Explainable Knowledge Graph Retrieval-Augmented Generation (KG-RAG) with KG-SMILE

Generative AI, such as Large Language Models (LLMs), has achieved impressive progress but still produces hallucinations and unverifiable claims, limiting reliability in sensitive domains. Retrieval-Augmented Generation (RAG) improves accuracy by grounding outputs in external knowledge, especially in domains like healthcare, where precision is vital. However, RAG remains opaque and essentially a black box, heavily dependent on data quality. We developed a method-agnostic, perturbation-based framework that provides token and component-level interoperability for Graph RAG using SMILE and named it as Knowledge-Graph (KG)-SMILE. By applying controlled perturbations, computing similarities, and training weighted linear surrogates, KG-SMILE identifies the graph entities and relations most influential to generated outputs, thereby making RAG more transparent. We evaluate KG-SMILE using comprehensive attribution metrics, including fidelity, faithfulness, consistency, stability, and accuracy. Our findings show that KG-SMILE produces stable, human-aligned explanations, demonstrating its capacity to balance model effectiveness with interpretability and thereby fostering greater transparency and trust in machine learning technologies.

RAGuard: A Novel Approach for in-context Safe Retrieval Augmented Generation for LLMs

Accuracy and safety are paramount in Offshore Wind (OSW) maintenance, yet conventional Large Language Models (LLMs) often fail when confronted with highly specialised or unexpected scenarios. We introduce RAGuard, an enhanced Retrieval-Augmented Generation (RAG) framework that explicitly integrates safety-critical documents alongside technical manuals.By issuing parallel queries to two indices and allocating separate retrieval budgets for knowledge and safety, RAGuard guarantees both technical depth and safety coverage. We further develop a SafetyClamp extension that fetches a larger candidate pool, "hard-clamping" exact slot guarantees to safety. We evaluate across sparse (BM25), dense (Dense Passage Retrieval) and hybrid retrieval paradigms, measuring Technical Recall@K and Safety Recall@K. Both proposed extensions of RAG show an increase in Safety Recall@K from almost 0\% in RAG to more than 50\% in RAGuard, while maintaining Technical Recall above 60\%. These results demonstrate that RAGuard and SafetyClamp have the potential to establish a new standard for integrating safety assurance into LLM-powered decision support in critical maintenance contexts.

Safer Skin Lesion Classification with Global Class Activation Probability Map Evaluation and SafeML

Recent advancements in skin lesion classification models have significantly improved accuracy, with some models even surpassing dermatologists' diagnostic performance. However, in medical practice, distrust in AI models remains a challenge. Beyond high accuracy, trustworthy, explainable diagnoses are essential. Existing explainability methods have reliability issues, with LIME-based methods suffering from inconsistency, while CAM-based methods failing to consider all classes. To address these limitations, we propose Global Class Activation Probabilistic Map Evaluation, a method that analyses all classes' activation probability maps probabilistically and at a pixel level. By visualizing the diagnostic process in a unified manner, it helps reduce the risk of misdiagnosis. Furthermore, the application of SafeML enhances the detection of false diagnoses and issues warnings to doctors and patients as needed, improving diagnostic reliability and ultimately patient safety. We evaluated our method using the ISIC datasets with MobileNetV2 and Vision Transformers.

Evaluating Fairness and Mitigating Bias in Machine Learning: A Novel Technique using Tensor Data and Bayesian Regression

Fairness is a critical component of Trustworthy AI. In this paper, we focus on Machine Learning (ML) and the performance of model predictions when dealing with skin color. Unlike other sensitive attributes, the nature of skin color differs significantly. In computer vision, skin color is represented as tensor data rather than categorical values or single numerical points. However, much of the research on fairness across sensitive groups has focused on categorical features such as gender and race. This paper introduces a new technique for evaluating fairness in ML for image classification tasks, specifically without the use of annotation. To address the limitations of prior work, we handle tensor data, like skin color, without classifying it rigidly. Instead, we convert it into probability distributions and apply statistical distance measures. This novel approach allows us to capture fine-grained nuances in fairness both within and across what would traditionally be considered distinct groups. Additionally, we propose an innovative training method to mitigate the latent biases present in conventional skin tone categorization. This method leverages color distance estimates calculated through Bayesian regression with polynomial functions, ensuring a more nuanced and equitable treatment of skin color in ML models.

Incorporating Failure of Machine Learning in Dynamic Probabilistic Safety Assurance

Machine Learning (ML) models are increasingly integrated into safety-critical systems, such as autonomous vehicle platooning, to enable real-time decision-making. However, their inherent imperfection introduces a new class of failure: reasoning failures often triggered by distributional shifts between operational and training data. Traditional safety assessment methods, which rely on design artefacts or code, are ill-suited for ML components that learn behaviour from data. SafeML was recently proposed to dynamically detect such shifts and assign confidence levels to the reasoning of ML-based components. Building on this, we introduce a probabilistic safety assurance framework that integrates SafeML with Bayesian Networks (BNs) to model ML failures as part of a broader causal safety analysis. This allows for dynamic safety evaluation and system adaptation under uncertainty. We demonstrate the approach on an simulated automotive platooning system with traffic sign recognition. The findings highlight the potential broader benefits of explicitly modelling ML failures in safety assessment.

Explainability of Large Language Models using SMILE: Statistical Model-agnostic Interpretability with Local Explanations

Large language models like GPT, LLAMA, and Claude have become incredibly powerful at generating text, but they are still black boxes, so it is hard to understand how they decide what to say. That lack of transparency can be problematic, especially in fields where trust and accountability matter. To help with this, we introduce SMILE, a new method that explains how these models respond to different parts of a prompt. SMILE is model-agnostic and works by slightly changing the input, measuring how the output changes, and then highlighting which words had the most impact. Create simple visual heat maps showing which parts of a prompt matter the most. We tested SMILE on several leading LLMs and used metrics such as accuracy, consistency, stability, and fidelity to show that it gives clear and reliable explanations. By making these models easier to understand, SMILE brings us one step closer to making AI more transparent and trustworthy.

Mapping the Mind of an Instruction-based Image Editing using SMILE

Despite recent advancements in Instruct-based Image Editing models for generating high-quality images, they are known as black boxes and a significant barrier to transparency and user trust. To solve this issue, we introduce SMILE (Statistical Model-agnostic Interpretability with Local Explanations), a novel model-agnostic for localized interpretability that provides a visual heatmap to clarify the textual elements' influence on image-generating models. We applied our method to various Instruction-based Image Editing models like Pix2Pix, Image2Image-turbo and Diffusers-Inpaint and showed how our model can improve interpretability and reliability. Also, we use stability, accuracy, fidelity, and consistency metrics to evaluate our method. These findings indicate the exciting potential of model-agnostic interpretability for reliability and trustworthiness in critical applications such as healthcare and autonomous driving while encouraging additional investigation into the significance of interpretability in enhancing dependable image editing models.

Explainability of Point Cloud Neural Networks Using SMILE: Statistical Model-Agnostic Interpretability with Local Explanations

In today's world, the significance of explainable AI (XAI) is growing in robotics and point cloud applications, as the lack of transparency in decision-making can pose considerable safety risks, particularly in autonomous systems. As these technologies are integrated into real-world environments, ensuring that model decisions are interpretable and trustworthy is vital for operational reliability and safety assurance. This study explores the implementation of SMILE, a novel explainability method originally designed for deep neural networks, on point cloud-based models. SMILE builds on LIME by incorporating Empirical Cumulative Distribution Function (ECDF) statistical distances, offering enhanced robustness and interpretability, particularly when the Anderson-Darling distance is used. The approach demonstrates superior performance in terms of fidelity loss, R2 scores, and robustness across various kernel widths, perturbation numbers, and clustering configurations. Moreover, this study introduces a stability analysis for point cloud data using the Jaccard index, establishing a new benchmark and baseline for model stability in this field. The study further identifies dataset biases in the classification of the 'person' category, emphasizing the necessity for more comprehensive datasets in safety-critical applications like autonomous driving and robotics. The results underscore the potential of advanced explainability models and highlight areas for future research, including the application of alternative surrogate models and explainability techniques in point cloud data.

SafeLLM: Domain-Specific Safety Monitoring for Large Language Models: A Case Study of Offshore Wind Maintenance

The Offshore Wind (OSW) industry is experiencing significant expansion, resulting in increased Operations \& Maintenance (O\&M) costs. Intelligent alarm systems offer the prospect of swift detection of component failures and process anomalies, enabling timely and precise interventions that could yield reductions in resource expenditure, as well as scheduled and unscheduled downtime. This paper introduces an innovative approach to tackle this challenge by capitalising on Large Language Models (LLMs). We present a specialised conversational agent that incorporates statistical techniques to calculate distances between sentences for the detection and filtering of hallucinations and unsafe output. This potentially enables improved interpretation of alarm sequences and the generation of safer repair action recommendations by the agent. Preliminary findings are presented with the approach applied to ChatGPT-4 generated test sentences. The limitation of using ChatGPT-4 and the potential for enhancement of this agent through re-training with specialised OSW datasets are discussed.

Scope Compliance Uncertainty Estimate

The zeitgeist of the digital era has been dominated by an expanding integration of Artificial Intelligence~(AI) in a plethora of applications across various domains. With this expansion, however, questions of the safety and reliability of these methods come have become more relevant than ever. Consequently, a run-time ML model safety system has been developed to ensure the model's operation within the intended context, especially in applications whose environments are greatly variable such as Autonomous Vehicles~(AVs). SafeML is a model-agnostic approach for performing such monitoring, using distance measures based on statistical testing of the training and operational datasets; comparing them to a predetermined threshold, returning a binary value whether the model should be trusted in the context of the observed data or be deemed unreliable. Although a systematic framework exists for this approach, its performance is hindered by: (1) a dependency on a number of design parameters that directly affect the selection of a safety threshold and therefore likely affect its robustness, (2) an inherent assumption of certain distributions for the training and operational sets, as well as (3) a high computational complexity for relatively large sets. This work addresses these limitations by changing the binary decision to a continuous metric. Furthermore, all data distribution assumptions are made obsolete by implementing non-parametric approaches, and the computational speed increased by introducing a new distance measure based on the Empirical Characteristics Functions~(ECF).