Research Papers

As AI systems increasingly influence critical sectors like telecommunications, finance, healthcare, and public services, ensuring fairness in decision-making is essential to prevent biased or unjust outcomes that disproportionately affect vulnerable entities or result in adverse impacts. This need is particularly pressing as the industry approaches the 6G era, where AI will drive complex functions like autonomous network management and hyper-personalized services. The TEC Standard for Fairness Assessment and Rating of AI Systems provides guidelines for evaluating fairness in AI, focusing primarily on tabular data and supervised learning models. However, as AI applications diversify, this standard requires enhancement to strengthen its impact and broaden its applicability. This paper proposes an expansion of the TEC Standard to include fairness assessments for images, unstructured text, and generative AI, including large language models, ensuring a more comprehensive approach that keeps pace with evolving AI technologies. By incorporating these dimensions, the enhanced framework will promote responsible and trustworthy AI deployment across various sectors.

Harmful text detection has become a crucial task in the development and deployment of large language models, especially as AI-generated content continues to expand across digital platforms. This study proposes a joint retrieval framework that integrates pre-trained language models with knowledge graphs to improve the accuracy and robustness of harmful text detection. Experimental results demonstrate that the joint retrieval approach significantly outperforms single-model baselines, particularly in low-resource training scenarios and multilingual environments. The proposed method effectively captures nuanced harmful content by leveraging external contextual information, addressing the limitations of traditional detection models. Future research should focus on optimizing computational efficiency, enhancing model interpretability, and expanding multimodal detection capabilities to better tackle evolving harmful content patterns. This work contributes to the advancement of AI safety, ensuring more trustworthy and reliable content moderation systems.

Functional fixedness, a cognitive bias that restricts users' interactions with a new system or tool to expected or familiar ways, limits the full potential of Large Language Model (LLM)-enabled chat search, especially in complex and exploratory tasks. To investigate its impact, we conducted a crowdsourcing study with 450 participants, each completing one of six decision-making tasks spanning public safety, diet and health management, sustainability, and AI ethics. Participants engaged in a multi-prompt conversation with ChatGPT to address the task, allowing us to compare pre-chat intent-based expectations with observed interactions. We found that: 1) Several aspects of pre-chat expectations are closely associated with users' prior experiences with ChatGPT, search engines, and virtual assistants; 2) Prior system experience shapes language use and prompting behavior. Frequent ChatGPT users reduced deictic terms and hedge words and frequently adjusted prompts. Users with rich search experience maintained structured, less-conversational queries with minimal modifications. Users of virtual assistants favored directive, command-like prompts, reinforcing functional fixedness; 3) When the system failed to meet expectations, participants generated more detailed prompts with increased linguistic diversity, reflecting adaptive shifts. These findings suggest that while preconceived expectations constrain early interactions, unmet expectations can motivate behavioral adaptation. With appropriate system support, this may promote broader exploration of LLM capabilities. This work also introduces a typology for user intents in chat search and highlights the importance of mitigating functional fixedness to support more creative and analytical use of LLMs.

In this research, we introduce BEATS, a novel framework for evaluating Bias, Ethics, Fairness, and Factuality in Large Language Models (LLMs). Building upon the BEATS framework, we present a bias benchmark for LLMs that measure performance across 29 distinct metrics. These metrics span a broad range of characteristics, including demographic, cognitive, and social biases, as well as measures of ethical reasoning, group fairness, and factuality related misinformation risk. These metrics enable a quantitative assessment of the extent to which LLM generated responses may perpetuate societal prejudices that reinforce or expand systemic inequities. To achieve a high score on this benchmark a LLM must show very equitable behavior in their responses, making it a rigorous standard for responsible AI evaluation. Empirical results based on data from our experiment show that, 37.65\% of outputs generated by industry leading models contained some form of bias, highlighting a substantial risk of using these models in critical decision making systems. BEATS framework and benchmark offer a scalable and statistically rigorous methodology to benchmark LLMs, diagnose factors driving biases, and develop mitigation strategies. With the BEATS framework, our goal is to help the development of more socially responsible and ethically aligned AI models.

The malicious use or malfunction of advanced general-purpose AI (GPAI) poses risks that, according to leading experts, could lead to the 'marginalisation or extinction of humanity.' To address these risks, there are an increasing number of proposals for international agreements on AI safety. In this paper, we review recent (2023-) proposals, identifying areas of consensus and disagreement, and drawing on related literature to assess their feasibility. We focus our discussion on risk thresholds, regulations, types of international agreement and five related processes: building scientific consensus, standardisation, auditing, verification and incentivisation. Based on this review, we propose a treaty establishing a compute threshold above which development requires rigorous oversight. This treaty would mandate complementary audits of models, information security and governance practices, overseen by an international network of AI Safety Institutes (AISIs) with authority to pause development if risks are unacceptable. Our approach combines immediately implementable measures with a flexible structure that can adapt to ongoing research.

Semantic-based test generators are widely used to produce failure-inducing inputs for Deep Learning (DL) systems. They typically generate challenging test inputs by applying random perturbations to input semantic concepts until a failure is found or a timeout is reached. However, such randomness may hinder them from efficiently achieving their goal. This paper proposes XMutant, a technique that leverages explainable artificial intelligence (XAI) techniques to generate challenging test inputs. XMutant uses the local explanation of the input to inform the fuzz testing process and effectively guide it toward failures of the DL system under test. We evaluated different configurations of XMutant in triggering failures for different DL systems both for model-level (sentiment analysis, digit recognition) and system-level testing (advanced driving assistance). Our studies showed that XMutant enables more effective and efficient test generation by focusing on the most impactful parts of the input. XMutant generates up to 125% more failure-inducing inputs compared to an existing baseline, up to 7X faster. We also assessed the validity of these inputs, maintaining a validation rate above 89%, according to automated and human validators.

The rise of social media has significantly increased the prevalence of cyberbullying (CB), posing serious risks to both mental and physical well-being. Effective detection systems are essential for mitigating its impact. While several machine learning (ML) models have been developed, few incorporate victims' psychological, demographic, and behavioral factors alongside bullying comments to assess severity. In this study, we propose an AI model intregrating user-specific attributes, including psychological factors (self-esteem, anxiety, depression), online behavior (internet usage, disciplinary history), and demographic attributes (race, gender, ethnicity), along with social media comments. Additionally, we introduce a re-labeling technique that categorizes social media comments into three severity levels: Not Bullying, Mild Bullying, and Severe Bullying, considering user-specific factors.Our LSTM model is trained using 146 features, incorporating emotional, topical, and word2vec representations of social media comments as well as user-level attributes and it outperforms existing baseline models, achieving the highest accuracy of 98\% and an F1-score of 0.97. To identify key factors influencing the severity of cyberbullying, we employ explainable AI techniques (SHAP and LIME) to interpret the model's decision-making process. Our findings reveal that, beyond hate comments, victims belonging to specific racial and gender groups are more frequently targeted and exhibit higher incidences of depression, disciplinary issues, and low self-esteem. Additionally, individuals with a prior history of bullying are at a greater risk of becoming victims of cyberbullying.

Deploying large scale language models on edge devices faces inherent challenges such as high computational demands, energy consumption, and potential data privacy risks. This paper introduces the Shakti Small Language Models (SLMs) Shakti-100M, Shakti-250M, and Shakti-500M which target these constraints headon. By combining efficient architectures, quantization techniques, and responsible AI principles, the Shakti series enables on-device intelligence for smartphones, smart appliances, IoT systems, and beyond. We provide comprehensive insights into their design philosophy, training pipelines, and benchmark performance on both general tasks (e.g., MMLU, Hellaswag) and specialized domains (healthcare, finance, and legal). Our findings illustrate that compact models, when carefully engineered and fine-tuned, can meet and often exceed expectations in real-world edge-AI scenarios.

The capabilities of artificial intelligence systems have been advancing to a great extent, but these systems still struggle with failure modes, vulnerabilities, and biases. In this paper, we study the current state of the field, and present promising insights and perspectives regarding concerns that challenge the trustworthiness of AI models. In particular, this paper investigates the issues regarding three thrusts: safety, privacy, and bias, which hurt models' trustworthiness. For safety, we discuss safety alignment in the context of large language models, preventing them from generating toxic or harmful content. For bias, we focus on spurious biases that can mislead a network. Lastly, for privacy, we cover membership inference attacks in deep neural networks. The discussions addressed in this paper reflect our own experiments and observations.

Extracting and Understanding the Superficial Knowledge in Alignment

Runjin Chen, Gabriel Jacob Perin, Xuxi Chen +5

Alignment of large language models (LLMs) with human values and preferences, often achieved through fine-tuning based on human feedback, is essential for ensuring safe and responsible AI behaviors. However, the process typically requires substantial data and computation resources. Recent studies have revealed that alignment might be attainable at lower costs through simpler methods, such as in-context learning. This leads to the question: Is alignment predominantly superficial? In this paper, we delve into this question and provide a quantitative analysis. We formalize the concept of superficial knowledge, defining it as knowledge that can be acquired through easily token restyling, without affecting the model's ability to capture underlying causal relationships between tokens. We propose a method to extract and isolate superficial knowledge from aligned models, focusing on the shallow modifications to the final token selection process. By comparing models augmented only with superficial knowledge to fully aligned models, we quantify the superficial portion of alignment. Our findings reveal that while superficial knowledge constitutes a significant portion of alignment, particularly in safety and detoxification tasks, it is not the whole story. Tasks requiring reasoning and contextual understanding still rely on deeper knowledge. Additionally, we demonstrate two practical advantages of isolated superficial knowledge: (1) it can be transferred between models, enabling efficient offsite alignment of larger models using extracted superficial knowledge from smaller models, and (2) it is recoverable, allowing for the restoration of alignment in compromised models without sacrificing performance.

An Explainable Biomedical Foundation Model via Large-Scale Concept-Enhanced Vision-Language Pre-training

Yuxiang Nie, Sunan He, Yequan Bie +14

The clinical adoption of artificial intelligence (AI) in medical imaging requires models that are both diagnostically accurate and interpretable to clinicians. While current multimodal biomedical foundation models prioritize performance, their black-box nature hinders explaining the decision-making process in clinically meaningful concepts. Here, we present ConceptCLIP, the first explainable biomedical foundation model that achieves state-of-the-art diagnostic accuracy while delivering human-interpretable explanations across diverse imaging modalities. We curate MedConcept-23M, the largest pre-training dataset comprising 23 million image-text-concept triplets across diverse medical modalities, where clinical concepts are derived from the Unified Medical Language System. Leveraging this dataset, we develop ConceptCLIP through a novel dual-alignment approach that simultaneously learns global image-text representations and fine-grained region-concept associations for precise and interpretable medical image analysis. We curate the most extensive evaluation benchmark for multimodal biomedical foundation models, covering 52 clinical tasks spanning 10 imaging modalities. Extensive experiments demonstrate that ConceptCLIP outperforms existing state-of-the-art multimodal biomedical foundation models. Importantly, ConceptCLIP demonstrates superior diagnostic performance while providing human-understandable explanations validated by clinical experts. As the first precise and interpretable biomedical foundation model, ConceptCLIP represents a critical milestone toward the widespread clinical adoption of AI, thereby advancing trustworthy AI in medicine.

Skin cancer is one of the most prevalent and potentially life-threatening diseases worldwide, necessitating early and accurate diagnosis to improve patient outcomes. Conventional diagnostic methods, reliant on clinical expertise and histopathological analysis, are often time-intensive, subjective, and prone to variability. To address these limitations, we propose a novel hybrid deep learning framework that integrates convolutional neural networks (CNNs) with Radial Basis Function (RBF) Networks to achieve high classification accuracy and enhanced interpretability. The motivation for incorporating RBF Networks lies in their intrinsic interpretability and localized response to input features, which make them well-suited for tasks requiring transparency and fine-grained decision-making. Unlike traditional deep learning models that rely on global feature representations, RBF Networks allow for mapping segments of images to chosen prototypes, exploiting salient features within a single image. This enables clinicians to trace predictions to specific, interpretable patterns. The framework incorporates segmentation-based feature extraction, active learning for prototype selection, and K-Medoids clustering to focus on these salient features. Evaluations on the ISIC 2016 and ISIC 2017 datasets demonstrate the model's effectiveness, achieving classification accuracies of 83.02\% and 72.15\% using ResNet50, respectively, and outperforming VGG16-based configurations. By generating interpretable explanations for predictions, the framework aligns with clinical workflows, bridging the gap between predictive performance and trustworthiness. This study highlights the potential of hybrid models to deliver actionable insights, advancing the development of reliable AI-assisted diagnostic tools for high-stakes medical applications.