite bidirectional encoder representations from transformers (ALBERT), trained on biomedical and clinical corpora and fine-tuned for 6 different tasks across 20 benchmark datasets are presented.
- A Comparative Analysis of Active Learning for Biomedical Text Mining64
Experiments show that AL has the potential to significantly reducing the cost of manual labelling, and pre-labelling performed using AL expediates the labelling process by reducing the time required for labelling.
- Enhancing textual textbook question answering with large language models and retrieval augmented generation53
A framework (PLRTQA) is proposed that incorporates the retrieval augmented generation (RAG) technique to handle the out-of-domain scenario where concepts are spread across different lessons, and utilize transfer learning to handle the long context and enhance reasoning abilities.
- MultiHateClip: A Multilingual Benchmark Dataset for Hateful Video Detection on YouTube and Bilibili53
Evaluations of state-of-the-art video classification models, such as VLM, GPT-4V and Qwen-VL, on MultiHateClip highlight the existing challenges in accurately distinguishing between hateful and offensive content and the urgent need for models that are both multimodally and culturally nuanced.
- Performance Evaluation of Next-Generation Wireless (5G) UAV Relay53
This study investigates, analyzes and describes the distinctive rich characteristics of mmWave propagation in Access and backhaul network simultaneously using UAV, and highlights the impact of UAV location to maximize the performance of an Amplify-and-Forward UAV based relay for providing enhanced coverage to the users.
- MoCFL: Mobile Cluster Federated Learning Framework for Highly Dynamic Network52
MoCFL enhances feature aggregation by introducing an affinity matrix that quantifies the similarity between local feature extractors from different clients, addressing dynamic data distribution changes caused by frequent client churn and topology changes.
- 52
- Navigating the Web of Disinformation and Misinformation: Large Language Models as Double-Edged Swords51
The dual role of Large Language Models in the context of online misinformation and disinformation and their potential to generate and counter misinformation and disinformation in sensitive topics such as healthcare, COVID-19, and political agendas are explored.
- Identifying Creative Harmful Memes via Prompt based Approach48
This paper reformulates a harmful meme analysis as an auto-filling and presents a prompt-based approach to identify harmful memes, which outperformed state-of-the-art methods on two benchmark harmful memes datasets.
- 47
- A Multi-Criteria Approach for Arabic Dialect Sentiment Analysis for Online Reviews: Exploiting Optimal Machine Learning Algorithm Selection45
A multi-criteria method to empirically assess and rank classifiers for Arabic sentiment analysis, which suggests that deep learning and support vector machine (SVM) classifiers perform best and outperform decision trees, K-nearest neighbours, and Naïve Bayes classifiers.
- Hybrid Words Representation for Airlines Sentiment Analysis45
This paper presents hybrid words representation and Bi-directional Long Short Term Memory with attention modeling resulting in improvement in tweet quality by not only treating the noise within the textual context but also considers polysemy, semantics, syntax, out of vocabulary (OOV) words as well as words sentiments within a tweet.
- A Multi-Modal Dataset for Hate Speech Detection on Social Media: Case-study of Russia-Ukraine Conflict44
Experimental analysis of the presented dataset has shown that understanding both modalities is essential for detecting hate speech detection techniques, and this work plans to develop new multi-modal models tailored explicitly to hate-speech detection.
- 43
- Evaluating Personality Traits in Large Language Models: Insights from Psychological Questionnaires42
This work applies psychological tools to LLMs in diverse scenarios to generate personality profiles and reveals that LLMs exhibit unique dominant traits, varying characteristics, and distinct personality profiles even within the same family of models.
- CHUNAV: Analyzing Hindi Hate Speech and Targeted Groups in Indian Election Discourse41
The CHUNAV dataset—a collection of 11,457 Hindi tweets gathered during assembly elections in various states is introduced—a collection of hate speech categorization and the identification of target groups aimed at effectively addressing hate speech and target identification in the Hindi language.
- 39
- Automatic COVID-19 Lung Infection Segmentation through Modified Unet Model39
This research work proposes a segmentation approach to identifying ground glass opacity or ROI in CT images developed by coronavirus, with a modified structure of the Unet model having been used to classify the region of interest at the pixel level.
- Detecting Alzheimer’s Disease by Exploiting Linguistic Information from Nepali Transcript38
The proposed study makes a convincing conclusion that the difficulty in processing information in AD patients reflects in their speech while describing a picture, and sets a baseline for the early detection of AD using NLP in the Nepali language.
- Uncovering Political Hate Speech During Indian Election Campaign: A New Low-Resource Dataset and Baselines37
A new dataset named IEHate is introduced, which contains 11,457 manually annotated Hindi tweets related to the Indian Assembly Election Campaign from November 1, 2021, to March 9, 2022, and performs a detailed analysis of the dataset, focusing on the prevalence of hate speech in political communication and the different forms of hateful language used.
- 37
- Colouring of COVID-19 Affected Region Based on Fuzzy Directed Graphs37
The colouring of directed fuzzy graphs is defined based on the influence of relationship, the chromatic number and strong Chromatic number are provided, and related properties are investigated An application regarding COVID-19 infection is presented using the colouring.
- 36
- 36
- Towards Improved Deep Contextual Embedding for the identification of Irony and Sarcasm36
This paper presents the identification of irony and sarcasm in social media posts through transformer-based deep, intelligent contextual embedding – T-DICE – which improves noise within contexts and outperforms existing state-of-the-art methods.
- MSynFD: Multi-hop Syntax Aware Fake News Detection35
A novel multi-hop syntax aware fake news detection method, which incorporates complementary syntax information to deal with subtle twists in fake news, and introduces a syntactical dependency graph and a multi-hop subgraph aggregation mechanism to capture multi-hop syntax.
- Explainable hybrid word representations for sentiment analysis of financial news34
An explainable hybrid word representation that first augments the data to address the class imbalance issue and then integrates three embeddings to involve polysemy in context, semantics, and syntax in a context is presented.
- Deep Context-Aware Embedding for Abusive and Hate Speech detection on Twitter.34
Deep Context-Aware Embedding is presented for the detection of Hate speech and abusive language on twitter by considering polsemy, syntax, semantic, OOV words as well as sentiment knowledge and concatenated to form input vector.
- 33
- Large language models vs human for classifying clinical documents32
Combining ChatGPT with methods like SNOMED mapping could further enhance clinical coding accuracy, particularly for complex scenarios, and underscores the potential of integrating advanced language models with clinical coding processes to improve documentation accuracy.
- A Multimodal Framework for the Identification of Vaccine Critical Memes on Twitter31
A new multimodal framework designed to improve the memes' representation by learning the global and local representations of memes, and discusses how understanding memes can be useful in designing shareable vaccination promotion, myth debunking memes and monitoring their uptake on social media platforms.
- Deep AutoEncoder-Decoder Framework for Semantic Segmentation of Brain Tumor.30
This paper performed the semantic segmentation on the MRI brain tumor Figshare - dataset and achieved the state of the arts results (99.93% global accuracy) in comparison to traditional CNN models.
- Deep learning, radiomics and radiogenomics applications in the digital breast tomosynthesis: a systematic review29
A comprehensive guide to understanding the current state of early breast cancer detection using DBT images is provided and investigators with various backgrounds can easily seek interdisciplinary science and new DL, radiomics, and radiogenomics directions towards DBT.
- 29
- Medical Question Summarization with Entity-driven Contrastive Learning28
This article proposes a novel medical question summarization framework based on entity-driven contrastive learning (ECL), which employs medical entities present in frequently asked questions as focuses and devises an effective mechanism to generate hard negative samples.
- 28
- CapAlign: Improving Cross Modal Alignment via Informative Captioning for Harmful Meme Detection27
Experimental findings show that the proposed instructions-based abstracting approach CapAlign produces informative captions and outperforms state-of-the-art methods for detecting harmful memes.
- Multiview Multimodal Feature Fusion for Breast Cancer Classification Using Deep Learning26
This study compiled a multimodal dataset comprising imaging and textual data (combination of clinical and radiological features) and proposed a DL-based multiview multimodal feature fusion (MMFF) strategy for breast cancer classification that utilizes images and tabular data from an in-house dataset.
- 25
- Rumor Detection With Hierarchical Representation on Bipartite Ad Hoc Event Trees24
This article proposes a novel rumor detection model with hierarchical representation on the bipartite ad hoc event trees called BAET, and introduces word embedding and feature encoder for the author and post tree, respectively, and design a root-aware attention module to perform node representation.
- Identification of Disease or Symptom terms in Reddit to Improve Health Mention Classification24
This work presents a Reddit health mention dataset (RHMD), a new dataset of multi-domain Reddit data for the health mention classification (HMC) task, and proposes HMCNET that combines a target keyword (disease or symptom term) identification and user behavior hierarchically to improve HMC.
- Better to Ask in English: Evaluation of Large Language Models on English, Low-resource and Cross-Lingual Settings23
This study extensively investigates LLMs in low-resource language contexts to highlight the improvements required in LLMs and language-specific resources to develop more generally purposed NLP applications.
- 23
- Travel behaviour prediction amid covid-19 underlaying situational awareness theory and health belief model23
It is demonstrated that situational awareness has an effect on the postulates of health-protective measures and plays a key role in the implementation of protective measures such as canceling travel plans to ensure protection.
- A Survey on Progress in LLM Alignment from the Perspective of Reward Design22
This survey provides a structured organization of reward modeling and develops a macro-level taxonomy that characterizes reward mechanisms along complementary dimensions, thereby offering both conceptual clarity and practical guidance for alignment research.
- 21
- Do Large Language Models Speak All Languages Equally? A Comparative Study in Low-Resource Settings20
The findings indicate that GPT-4 consistently outperforms Llama 2 and Gemini, with English consistently demonstrating superior performance across diverse tasks compared to low-resource languages, and natural language inference (NLI) exhibits the highest performance among the evaluated tasks.
- 20
- Unleashing the Potential of Diffusion Models Towards Diversified Sequential Recommendations19
A new diversity-aware guidance learning mechanism is devised to direct the training of DMs to effectively capture users' diversified preferences from their historical interactions, which boosts the recommendation diversity of SRSs.
- Assessing Political Inclination of Bangla Language Models19
This work comprehensively analyze biases present in Bangla language models, specifically focusing on social and economic dimensions, and reveals the inclinations of various LMs, which will provide insights into ethical considerations and limitations associated with deploying Bangla LMs.
- Incorporating Medical Knowledge to Transformer-based Language Models for Medical Dialogue Generation19
This study presents a new method that addresses the challenges of medical dialogue generation by incorporating medical knowledge into transformer-based language models that leverages an external medical knowledge graph and injects triples as domain knowledge into the utterances.
- Abusive Language Detection: A Comprehensive Review18
This survey study contains all the relevant references related to detection of abusive language on social media using NLP and machine learning methods.
- 17
- 16
- Wills Aligner: Multi-Subject Collaborative Brain Visual Decoding15
Wills Aligner begins by aligning the fMRI data from different subjects at the anatomical level, and employs delicate mixture-of-brain-expert adapters and a meta-learning strategy to account for individual fMRI pattern differences, enabling visual decoding for each subject to draw insights from other subjects' data.
- Unveiling Misogyny Memes: A Multimodal Analysis of Modality Effects on Identification15
It is revealed that while both text and image modalities contribute to identifying misogynous memes, text plays a significant role in misogyny identification, while image contributes further in terms of fairness.
- Classifying vaccine sentiment tweets by modelling domain-specific representation and commonsense knowledge into context-aware attentive GRU15
This study presents a novel end-to-end framework consisting of interconnected components that use domain-specific LM trained on vaccine-related tweets and models commonsense knowledge into a bidirectional gated recurrent network (CK-BiGRU) with context-aware attention to classify vaccine sentiment tweets with limited information.
- Survey on LLM Safety: Attacks, Defenses, Alignment, Metrics, and Guardrails14
This survey reviews the end-to-end security and safety pipeline of LLMs, focusing on the interaction between users and model responses, and categorizes the system into five key components: attacks, defenses, safety alignment, metrics and guarding mechanisms.
- Ask, Acquire, Understand: A Multimodal Agent-based Framework for Social Abuse Detection in Memes14
This work presents a multimodal, agent-based framework designed to generate informative visual descriptions of memes by asking insightful questions to improve visual descriptions in zero-shot visual question-answering settings, demonstrating strong generalizability and improved detection of social abuse in memes.
- 14
- Review on enhancing clinical decision support system using machine learning14
An overview of state‐of‐the‐art healthcare CDSS, a generic architecture of computer‐based decision support systems using ML, and the Clinical Decision Support Systems enhanced using Machine Learning (ML) approaches are described are described.
- Show Me The Best Outfit for A Certain Scene: A Scene-aware Fashion Recommender System14
The scene-aware fashion recommender system (SAFRS), which uncovers a hitherto unexplored avenue where scene information is taken into account when constructing the FR model, and which outperforms all baselines for every evaluated metric.
- Mechanistic Interpretability for Large Language Model Alignment: Progress, Challenges, and Future Directions13
This paper surveys recent progress in mechanistic interpretability techniques applied to LLM alignment, examining methods ranging from circuit discovery to feature visualization, activation steering, and causal intervention and proposes future research directions focusing on automated interpretability, cross-model generalization of circuits, and the development of interpretability-driven alignment techniques that can scale to frontier models.
- Can Large Language Models Enhance Predictions of Disease Progression? Investigating Through Disease Network Link Prediction13
This work introduces a disease Comorbidity prediction model using LLM, named ComLLM, which leverages domain knowledge to enhance the prediction performance and serves as a proof-of-concept for LLM-based systems in disease progression prediction, highlighting its potential for broad applications in healthcare.
- TECRR: a benchmark dataset of radiological reports for BI-RADS classification with machine learning, deep learning, and large language model baselines13
A curated and annotated benchmark dataset that can be used for BI-RADS and breast density category classification is proposed and baseline results of most ML, DL and LLMs models for BI-RADS classification are provided that can be used as a starting point for future investigation.
- A novel multiple kernel fuzzy topic modeling technique for biomedical data13
MKFTM is a new approach to topic modeling, which has the flexibility to work with a variety of clustering methods, which has the flexibility to work with a variety of clustering methods.
- Fuzzy mixed graphs and its application to identification of COVID19 affected central regions in India13
Representation of fuzzy mixed graphs as matrix and isomorphism theorems on fuzzy mixed graphs are developed and a network of COVID19 affected areas in India and central regions are identified as per the proposed theory.
- PersoBench: Benchmarking Personalized Response Generation in Large Language Models12
An automated benchmarking pipeline is presented to evaluate the personalization ability of LLMs in persona-aware dialogue generation within a zero-shot setting, revealing that while LLMs excel at generating fluent and diverse responses, they are far from satisfactory in delivering personalized and coherent responses, considering both the conversation context and the provided personas.
- A Multimodal Prompt-based Framework for Analyzing Code-Mixed and Low-Resource Memes12
The first Nepali meme dataset annotated for hate speech and sentiment is introduced and a novel multimodal framework employing prompt-assisted learning to effectively categorize Nepali memes is introduced, which promotes inclusivity in content moderation research and aligns with UN Sustainable Development Goals.
- Benchmarking Large Language Models for Cryptanalysis and Side-Channel Vulnerabilities12
This research introduces a benchmark dataset of diverse plaintexts, spanning multiple domains, lengths, writing styles, and topics, paired with their encrypted versions, and reveals key insights into LLMs'sths and limitations in side-channel scenarios and raises concerns about their susceptibility to under-generalization-related attacks.
- RUHate-MM: Identification of Hate Speech and Targets using Multimodal Data from Russia-Ukraine Crisis12
A comprehensive multimodal dataset consisting of 20,675 posts related to the Russia-Ukraine crisis, which were manually annotated as either 'Hate Speech' or 'No Hate Speech', which categorize the hate speech data into three targets: 'Individual', 'Organization', and 'Community'.
- Decoding Memes: A Comprehensive Analysis of Late and Early Fusion Models for Explainable Meme Analysis12
In comprehensive experiments, various classifiers such as XGBoost and Random Forest, along with combinations of different vision and text features across multiple sentiment scenarios, showcased the superior effectiveness of late fusion over early fusion.
- 12
- LLM2CLIP: Powerful Language Model Unlocks Richer Cross-Modality Representation11
This work introduces an efficient fine-tuning framework that embeds an LLM into a pretrained CLIP while incurring almost the same training cost as regular CLIP fine-tuning.
- Advancing Accuracy in Multimodal Medical Tasks Through Bootstrapped Language-Image Pretraining (BioMedBLIP): Performance Evaluation Study11
The proposed BioMedBLIP models demonstrate their potential to advance medical image analysis, impacting diagnosis, medical education, and research and suggest that incorporating medical knowledge through pretraining with domain-specific medical data sets helps models achieve higher performance.
- Graph-Based Hierarchical Attention Network for Suicide Risk Detection on Social Media11
A new graph-based hierarchical attention network that uses a graph convolutional neural network with an ordinal loss to improve suicide risk identification on social media and shows that the proposed GHAN outperformed state-of-the-art methods on a public Reddit dataset.
- Diabetic Retinopathy Detection Using Multi-layer Neural Networks and Split Attention with Focal Loss11
An automatic detection of DR using Multi-layer Neural Networks and Split Attention with Focal Loss outperformed state-of-the-art (SOTA) networks in early-stage detection and achieved 85.9% accuracy in DR classification.
- Application of Fuzzy Logic in the Ranking of Academic Institutions11
This study will focus on the ranking system based on Mamdani and Sugeno with fuzzy triangular numbers as input values of membership functions, which is compared with existing NIRF systems and considered Top 20 institutions of India for the year 2019 as per NIRf.
- VaxGuard: A Multi-Generator, Multi-Type, and Multi-Role Dataset for Detecting LLM-Generated Vaccine Misinformation10
Results show that GPT-3.5 and GPT-4o consistently outperform other LLMs in detecting misinformation, especially when dealing with subtle or emotionally charged narratives, and suggest that VaxGuard can serve as a key resource for improving the detection of LLM-generated vaccine misinformation.
- Did You Tell a Deadly Lie? Evaluating Large Language Models for Health Misinformation Identification10
- 10
- MRGAgents: A Multi-Agent Framework for Improved Medical Report Generation with Med-LVLMs9
By curating subsets of the IU X-ray and MIMICCXR datasets to train disease-specific agents, the MRGAgents generates reports that effectively balance normal and abnormal findings while ensuring a comprehensive description of clinically relevant regions.
- Multimodal Generative AI with Autoregressive LLMs for Human Motion Understanding and Generation: A Way Forward9
This systematic survey underscores the transformative potential of text-to-motion GenAI and LLM architectures in applications such as healthcare, humanoids, gaming, animation, and assistive technologies, while addressing ongoing challenges in generating efficient and realistic human motion.
- Alleviating Textual Reliance in Medical Language-Guided Segmentation via Prototype-Driven Semantic Approximation9
ProLearn is proposed, the first Prototype-driven Learning framework for language-guided segmentation that fundamentally alleviates textual reliance and outperforms state-of-the-art language-guided methods when limited text is available.
- 9
- Incorporating historical information by disentangling hidden representations for mental health surveillance on social media9
Experimental results show that the incorporate a time-aware framework that jointly learns the context of the user’s historical posts and temporal posting irregularities outperforms state-of-the-art methods for mental health surveillance on social media.
- Temporal Tides of Emotional Resonance: A Novel Approach to Identify Mental Health on Social Media9
Experimental results demonstrate proposed method outperforms recent competitive methods, demonstrating the importance of capturing the user’s temporal emotional spectrum and time-aware emotional context through historical posts for social media users’ mental health surveillance.
- 9
- Can reasoning LLMs enhance clinical document classification?8
Evaluated LLMs in binary classification clinical discharge summaries based on ICD-10 codes suggest that a hybrid approach, leveraging the strengths of both model types, could optimize automated clinical coding by balancing accuracy and reliability.
- 8
- Should LLM Safety be More Than Refusing Harmful Instructions?8
A two-dimensional evaluation framework is introduced that separately assesses early instruction refusal and generation safety and reveals that most models struggle at balancing both dimensions effectively—prioritizing either instruction refusal or response suppression.
- Robust and explainable framework to address data scarcity in diagnostic imaging8
This work introduces a novel ensemble framework called 'Efficient Transfer and Self-supervised Learning based Ensemble Framework' (ETSEF), which leverages features from multiple pre-trained deep learning models to efficiently learn powerful representations from a limited number of data samples.
- 8
- Performance Evaluation of Deep Learning and Transformer Models Using Multimodal Data for Breast Cancer Classification8
A multimodal DL architecture for BC classification is proposed, utilising images (mammograms; four views) and textual data (radiological reports) from the authors' new in-house dataset, and different feature extractor and classifier arrangements are evaluated.
- 8
- 8
- Reducing Knowledge Noise for Improved Semantic Analysis in Biomedical Natural Language Processing Applications8
Two novel methods are proposed for integrating knowledge graph-based language model with nearest-neighbor models to incorporate semantic and category information from neighboring instances in the graph to enhance the performance of language models and improving NLP applications in the biomedical domain.
- 8
- Concepts on Coloring of Cluster Hypergraphs with Application8
This study analyzes cluster hypergraphs where cluster vertices too contain simple vertices and proper coloring and strong coloring of cluster hyper graphs have been defined.
- 7
- 7
- From Native Memes to Global Moderation: Cross-Cultural Evaluation of Vision-Language Models for Hateful Meme Detection7
Results show that the common ''translate-then-detect'' approach deteriorate performance, while culturally aligned interventions — native-language prompting and one-shot learning — significantly enhance detection.
- Cross Platform MultiModal Retrieval Augmented Distillation for Code-Switched Content Understanding7
This paper proposes a novel multimodal model, MM-RAD, designed to understand code-switched Nepali-English memes, leveraging both textual and visual content, and demonstrates superior performance in sentiment analysis and hate speech detection compared to individual baseline models.
- Steering Towards Fairness: Mitigating Political Bias in LLMs7
This work introduces a comprehensive activation extraction pipeline capable of layer-wise analysis across multiple ideological axes, revealing meaningful disparities linked to political framing and provides new insights into how political bias is encoded in LLMs.
- Doc‐KG: Unstructured documents to knowledge graph construction, identification and validation with Wikidata7
An enhanced Doc‐KG model is introduced, a sophisticated approach designed to transform unstructured documents into structured knowledge by generating local KGs and mapping these to a target KG, such as Wikidata, employing Wikidata's Unified Resource Identifiers for precise mapping.
- DialectNLU at NADI 2023 Shared Task: Transformer Based Multitask Approach Jointly Integrating Dialect and Machine Translation Tasks in Arabic7
This paper addresses this need by presenting a system description of the approaches employed for the subtasks outlined in the Nuanced Arabic Dialect Identification (NADI) task at EMNLP 2023.
- 7
- LowResourceNLU at BLP-2023 Task 1 & 2: Enhancing Sentiment Classification and Violence Incitement Detection in Bangla Through Aggregated Language Models7
This paper presents an innovative approach that leverages aggregated BERT models for two tasks at the BLP workshop in EMNLP 2023, specifically tailored for Bangla, and significantly enhances the accuracy of sentiment classification and violence incitement detection in Bangla text.
- 7
- 7
- 7
- Health-ORSC-Bench: A Benchmark for Measuring Over-Refusal and Safety Completion in Health Context6
Health-ORSC-Bench provides a rigorous standard for calibrating the next generation of medical AI assistants toward nuanced, safe, and helpful completions, and facilitates reproducible evaluation, encourages safety calibration, and supports development of clinically reliable, context-aware, human-aligned medical AI systems.
- 6
- Seeing the Threat: Vulnerabilities in Vision-Language Models to Adversarial Attack6
A systematic representational analysis is conducted to uncover why conventional adversarial attacks can circumvent the safety mechanisms embedded in LVLMs, and a novel two-stage evaluation framework for adversarial attacks on LVLMs is proposed.
- SemEval-2026 Task 9: Detecting Multilingual, Multicultural and Multievent Online Polarization6
The baseline results and performance of the best-performing systems are reported and the performance of the best-performing systems are analyzed, highlighting the most common approaches and the most effective methods across different subtasks and languages.
- Simulating Influence Dynamics with LLM Agents6
This paper introduces a simulator designed for opinion dynamics researchers to model competing influences within social networks in the presence of LLM-based agents, enabling the study of influence propagation and counter-misinformation strategies.
- Explainable Medical Image Captioning6
This research aims to transform the landscape of medical imaging by introducing an interpretable approach for generating precise medical captions from X-ray images and associated data by proposing a novel framework that integrates advanced deep learning techniques with explainable AI.
- Arabic text detection: a survey of recent progress challenges and opportunities6
This study provides a survey of research in the field of Arabic Text Detection (ATD) published from 2017 to 2023 in a two-fold manner and highlights new directions and discusses the open challenges and opportunities which assist researchers in identifying future work.
- NLP as a Lens for Causal Analysis and Perception Mining to Infer Mental Health on Social Media6
This position paper advocates for a more explainable approach toward modeling computational psychology problems through the lens of language as an increased number of research contributions in dataset and problem formulation for causal relation extraction and perception enhancements while inferring mental states.
- LowResContextQA at Qur’an QA 2023 Shared Task: Temporal and Sequential Representation Augmented Question Answering Span Detection in Arabic6
The approach to task B of Qur’an QA 2023, part of EMNLP 2023 is presented, addressing this challenge by proposing a robust method for extracting answers from Qur‘anic passages by employing innovative techniques and advanced models to improve the precision and contextuality of answers derived from Qur'anic passages.
- 6
- UICD: A new dataset and approach for urdu image captioning5
This study creates a new Urdu Image Captioning Dataset (UCID) called UC-23-RY to fill in the gaps in Urdu image captioning and suggests deep learning architectures designed especially for Urdu image captioning, including NASNetLarge-LSTM and ResNet-50-LSTM.
- Competing LLM Agents in a Non-Cooperative Game of Opinion Polarisation5
These findings demonstrate that while weaker confirmation bias strengthens opinion alignment within groups, it also exacerbates overall polarisation and investing heavily in a high-resource debunking strategy can initially align the population with the debunking agent, but risks rapid resource depletion and diminished long-term influence.
- SHIELD: Classifier-Guided Prompting for Robust and Safer LVLMs5
This work proposes SHIELD, a lightweight, model-agnostic preprocessing framework that couples fine-grained safety classification with category-specific guidance and explicit actions (Block, Reframe, Forward) that consistently lowers jailbreak and non-following rates while preserving utility.
- 5
- A Linguistic Grounding-Infused Contrastive Learning Approach for Health Mention Classification on Social Media5
A novel method that uses a transformer-based PLM to obtain a contextual representation of target (disease or symptom) terms coupled with a contrastive loss to establish a larger gap between target terms' literal and figurative uses using linguistic theories is proposed.
- A Causal Approach to Mitigate Modality Preference Bias in Medical Visual Question Answering5
A Medical CounterFactual VQA (MedCFVQA) model is proposed, which trains with bias and leverages causal graphs to eliminate the modality preference bias during inference and significantly outperforms its non-causal counterpart on both SLAKE, RadVQA and SLAKE-CP, RadVQA-CP datasets.
- 5
- Automated Citation Function Classification and Context Extraction in Astrophysics: Leveraging Paraphrasing and Question Answering5
An automated approach is introduced that leverages state-of-the-art language models, including ALBERT, RoBERTa, BERT, and DistillBERT, to classify citation functions and extract context within astrophysical para-graphs and demonstrates the robustness of this approach.
- A New Method for Centrality Measurement Using Generalized Fuzzy Graphs5
An application to detect the central person in any online group like WhatsApp is described in this study by using generalized fuzzy graphs and a few important results are established.
- 5
- Bias Beyond Borders: Political Ideology Evaluation and Steering in Multilingual LLMs4
A complementary post-hoc mitigation framework, Cross-Lingual Alignment Steering (CLAS) is introduced, designed to augment existing steering methods by aligning ideological representations across languages and dynamically regulating intervention strength, establishing a scalable and interpretable paradigm for fairness-aware multilingual LLM governance.
- Do Personality Traits Interfere? Geometric Limitations of Steering in Large Language Models4
This work analyses the geometric relationships between Big Five personality steering directions to suggest that personality traits in LLMs occupy a slightly coupled subspace, limiting fully independent trait control.
- Uno-Orchestra: Parsimonious Agent Routing via Selective Delegation4
Uno-Orchestra is introduced, a unified orchestration policy that selectively decomposes a task and dispatches each subtask to an admissible (model, primitive) pair, with both decisions learned together from curated RL trajectories grounded in real worker interactions.
- Activation-Space Personality Steering: Hybrid Layer Selection for Stable Trait Control in LLMs4
The findings reveal that personality traits occupy a low-rank shared subspace, and that these latent structures can be transformed into actionable mechanisms for effective steering through careful perturbations without impacting the fluency, variance and general capabilities, helping to bridge the gap between psychological theory and practical model alignment.
- We Think, Therefore We Align LLMs to Helpful, Harmless and Honest Before They Go Wrong4
Adaptive Multi-Branch Steering (AMBS), a two-stage framework in a 1-to-N Transformer setting that parameterizes objective-specific transformations relative to a shared representation, is proposed.
- MRG-R1: Reinforcement Learning for Clinically Aligned Medical Report Generation4
A semantic-driven reinforcement learning (SRL) framework for medical report generation, named MRG-R1, which directly optimizes report-level clinical correctness rather than token-level likelihood, and improves the accuracy and coverage of clinically relevant findings in generated reports.
- Vaccine Misinformation Detection in X using Cooperative Multimodal Framework4
The experimental results show that the VaxMine approach outperforms state-of-the-art methods with an F1-Score of 0.94, demonstrating that extracting relevant content from users' historical posts and understanding both modalities are essential to detecting anti-vaccine users on social media.
- 4
- Vision-Language Models for Biomedical Applications4
An overview of the inaugural edition of the VLM4Bio workshop is provided, summarizing the key discussions, contributions, and future directions for expanding the workshop's scope and influence in subsequent editions.
- 4
- Structural Representation Learning and Disentanglement for Evidential Chinese Patent Approval Prediction4
A novel framework called DiSPat is proposed, which focuses on structural representation learning and disentanglement to predict the approval of Chinese patents and offer decision-making evidence, and surpasses state-of-the-art baselines on patent approval prediction, while also exhibiting enhanced evidentiality.
- Incorporating Embedding to Topic Modeling for More Effective Short Text Analysis4
An Embedding-based topic modeling approach that incorporates word embedding and hierarchical clustering to identify significant topics is proposed that indicates a superiority of EmTM over baseline topic models by its exceptional performance in both classification accuracy and topic coherence metrics.
- 4
- KnowTellConvince at ArAIEval Shared Task: Disinformation and Persuasion Detection in Arabic using Similar and Contrastive Representation Alignment4
This work fills a critical research gap in content analysis for Arabic, empowering individuals, communities, and digital platforms to combat deceptive content effectively and preserve the credibility of information sources within the Arabic-speaking world.
- A Proposed Framework for the Security of Financial Systems4
A framework for the enhancement the security level of the eBanking or online banking systems and defines parameters of Wireless Security design and deploy security algorithms and techniques to offer secure network.
- 3
- Robust Harmful Meme Detection under Missing Modalities via Shared Representation Learning3
This work presents the first-of-its-kind work to comprehensively investigate the behavior of harmful meme detection methods in the presence of modal-incomplete data, and proposes a new baseline method that learns a shared representation for multiple modalities by projecting them independently when data is modal-incomplete.
- When the Model Said 'No Comment', We Knew Helpfulness Was Dead, Honesty Was Alive, and Safety Was Terrified3
AlignX, a two-stage framework that uses prompt-injected fine-tuning to extract axis-specific task features, mitigating catastrophic forgetting, and a MoCaE module that calibrates expert routing using fractal and natural geometry, improving inference reliability is proposed.
- They Said Memes Were Harmless --- We Found the Ones That Hurt: Decoding Jokes, Symbols, and Cultural References3
CROSS-ALIGN+ is introduced, a three-stage framework that systematically addresses cultural blindness, boundary ambiguity, and lack of interpretability in meme-based social abuse detection while providing interpretable justifications for each decision.
- Fairness Evaluation and Inference Level Mitigation in LLMs3
This work proposes a dynamic, reversible, pruning-based framework that detects context-aware neuron activations and applies adaptive masking to modulate their influence during generation, enabling dynamic fairness control in real-world conversational AI.
- ExBigBang: A Dynamic Approach for Explainable Persona Classification through Contextualized Hybrid Transformer Analysis3
This work presents ExBigBang (Explainable BigBang), a hybrid text-tabular approach that uses transformer-based architectures to model rich contextual features for persona classification and incorporates metadata, domain knowledge, and user profiling to embed deeper context into predictions.
- Evaluating Multimodal Large Language Models on Educational Textbook Question Answering3
The first evaluation of state-of-the-art MLLMs, including LLaVA-1.5 and LLaMA 3.2-Vision, on the textbook question answering (TQA) task using the CK12-QA dataset is presented, providing a benchmark and pointing to key directions for developing more robust AI-driven educational tools.
- 3
- Artificial Intelligence for Text Analysis in the Arabic and Related Middle Eastern Languages: Progress, Trends, and Future Recommendations3
A comprehensive review of recent developments in Arabic text classification (ATC) and Arabic text representation (ATR) finds that while deep learning models, particularly transformer‐based architectures, are increasingly effective for ATC, challenges such as dialectal variations and insufficient labeled datasets remain key obstacles.
- 3
- 3
- KG-UQ: Knowledge Graph-Based Uncertainty Quantification for Long Text in Large Language Models3
This work proposes KG-UQ, a UQ method leveraging KGs to address the semantic intricacies of long text, and demonstrates that knowledge graphs are an effective tool for decomposing long text into fundamental statements.
- Multimodal Hyperbolic Graph Learning for Alzheimer’s Disease Detection3
This work proposes a novel multimodal Poincare Frechet mean graph convolutional network (MochaGCN), which combines hyperbolic graph convolution and Poincare Frechet mean to extract features from multimodal brain networks, enhancing their representations in hyperbolic space.
- Breaking Barriers: Exploring the Diagnostic Potential of Speech Narratives in Hindi for Alzheimer’s Disease3
The analysis suggests that speech narratives in the Hindi language have the potential to aid in the diagnosis of AD, and this work explores the possibility of using a low-resourced language, i.e., Hindi language, to diagnose AD.
- Self-Supervised Learning for Multilevel Skeleton-Based Forgery Detection via Temporal-Causal Consistency of Actions3
This work proposes an approach to self-supervised learning of the temporal causality behind human action, which can effectively check TII in skeletal sequences and designs a multilevel skeleton-based forgery detection framework to recognize the forgery on frame level, clip level, and action level in terms of learning the corresponding temporal-causal skeleton representations for each level.
- VISPA: Pluralistic Alignment via Automatic Value Selection and Activation2
This work introduces VISPA, a training-free pluralistic alignment framework that enables direct control over value expression by dynamic selection and internal model activation steering, and suggests that pluralistic alignment can be achieved through internal activation mechanisms, offering a scalable path toward language models that serves all.
- Dynamic Traceback Learning for Medical Report Generation2
This study proposes a novel multimodal dynamic traceback learning framework (DTrace), introducing a traceback mechanism to supervise the semantic validity of generated content and a dynamic learning strategy to adapt to various proportions of image and text input, enabling text generation without strong reliance on the input from both modalities during inference.
- Intersectional fairness in vision-language models for medical image disease classification2
Cross-Modal Alignment Consistency (CMAC-MMD) is developed, a training framework that standardises diagnostic certainty across intersectional patient subgroups and establishes a scalable framework for developing high-stakes clinical decision support systems that are both accurate and can perform equitably across diverse patient subgroups.
- Do Large Language Models Reflect Demographic Pluralism in Safety?2
Demo-SafetyBench addresses demographic pluralism directly at the prompt level by modeling demographic pluralism directly at the prompt level, decoupling value framing from responses, confirming that pluralistic safety evaluation can be both scalable and demographically robust.
- Can Argus Judge Them All? Comparing VLMs Across Domains2
ARGUS-EVAL, a capability-reliability-oriented evaluation framework for VLMs that characterizes model behavior through Benchmark Capability P, Cross-Dataset Consistency CDC, Robustness Retention RR, and Efficiency E(M), is presented.
- Addressing the Challenges of Mental Health Conversations with Large Language Models2
Modifications to ChatMGL are proposed, including fine-tuning the model on a mental health dataset without proximal policy optimization and incorporating dialogue act labels, to enhance its ability to handle the complex nature of mental health dialogues.
- 2
- 2
- DiGrI: Distorted Greedy Approach for Human-Assisted Online Suicide Ideation Detection2
A novel model called DiGrI, or Distorted Greedy Approach for Human-Assisted Online Suicide Ideation Detection, which reformulates suicide ideation assessment as a selective, prioritized prediction problem and outperforms strong comparative models including large language models in detecting mental health issues on a publicly available Reddit dataset.
- 2
- PoliMeme: Exploring Offensive Meme Propagation in the Israel-Palestine Conflict2
PoliMeme is introduced, a novel multimodal dataset encompassing text-embedded memes relevant to the Israel-Palestine conflict, meticulously annotated for sentiment analysis and offensive content detection, further subdivided into direction of offense (directed vs. undirected) and targeted entities (individuals, organizations, communities).
- 2
- NepAES: exploring the promise of automated essay scoring for Nepali essays2
The findings demonstrate that the effectiveness of AES systems is greatly influenced by the quality of translations, as the accuracy and precision of the translation process have a direct impact on the overall performance of the AES models.
- Agentic Moderation: Multi-Agent Design for Safer Vision-Language Models2
This work introduces Agentic Moderation, a model-agnostic framework that leverages specialized agents to defend multimodality models against jailbreak attacks and provides modular, scalable, and fine-grained safety enforcement, highlighting the potential of agentic systems as a foundation for automated safety governance.
- Foreword for the First International Workshop on Multimodal Content Analysis for Social Good (MM4SG 2024)2
The first international workshop on multimedia content analysis for social good (MM4SG) was held in conjunction with the Web Conference 2024 and describes the inaugural edition.
- Debunking Biases in Attention2
The results indicate that the majority of attention mechanisms that have been assessed can improve the fairness performance of Biddirectional Gated Recurrent Unit (BiGRU) and Bidirectional Long Short-Term Memory (BiLSTM) in all three datasets regarding religious and gender-sensitive groups, however, with varying degrees of trade-offs in accuracy measures.
- A Dynamics and Task Decoupled Reinforcement Learning Architecture for High-Efficiency Dynamic Target Intercept2
This paper proposes a dynamics and task decoupled RL architecture to address the inefficient learning procedure, where the RL module focuses on modeling the DTI task without involving physical dynamics, and the design of states, actions, and rewards are completely task-oriented.
- A Summary of Covid-19 Datasets2
The purpose of this research is to summarize the open COVID-19 datasets to make them more accessible to the research community for health systems design and analysis.
- 2
- Newly-Published Paper Recommendation With a Joint Multi-Relational Model2
A time-aware joint model is designed to infer users’ preference for the newly-published papers with the help of subsidiary relations of social and article linkages and outperforms the state-of-the-art methods.
- An Approach to Ensure Fairness in News Articles2
Dbias is a trained Machine Learning pipeline that can take a text and detects if the text is biased or not, and recommends a set of sentences with new words that are bias-free or at least less biased.
- 2
- Classification of Neuroblastoma Histopathological Images Using Machine Learning2
Using dataset gathered from The Tumour Bank at Kids Research at The Children’s Hospital at Westmead, which has been used in previous research, a range of feature extraction and data undersampling and over-sampling techniques are explored to improve classification accuracy.
- A Study on Discrete Mathematics: Sum Distance in Neutrosophic Graphs with Application2
This study defined sum distance in a neutrosophic graph based on the idea of strong sum distance and a real-life application on the travelling salesman problem is illustrated.
- No Evidence, No Problem: When Less is More for Out of Context Multimodal Misinformation Detection1
This work introduces a lightweight and evidence-free framework for OOC misinformation detection that achieves competitive performance with high efficiency, while offering low computational overhead and strong interpretability, making it well-suited for real-world applications.
- Are Aligned Large Language Models Still Misaligned?1
This work constructs SAVACU, an English misaligned-aligned dataset of 382,424 samples spanning 112 domains, and benchmarks general-purpose, fine-tuned, and open-weight LLMs, enabling systematic evaluation of misalignment under three dimensions.
- Beyond Uniform Forgetting: A Study of Sequential Direct Preference Optimization Across Preference Settings1
Future sequential alignment pipelines should account for objective compatibility and signal strength, rather than assuming that later objectives affect earlier preferences uniformly, according to objective relationship, signal strength, and training order.
- 1
- 1
- A comprehensive review of use cases, misuses, and potential mitigation techniques in generative artificial intelligence1
A comprehensive review of the history of GenAI, its use cases, recent developments, downsides, and solutions to address the problem of its misuse in various sectors are provided.
- SafeConstellations: Mitigating Over-Refusals in LLMs Through Task-Aware Representation Steering1
SafeConstellations is introduced, an inference-time trajectory-shifting approach that tracks task-specific trajectory patterns and guides representations toward non-refusal pathways and reduces over-refusal with minimal impact on utility.
- AlignCultura: Towards Culturally Aligned Large Language Models?1
This work built Align-Cultura, two-stage pipeline for cultural alignment, which benchmarks CULTURAX on general-purpose models, culturally fine-tuned models, and open-weight LLMs (Qwen3-8B and DeepSeek-R1-Distill-Qwen-7B).
- Med-R2: Perception and Reflection-driven Complex Reasoning for Medical Report Generation1
The proposed Med-R2 introduces a perception-driven long reasoning process that precedes report generation and incorporates radiology-specific knowledge as guidance and a reflection mechanism is introduced to refine the perception of pathological features and the generated report.
- 1
- 1
- 1
- CUFG: Curriculum Unlearning Guided by the Forgetting Gradient1
This work proposes CUFG (Curriculum Unlearning via Forgetting Gradients), a novel framework that enhances the stability of approximate unlearning through innovations in both forgetting mechanisms and data scheduling strategies that narrow the gap with the gold-standard Retrain method.
- PersoPilot: An Adaptive AI-Copilot for Transparent Contextualized Persona Classification and Personalized Response Generation1
PersoPilot is an agentic AI-Copilot that integrates persona understanding with contextual analysis to support both end users and analysts, and is applicable to a broad range of service personalization scenarios.
- ChildGuard: A Specialized Dataset for Combatting Child-Targeted Hate Speech1
ChildGuard is a large-scale English dataset for child-targeted hate speech containing 351,877 annotated instances collected from X, Reddit, and YouTube, and contains two subsets such as a contextual subset and a lexical subset.
- Third International Workshop on Multimodal Content Analysis for Social Good1
This paper highlights the key themes, discussions, and contributions of the third edition of the MM4SG workshop, with a particular focus on the intersection of computational linguistics and multimodal content analysis.
- 1
- 1
- 1
- 1
- How Can Multimodal Remote Sensing Datasets Transform Classification Via Spatialnet-ViT?1
A novel model is proposed, SpatialNet-ViT, leveraging the power of Vision Transformers (ViTs) and Multi-Task Learning (MTL), which combines spatial awareness with contextual understanding, improving both classification accuracy and scalability.
- Medical Domain Knowledge Collaborative Graph Learning for Healthcare Event Prediction1
The proposed MED‐CGL model incorporates external medical knowledge bases to enhance the predictive power of unstructured clinical notes and extracts learnable features from the MIMIC‐III health record dataset using medical domain knowledge and collaborative graph learning.
- Entity-Aware Optimal Transport and Residual Attention for Multimodal Content Moderation1
The findings highlight that an approach that balances local token-level alignment with broader contextual modeling can effectively detect harmful memes across diverse topical domains, paving the way for more robust and transparent content moderation on social media.
- 1
- 1
- VR-DiagNet: Medical Volumetric and Radiomic Diagnosis Networks with Interpretable Clinician-like Optimizing Visual Inspection1
The findings showcase superior performance, as evidenced by heightened accuracy and area under the curve (AUC) metrics, reduced computational overhead, and expedited convergence while conclusively illustrating the immense value of integrating volumetric and radiomic modalities for the authors' current problem setup.
- Automated Detection and Segmentation of Glioblastoma in MRI Using Multi-Level Diffusion Transformer U-Net1
A multi-level Diffusion Transformer U-Net (mDT-UNet) with a modified cross-attention module is proposed with an average accuracy of 95% on the BraTS 2020 dataset whereas U-Net and Swin Transformer achieved 86% accuracy.
- 1
- 1
- Deep Learning Based Multi-Scale Approach for Precision Medicine and Quantitative Imaging in Glioblastoma1
A multiscale multilevel approach based on deep learning for precision medicine and quantitative imaging in GBM capturing image feature and providing wide-ranging contextual information and surpasses state-of-the-art methods in Glioblastoma segmentation.
- 1
- Coherent Topic Modeling for Creative Multimodal Data on Social Media1
A Coherent Topic modeling for Multimodal Data (CTM-MM), which takes into account that text in social media posts typically relates to one topic, while images can contain information about multiple topics, outperforms traditional multi- topic models in terms of classification and topic coherence.
- Data Mining ENCODE Data Predicts a Significant Role of SINA3 in Human Liver Cancer1
A bioinformatics pipeline to investigate Chip-seq DNA binding proteins datasets for HepG2 liver cancer cell line downloaded from ENCODE project showed a strong enrichment of DNA-binding protein SIN3A to activator and repressor indicating Sin3A plays has an important regulatory role in vital liver functions.
- –
- –
- –
- Can Large Language Models Make Everyone Happy?–
The MisAlign-Profile is introduced, a unified benchmark for measuring misalignment trade-offs inspired by mechanistic profiling, and general-purpose, fine-tuned, and open-weight LLMs are benchmarked on MISALIGNTRADE, revealing 12%-34% misalignment trade-offs across dimensions.
- –
- –
- –
- Counterfactual Augmented Causal Reasoning for Aspect-Based Sentiment Analysis–
A new counterfactual-enhanced causal reasoning (CECR) framework to reduce spurious correlations in aspect-based sentiment analysis and achieves outstanding performance on state-of-the-art baselines.
- –
- –
- –
- –
- –
- –
- Towards safe and trustworthy agentic AI: foundations, taxonomy, technologies, applications, and future directions–
This work synthesizes perspectives from philosophy, cognitive science, and AI to define agency, outline its key properties, and situate it in relation to existing paradigms such as reinforcement learning, symbolic reasoning, Belief–Desire–Intention (BDI) architectures, and embodied cognition.
- –
- –
- Concept-Grounded Detection of Vaccine Misinformation in Multimodal Content Using Interpretable Vision-Language Models–
Results show that grounding stance prediction in structured concept representations consistently outperforms direct LVLM classification, yielding accuracy improvements of approximately 10--17% while enabling explicit inspection of the visual and rhetorical cues driving model decisions.
- Divergent Paths to Depolarization: Dialogue Design Shapes the Intergroup Attitudinal Effects of AI-Assisted Political Argumentation–
Experimental findings show attitude-congruent dialogues produced greater immediate reduction in both affective and opinion polarization than attitude-incongruent dialogues, highlighting dialogue design as a key determinant of effective AI-mediated behavioral interventions.
- –
- Multimodal Content Analysis for Social Good–
A major theme of this edition was the growing role of agentic multimodal systems and large vision-language models, including emerging agentic LVLM paradigms, for scalable reasoning, moderation, and decision-making in real-world settings.
- –
- Exploring Internal Emphasis for Robust Noisy RAG–
The key insight is that, because relevance signals reside in only a few modules, emphasis should be applied directly within these modules rather than only at the text level, enabling more reliable passage-denoising.
- –
- –
- –
- –
- LLMs as Historical Actors: How AI Systems Influence the Web's Evolution–
It is argued that LLMs can be framed as historical actors, meaning they are not only facilitators of human endeavors but also independent forces that influence the web's evolution.
- –
- –
- –
- –
- Few Labels with Active Learning: From Weak to Strong Labels for Misinformation Detection–
This work proposes a framework called Few Labels with Active Learning (FLAL), which leverages a margin-sampling technique within the active learning paradigm to enable the model to prioritize and learn from the most informative yet uncertain instances, enhancing its performance by focusing iteratively on examples that contribute the most to learning.
- –
- BEAST: Leveraging Contrastive Learning and Unsupervised Sentence Embeddings for Improved Drug Abuse Detection–
BEAST is proposed, a novel approach that leverages contrastive learning to improve the detection of drug abuse references hidden within figurative language on social media.
- RF-DTR: A Multi-Stage DCT Token Regression Network for Progressive Rib Fracture Mask Refinement–
A novel regression network, RF-DTR, incorporating a gated regressor mechanism and operating entirely in the frequency domain is proposed, incorporating a Mahalanobis regularization technique to enhance the model and learn high-frequency DCT components relevant to clinical tasks.
- Enforcing Consistency and Fairness in Multi-level Hierarchical Classification with a Mask-based Output Layer–
This paper introduces a fair, model-agnostic layer designed to enforce taxonomy and optimize specific objectives, including consistency, fairness, and exact match in hierarchical classification, achieving better outcomes in both fairness and accuracy.
- –
- –
- –
- Bridging the Semantic Gaps: Improving Medical VQA Consistency with LLM-Augmented Question Sets–
A Semantically Equivalent Question Augmentation (SEQA) framework is proposed, which leverages large language models (LLMs) to generate diverse yet semantically equivalent rephrasings of questions to enriches linguistic diversity while preserving semantic meaning.
- –
- –
- –
- A Systematic Overview and Comparative Analysis of Service Discovery Protocols and Middlewares–
This paper overviews context aware systems, their sensing capabilities in location or beyond location along with COIVA (a context aware system), and named CoCaMAAL as better middleware as compared to others.
- –
- –
Publication data from OpenAlex, with missing venues and authors filled in from Crossref; citation counts are the higher of OpenAlex and Semantic Scholar, last synced 2026-10-12. One-sentence summaries under some papers are written by Semantic Scholar’s model. Citation counts may be lower than on Google Scholar, which indexes more sources.
Report an error
Wrong papers, two people merged into one, or a profile that should not be here? Tell us and we will fix or hide it.