Research Papers

Finetuning has become the gold standard for refining existing behaviors and inducing new ones in language models, yet it often remains unclear exactly which behaviors emerge during this process. As models grow ever more capable, understanding finetuning better becomes increasingly important, particularly since unwanted behaviors may arise during finetuning. In this paper, we introduce Diff Mining, a simple yet effective framework for identifying what a finetuned model has learned by comparing its logits to those of its base model. Diff Mining effectively surfaces salient tokens that are amplified in the finetuned model, serving as a fingerprint of its training -- even on text unrelated to the finetuning domain. Unlike many existing model diffing methods which require model internals, Diff Mining only needs access to output logits and scales to large models. The framework consists of two modular stages: (i) extracting per-context logit differences between the finetuned and base models on a reference corpus, and (ii) aggregating the resulting signals to construct an interpretable token set representing the finetune. For aggregation, we explore both a simple Top-K frequency method and a Non-negative Matrix Factorization (NMF)-based approach for disentangling multiple finetuning objectives into distinct token clusters. Empirically, Diff Mining succeeds across diverse settings: on finetune domain detection, it significantly outperforms state-of-the-art model diffing methods both in identifying relevant tokens and in downstream performance when an interpretability agent is given access to the extracted token set; on models with injected biases, it identifies more than one third of the biases without targeted probing. Overall, our framework shows promise in developing auditing tools to detect finetuning objectives.

Postpartum depression (PPD) poses a major burden on maternal and child health, especially in low- and middle-income countries where prevalence exceeds 19%. Despite advancements in machine learning for PPD prediction, current approaches are limited by opaque global explanations that lack clinical usefulness at the patient level, unstable feature selection, and poor generalization under class imbalance. We propose SAGE, a Stability-Aware Graph-Based Ensemble feature selection system that incorporates both local explainable AI and a genetically optimized artificial neural network (GA-ANN). Using a primary cohort of 766 postpartum women, SAGE combines information-theoretic relevance, PCA-based structure, and graph-based interactions with bootstrap stability weighting to identify robust and non-redundant predictors. The GA-ANN architecture, optimized using a genetic algorithm and enhanced with GAN based oversampling, achieved strong performance with 87.96% accuracy, 86.32% F1 score, and 0.88 AUC using only 16 features, outperforming baseline and other feature selection methods. Psychological and socioeconomic factors such as EPDS score, PHQ-9 score, feelings about motherhood, and abuse history are the main predictors, while demographic factors have less influence. The LIME-based explanations allow instance-based insight into selected features from the graph, enabling personalized risk assessment. The findings make SAGE a scalable, interpretable, and clinical tool for early identification of PPD in health-care limited resources.

Explainable AI (XAI) research has produced a plethora of explanation techniques, yet user studies repeatedly show that available explanations are not effective in practice. We argue that, given the siloed nature of conventional XAI, users are struggling to select the appropriate XAI technique. Viewing XAI through a philosophical lens, we offer a formalization of what we call the selection problem: the systematic failure of XAI interfaces to bridge the gap between a user's natural-language uncertainty and the explanation technique that resolves it. Following a logical premise-conclusion format, we show that conventional interfaces require users to translate their uncertainty into a technique selection, a challenging prerequisite to meet. We also propose a structural solution: a multi-agent LLM orchestration tool that translates the user's query to the proper XAI explanation technique. We provide an example of how this structural solution could be instantiated to address the selection problem.

Financial fraud detection relies heavily on centralized machine learning models. This creates serious data privacy risks. Federated Learning (FL) decentralizes data processing, but financial regulations still require models to be transparent. This means using Explainable AI (XAI) tools such as TreeSHAP. Recent cybersecurity research shows a problem with this approach. Sharing high-fidelity SHAP explanations exposes the federated network to Membership Inference Attacks (MIAs). This dissertation proposes and evaluates DP-FedSHAP. It is a new architecture that applies client-level differential privacy only to post-hoc TreeSHAP vectors. It is compared against a Weight-Level DP baseline, which perturbs the trained model directly instead. Using the highly imbalanced IEEE-CIS Fraud Detection dataset, this study measures the trade-off between explanation fidelity, privacy preservation, and the model's Area Under the Precision-Recall Curve (AUPRC).

Multimodal neuroimaging combining structural MRI and positron emission tomography (PET) captures complementary structure-function relationships across the Alzheimer's disease (AD) continuum, yet existing artificial intelligence systems produce a single diagnostic label without quantifying which imaging modality drove that decision for a specific patient. We introduce the Modality Contribution Network (MCNet) and the Modality Contribution Score (MCS), the first per-patient attribution framework quantifying the shift in modality dominance from structural atrophy to amyloid and metabolic dysfunction across the cognitively normal to MCI to AD continuum. MCS is normalised to unity per subject via modality ablation (MCS_MRI_i + MCS_PET_i = 1.0 for every subject i), providing an interpretable, clinically actionable score that fluid biomarkers cannot supply. Applied to 327 ADNI-3 participants balanced across cognitively normal, mild cognitive impairment, and AD groups, MCNet achieved competitive three-class staging performance (AUC=0.881). The MCS revealed a statistically significant monotonic gradient (Kruskal-Wallis p<0.0001), with increasing PET dominance from cognitively normal (MCS_PET 0.412+/-0.229) through MCI (0.489+/-0.289) to AD (0.671+/-0.426), validated against amyloid SUVR (r=0.172, p=0.006) and FDG metabolic biomarkers (r=-0.287, p=0.0005) from separate imaging pipelines. External replication in 1,073 independent OASIS-3 subjects confirmed cross-cohort generalisability (H=166.99, p<0.0001, eta^2=0.156). A mechanistic comparison with SHAP demonstrated that ablation-based MCS captures clinically meaningful modality dependence that deviation-based methods cannot. These findings position MCNet as a foundation for personalised imaging decisions, clinical trial stratification, and trustworthy AI in dementia care.

Accurate field plant disease diagnosis requires reliable fusion of uncertain and conflicting perceptual evidence. We present the Hybrid Hierarchical Multi-Agent Framework (H$^{2}$MAF), combining decision-level fusion of EfficientNet-B3 and ConvNeXt-Tiny with semantic arbitration by open-weight multimodal large language models (MLLMs), Gemma 4 E4B and Qwen3.5 4B, using structured JSON evidence to generate explainable diagnoses, risk levels, treatment urgency, and financial exposure. (H$^{2}$MAF) is evaluated on 14,364 images (1,370 test images) across PlantDoc (2,922 images, 27 classes) and two non-public, continuously captured Cornell robot-acquired field datasets: Stage 2 (20 GB; 4,215 images) and Stage 4 (40 GB; 7,227 images), covering Early Blight, Late Blight, and Septoria Leaf Spot under uncontrolled field conditions. On PlantDoc, Gemma improves accuracy from 63.9% to 68.5%, achieving +7.6 points on the 41.7% CNN-conflict subset. Cornell accuracies reach 99.3% and 98.9%, with only 1.7-4.1% disagreement, demonstrating conflict-dependent MLLM utility. The critical-risk error of gemma is 0.14-0.5 points, whereas Qwen overflags by 3.5-14.4 points. These results establish MLLM arbitration as a promising, yet calibration-dependent, approach for explainable agricultural AI and robotic field decision support. Github Link: https://github.com/Applied-AI-Research-Lab/Explainable-AI-Plant-Disease-Detection

As machine learning (ML) models are increasingly deployed in high-stakes environments, explainable AI (XAI) methods like SHAP and LIME have become essential for regulatory compliance and trust. However, the current auditing paradigm relies on an implicit "chain of trust" where third-party auditors are assumed to be trusted. Recent research demonstrates that this assumption is flawed and adversarial auditors can manipulate XAI explanations through manipulation attacks such as output shuffling or scaffolding out-of-distribution (OOD) to conceal model biases while maintaining high prediction accuracy aiming for fairwashed explanation. In this paper, we introduce a novel defense framework, ExplainGuard, that leverages a Zero-Trust architecture (ZTA) design to be incorporated within the XAI explanation supply chain and ensures the integrity of the generated explanation. This framework would help us to replace the ambiguous default assumption of "auditor is trustworthy," with a continuous "verify-then-trust" approach. Our design architecture establishes a Policy Decision Point (PDP) that enforces three distinct pillars of verification before any explanation is released to the user: (1) asset integrity via behavioral fingerprint to detect model substitution, (2) semantic validity using axiomatic consistency checks to reject mathematically impossible explanations, and (3) feature faithfulness verification utilizing a ranking stability approach with minimal computational overhead. Finally, we evaluate how ExplainGuard can effectively neutralize state- of-the-art explanation manipulation attacks while transforming the auditing process into a verifiable operation.

Artificial intelligence systems increasingly make consequential judgments - which patient is deteriorating, which building is safe to enter, whether an image is authentic and are trusted on the strength of how accurately and confidently they predict. The safeguards that certify them are correspondingly prediction-based: accuracy, calibration and conformal coverage all measure how well a model performs. Whether such checks are sufficient to establish model trustworthiness has remained unclear. Here we prove that they cannot. We establish a separation theorem showing that a reliable model and a compromised one can be identical under every prediction-side certificate, including accuracy, calibration and coverage, yet differ arbitrarily in explanation fidelity and deployment behaviour. Detecting this failure requires access to the model's decision mechanism in addition to its predictions. We introduce the competence envelope as an operational framework that combines prediction and explanation certification into a single deployable criterion. Across diverse datasets and model classes, the proposed framework reveals failure modes that prediction-side certification alone does not capture. Certification against failures that are invisible in prediction behaviour therefore requires evidence about the model's decision mechanism as well as its outputs.

Accurate electricity price forecasting is critical for smart grid stability, yet the heavy reliance on historical price lags in modern predictive models often obscures the fundamental physical drivers of market volatility. This paper proposes a regime-sensitive, explainable artificial intelligence (XAI) framework to unmask the hidden roles of load, climate, and calendar variables in the ERCOT Day-Ahead Market (2014-2024). Utilizing a Histogram-based Gradient Boosting Regressor (HGBR), we introduce a Suppression Ratio to quantify how price lags overshadow physical parameters. Our analysis reveals that while lags provide short-term memory during normal conditions, they fail to capture extreme dynamics; notably, their inclusion increased forecasting error (MAE/RMSE) during price spikes. By isolating these effects, the study revealed that price spikes are not merely the result of continuously rising temperatures, but rather emerge from the system reacting sharply within a specific temperature range. These findings underscore the importance of moving beyond lag-based models to uncover the true drivers of extreme price events.

Autonomous AI diagnostic agents, systems that analyse patient-specific clinical data and produce diagnostic outputs or triage decisions without mandatory real-time human review, are increasingly deployed across eHealth platforms in sub-Saharan Africa at a pace that has outrun the governance infrastructure needed to oversee them. While significant bodies of work address AI accountability, transparency and explainability in healthcare, existing frameworks are largely clinician-centered and assume regulatory conditions that do not uniformly exist in low-resource settings. A patient-centered analysis of the disparity in patient awareness regarding autonomous agents, which results in a structural accountability gap, is mostly missing from the literature. This paper synthesizes existing research on informed consent, algorithmic accountability, and explainable AI to highlight three distinct challenges introduced by deploying AI agents in the sub-Saharan African context. Drawing on three documented deployment cases, including computer-aided tuberculosis detection in Tanzania, diabetic retinopathy and TB screening in Zambia, and mobile health chat-bot triage in Ghana, it demonstrates that these gaps are already present in active deployments across the region. In response, the paper proposes three foundational principles; agent-aware informed consent, human override as a structural requirement and contextually adapted explainability. This triad of principles lays a practical minimum standard for developers, health system administrators and policymakers in contexts where formal AI regulation remains nascent.

Purpose: Nocturnal hypoglycemia (NH) remains a challenge for individuals with type 1 diabetes (T1D), particularly those who are physically active or on multiple daily injections (MDI). We leveraged an explainable evidential neural network model that forecasts minimum overnight glucose to identify NH risk factors and generate recommendations for NH prevention. Methods: We analyzed the impact of glucose and physical activity (PA) factors on predicted NH probability using SHapley Additive exPlanations (SHAP). Data were from 20 adults with T1D on MDI (10 females; mean age 39 years; HbA1c 7\%) who participated in a cross-over study (NCT05967260). Results: A total of 502 nights were analyzed. Bedtime glucose was the strongest predictor of NH. Other relevant factors included recent exposure to high or low glucose, glucose variability before bedtime, and timing of PA. Some associations appeared physiologically counterintuitive, possibly reflecting behavioral adjustments. Based on the identified risk factors and their SHAP values, we determined key decision points and developed recommendations that may help prevent NH, such as consuming a bedtime snack or discussing potential adjustments to insulin therapy with a healthcare provider. Conclusion: Identifying predictors of NH offers insights for managing NH risk in clinical decision support.

Time series arise in a wide range of application domains and are analyzed using machine learning in decision-critical settings. Time series classification (TSC) is one of the most widely studied and relevant tasks. In this context, ensuring the transparency and trustworthiness of TSC models has become an important requirement, motivating the use of explainable artificial intelligence (XAI) methods. Despite growing interest, research on XAI for TSC remains fragmented, and a systematic understanding of the available software frameworks for explanation generation, their evaluation practices, and practical limitations is still lacking. Prior work largely focused on individual explanation methods, while cross-framework consistency, time-series-specific evaluation, and reproducibility have received little attention. In this survey, we analyze existing software frameworks for explanation generation and evaluation in TSC. We compare them along multiple dimensions, including supported XAI methods, evaluation metrics, usability, benchmarking support, and reproducibility, providing the first time-series-specific survey of frameworks with implementation comparisons and an analysis of frequency-domain support. We identify six frameworks that explicitly support time series and reveal common limitations: only one method supports frequency-domain explanations despite their relevance; only two evaluation metrics have been developed specifically for time series; and identical XAI methods can yield substantially different explanations across frameworks. Based on these findings, we discuss open challenges and outline directions for future research, highlighting the need for unified, time-series-specific XAI frameworks that enable faithful, reproducible, and time-series-aware explanations.