Research Papers

Large language models (LLMs) aligned for safety through techniques like reinforcement learning from human feedback (RLHF) often exhibit emergent deceptive behaviors, where outputs appear compliant but subtly mislead or omit critical information. This paper introduces adversarial activation patching, a novel mechanistic interpretability framework that leverages activation patching as an adversarial tool to induce, detect, and mitigate such deception in transformer-based models. By sourcing activations from "deceptive" prompts and patching them into safe forward passes at specific layers, we simulate vulnerabilities and quantify deception rates. Through toy neural network simulations across multiple scenarios (e.g., 1000 trials per setup), we demonstrate that adversarial patching increases deceptive outputs to 23.9% from a 0% baseline, with layer-specific variations supporting our hypotheses. We propose six hypotheses, including transferability across models, exacerbation in multimodal settings, and scaling effects. An expanded literature review synthesizes over 20 key works in interpretability, deception, and adversarial attacks. Mitigation strategies, such as activation anomaly detection and robust fine-tuning, are detailed, alongside ethical considerations and future research directions. This work advances AI safety by highlighting patching's dual-use potential and provides a roadmap for empirical studies on large-scale models.

Effective AI governance requires structured approaches for stakeholders to access and verify AI system behavior. With the rise of large language models, Natural Language Explanations (NLEs) are now key to articulating model behavior, which necessitates a focused examination of their characteristics and governance implications. We draw on Explainable AI (XAI) literature to create an updated XAI taxonomy, adapted to prompt-based NLEs, across three dimensions: (1) Context, including task, data, audience, and goals; (2) Generation and Presentation, covering generation methods, inputs, interactivity, outputs, and forms; and (3) Evaluation, focusing on content, presentation, and user-centered properties, as well as the setting of the evaluation. This taxonomy provides a framework for researchers, auditors, and policymakers to characterize, design, and enhance NLEs for transparent AI systems.

Causality has gained increasing attention in recent years, notably for improving the interpretability of machine learning models. Yet the field of explainable artificial intelligence (XAI) has been criticized for emphasizing general tendencies rather than the situation-specific facts, which users typically expect as explanations. These expectations align with the notion of actual causes, which identify what made the observed outcome happen, in the specific context at hand. Halpern and Pearl provided a formal basis for actual causation, but identifying actual causes is NP-complete. Practical identification algorithms are extremely scarce, restricted to narrow classes of models, and typically identify only the shortest cause. We address this gap between the formal theory and its applicability through two main contributions. First, we introduce a baseline approximate polynomial-time algorithm with adjustable precision, together with two complementary algorithms that improve its efficiency. Second, we provide a theoretical result showing that the actual-cause identification problem can be decomposed into smaller sub-instances that preserve the set of solutions. This result directly motivates one of the complementary algorithms. Our experiments demonstrate that the baseline method can approximate the set of actual causes, notably for non-Boolean and stochastic models, and that the complementary algorithms further improve its performance in most cases.

Electroencephalography (EEG)-based auditory attention detection (AAD) offers a non-invasive way to enhance hearing aids, but conventional methods rely on too many electrodes, limiting wearability and comfort. This paper presents SHAP-AAD, a two-stage framework that combines DeepSHAP-based channel selection with a lightweight temporal convolutional network (TCN) for efficient AAD using fewer channels.DeepSHAP, an explainable AI technique, is applied to a Convolutional Neural Network (CNN) trained on topographic alpha-power maps to rank channel importance, and the top-k EEG channels are used to train a compact TCN. Experiments on the DTU dataset show that using 32 channels yields comparable accuracy to the full 64-channel setup (79.21% vs. 81.06%) on average. In some cases, even 8 channels can deliver satisfactory accuracy. These results demonstrate the effectiveness of SHAP-AAD in reducing complexity while preserving high detection performance.

Every publicly traded U.S. company files an annual 10-K report containing critical insights into financial health and risk. We propose Tiny eXplainable Risk Assessor (TinyXRA), a lightweight and explainable transformer-based model that automatically assesses company risk from these reports. Unlike prior work that relies solely on the standard deviation of excess returns (adjusted for the Fama-French model), which indiscriminately penalizes both upside and downside risk, TinyXRA incorporates skewness, kurtosis, and the Sortino ratio for more comprehensive risk assessment. We leverage TinyBERT as our encoder to efficiently process lengthy financial documents, coupled with a novel dynamic, attention-based word cloud mechanism that provides intuitive risk visualization while filtering irrelevant terms. This lightweight design ensures scalable deployment across diverse computing environments with real-time processing capabilities for thousands of financial documents which is essential for production systems with constrained computational resources. We employ triplet loss for risk quartile classification, improving over pairwise loss approaches in existing literature by capturing both the direction and magnitude of risk differences. Our TinyXRA achieves state-of-the-art predictive accuracy across seven test years on a dataset spanning 2013-2024, while providing transparent and interpretable risk assessments. We conduct comprehensive ablation studies to evaluate our contributions and assess model explanations both quantitatively by systematically removing highly attended words and sentences, and qualitatively by examining explanation coherence. The paper concludes with findings, practical implications, limitations, and future research directions. Our code is available at https://github.com/Chen-XueWen/TinyXRA.

Explainable AI for Radar Resource Management: Modified LIME in Deep Reinforcement Learning

Ziyang Lu, M. Cenk Gursoy, Chilukuri K. Mohan +1

Deep reinforcement learning has been extensively studied in decision-making processes and has demonstrated superior performance over conventional approaches in various fields, including radar resource management (RRM). However, a notable limitation of neural networks is their ``black box" nature and recent research work has increasingly focused on explainable AI (XAI) techniques to describe the rationale behind neural network decisions. One promising XAI method is local interpretable model-agnostic explanations (LIME). However, the sampling process in LIME ignores the correlations between features. In this paper, we propose a modified LIME approach that integrates deep learning (DL) into the sampling process, which we refer to as DL-LIME. We employ DL-LIME within deep reinforcement learning for radar resource management. Numerical results show that DL-LIME outperforms conventional LIME in terms of both fidelity and task performance, demonstrating superior performance with both metrics. DL-LIME also provides insights on which factors are more important in decision making for radar resource management.

The convergence of Information Technology and Operational Technology has exposed Industrial Control Systems to adaptive, intelligent adversaries that render static defenses obsolete. This paper introduces the Adversarial Resilience Co-evolution (ARC) framework, addressing the "Trinity of Trust" comprising model fidelity, data integrity, and analytical resilience. ARC establishes a co-evolutionary arms race within a Fortified Secure Digital Twin (F-SCDT), where a Deep Reinforcement Learning "Red Agent" autonomously discovers attack paths while an ensemble-based "Blue Agent" is continuously hardened against these threats. Experimental validation on the Tennessee Eastman Process (TEP) and Secure Water Treatment (SWaT) testbeds demonstrates superior performance in detecting novel attacks, with F1-scores improving from 0.65 to 0.89 and detection latency reduced from over 1200 seconds to 210 seconds. A comprehensive ablation study reveals that the co-evolutionary process itself contributes a 27% performance improvement. By integrating Explainable AI and proposing a Federated ARC architecture, this work presents a necessary paradigm shift toward dynamic, self-improving security for critical infrastructure.

We identify a critical vulnerability in autoregressive transformer language models where the em dash token induces recursive semantic drift, leading to clause boundary hallucination and embedding space entanglement. Through formal analysis of token-level perturbations in semantic lattices, we demonstrate that em dash insertion fundamentally alters the model's latent representations, causing compounding errors in long-form generation. We propose a novel solution combining symbolic clause purification via the phi-infinity operator with targeted embedding matrix realignment. Our approach enables total suppression of problematic tokens without requiring model retraining, while preserving semantic coherence through fixed-point convergence guarantees. Experimental validation shows significant improvements in generation consistency and topic maintenance. This work establishes a general framework for identifying and mitigating token-level vulnerabilities in foundation models, with immediate implications for AI safety, model alignment, and robust deployment of large language models in production environments. The methodology extends beyond punctuation to address broader classes of recursive instabilities in neural text generation systems.

Physics-informed neural networks (PINNs) offer a powerful approach to solving partial differential equations (PDEs), which are ubiquitous in the quantitative sciences. Applied to both forward and inverse problems across various scientific domains, PINNs have recently emerged as a valuable tool in the field of scientific machine learning. A key aspect of their training is that the data -- spatio-temporal points sampled from the PDE's input domain -- are readily available. Influence functions, a tool from the field of explainable AI (XAI), approximate the effect of individual training points on the model, enhancing interpretability. In the present work, we explore the application of influence function-based sampling approaches for the training data. Our results indicate that such targeted resampling based on data attribution methods has the potential to enhance prediction accuracy in physics-informed neural networks, demonstrating a practical application of an XAI method in PINN training.

Predictive Process Monitoring (PPM) often uses deep learning models to predict the future behavior of ongoing processes, such as predicting process outcomes. While these models achieve high accuracy, their lack of interpretability undermines user trust and adoption. Explainable AI (XAI) aims to address this challenge by providing the reasoning behind the predictions. However, current evaluations of XAI in PPM focus primarily on functional metrics (such as fidelity), overlooking user-centered aspects such as their effect on task performance and decision-making. This study investigates the effects of explanation styles (feature importance, rule-based, and counterfactual) and perceived AI accuracy (low or high) on decision-making in PPM. We conducted a decision-making experiment, where users were presented with the AI predictions, perceived accuracy levels, and explanations of different styles. Users' decisions were measured both before and after receiving explanations, allowing the assessment of objective metrics (Task Performance and Agreement) and subjective metrics (Decision Confidence). Our findings show that perceived accuracy and explanation style have a significant effect.

Accurate diagnosis of brain disorders such as Alzheimer's disease and brain tumors remains a critical challenge in medical imaging. Conventional methods based on manual MRI analysis are often inefficient and error-prone. To address this, we propose DGG-XNet, a hybrid deep learning model integrating VGG16 and DenseNet121 to enhance feature extraction and classification. DenseNet121 promotes feature reuse and efficient gradient flow through dense connectivity, while VGG16 contributes strong hierarchical spatial representations. Their fusion enables robust multiclass classification of neurological conditions. Grad-CAM is applied to visualize salient regions, enhancing model transparency. Trained on a combined dataset from BraTS 2021 and Kaggle, DGG-XNet achieved a test accuracy of 91.33\%, with precision, recall, and F1-score all exceeding 91\%. These results highlight DGG-XNet's potential as an effective and interpretable tool for computer-aided diagnosis (CAD) of neurodegenerative and oncological brain disorders.

Existing end-to-end sign-language animation systems suffer from low naturalness, limited facial/body expressivity, and no user control. We propose a human-centered, real-time speech-to-sign animation framework that integrates (1) a streaming Conformer encoder with an autoregressive Transformer-MDN decoder for synchronized upper-body and facial motion generation, (2) a transparent, editable JSON intermediate representation empowering deaf users and experts to inspect and modify each sign segment, and (3) a human-in-the-loop optimization loop that refines the model based on user edits and ratings. Deployed on Unity3D, our system achieves a 13 ms average frame-inference time and a 103 ms end-to-end latency on an RTX 4070. Our key contributions include the design of a JSON-centric editing mechanism for fine-grained sign-level personalization and the first application of an MDN-based feedback loop for continuous model adaptation. This combination establishes a generalizable, explainable AI paradigm for user-adaptive, low-latency multimodal systems. In studies with 20 deaf signers and 5 professional interpreters, we observe a +13 point SUS improvement, 6.7 point reduction in cognitive load, and significant gains in naturalness and trust (p $<$ .001) over baselines. This work establishes a scalable, explainable AI paradigm for accessible sign-language technologies.