Research Papers

Explaining Image Classifiers using Statistical Fault Localization

Youcheng Sun, Hana Chockler, Xiaowei Huang +1

The black-box nature of deep neural networks (DNNs) makes it impossible to understand why a particular output is produced, creating demand for "Explainable AI". In this paper, we show that statistical fault localization (SFL) techniques from software engineering deliver high quality explanations of the outputs of DNNs, where we define an explanation as a minimal subset of features sufficient for making the same decision as for the original input. We present an algorithm and a tool called DeepCover, which synthesizes a ranking of the features of the inputs using SFL and constructs explanations for the decisions of the DNN based on this ranking. We compare explanations produced by DeepCover with those of the state-of-the-art tools GradCAM, LIME, SHAP, RISE and Extremal and show that explanations generated by DeepCover are consistently better across a broad set of experiments. On a benchmark set with known ground truth, DeepCover achieves 76.7% accuracy, which is 6% better than the second best Extremal.

The advancement of natural language processing (NLP) has expanded AI-based text classification in the legal domain. However, accurately classifying legal documents remains challenging due to the complexity of legal texts and subtle differences between legal categories. This study evaluates legal text classification models ranging from traditional machine learning techniques to large language models (LLMs) using ten categories of Korean sexual offense precedents. The results show that fine-tuning small-scale models such as KLUE-BERT on legal data outperforms general-purpose models such as GPT-3.5 and GPT-4.0, as well as traditional machine learning models. KLUE-BERT achieved the highest accuracy of 99.3%, indicating that domain adaptation and fine-tuning can be more important than model size for legal document classification. We further employ explainable AI (XAI) techniques to analyze model predictions and misclassification cases. XAI analysis identifies linguistic features influencing model decisions and limitations in capturing subtle textual cues. Using KICS data, which closely resembles real-world legal case records, we further evaluate the model's generalization capabilities and find that it struggles to interpret implicit contextual cues. These findings highlight the importance of both performance and interpretability in legal AI and demonstrate how XAI can improve transparency in legal text classification. AI-assisted tools can support legal professionals in tasks including document classification, legal information retrieval, and case assessment.

While multimodal deep learning has advanced medical imaging analysis, existing black-box systems \textcolor{black}{may remain confined to isolated tasks, often overlooking} the trust-sensitive nature of clinical diagnosis as a multi-task process. We propose IMT-CXR (Interpretable Multi-task Transformer for Chest X-ray Analysis), a framework that emulates radiologists' diagnostic workflow through three evidence-driven stages: 1) Disease recognition; 2) Attribute characterization (e.g., size, location, severity quantification); 3) Evidence-integrated report generation with traceable decision pathways. The framework employs a unified transformer architecture optimized via medical-domain instruction tuning, sequentially executing four clinical tasks: multi-label disease classification, lesion localization, anatomical segmentation, and radiology report generation. Experimental validation demonstrates competitive performance on ten CXR benchmarks under direct inference and fine-tuning settings. In a blinded evaluation of 160 historical reports from four medical centers, three radiologists rated 66\% of AI-generated reports as comparable to or surpassing original clinical reports in diagnostic clarity, highlighting the framework's translational potential. By establishing traceable diagnostic pathways from anatomical findings to conclusions, this work bridges the gap between AI technical metrics and clinical utility, advancing trustworthy AI systems in medical imaging.

How large language models internally represent high-level behaviors is a core interpretability question with direct relevance to AI safety: it determines what we can detect, audit, or intervene on. Recent work has shown that traits such as evil or sycophancy correspond to linear directions in the internal activations, the so-called persona vectors. Although these vectors are now routinely utilized to inspect and steer model behavior in safety-relevant settings, how these representations are formed during training remains unknown. To address this gap, we trace persona vectors across the pretraining of OLMo-3-7B, finding that persona vectors form remarkably early -- within 0.22% of OLMo-3 pretraining -- and remain effective for steering the fully post-trained instruct models. Although core representations are formed early on, persona vectors continue to refine geometrically and semantically throughout pretraining. We further compare alternative elicitation strategies and find that all yield effective directions, with each strategy surfacing qualitatively distinct facets of the underlying persona. Replicating our analysis on Apertus-8B reveals that our findings transfer qualitatively beyond OLMo-3. Our results establish persona representations as stable features of early pretraining and open a path to studying how training forms, refines, and shapes them.

While Large Language Models (LLMs) have achieved strong performance across many NLP tasks, their opaque internal mechanisms hinder trustworthiness and safe deployment. Existing surveys in explainable AI largely focus on post-hoc explanation methods that interpret trained models through external approximations. In contrast, intrinsic interpretability, which builds transparency directly into model architectures and computations, has recently emerged as a promising alternative. This paper presents a systematic review of the recent advances in intrinsic interpretability for LLMs, categorizing existing approaches into five design paradigms: functional transparency, concept alignment, representational decomposability, explicit modularization, and latent sparsity induction. We further discuss open challenges and outline future research directions in this emerging field. The paper list is available at: https://github.com/PKU-PILLAR-Group/Survey-Intrinsic-Interpretability-of-LLMs.

Understanding Emergent Misalignment via Feature Superposition Geometry

Gouki Minegishi, Hiroki Furuta, Takeshi Kojima +2

Emergent misalignment, where fine-tuning on narrow, non-harmful tasks induces harmful behaviors, poses a key challenge for AI safety in LLMs. Despite growing empirical evidence, its underlying mechanism remains unclear. To uncover the reason behind this phenomenon, we propose a geometric account based on the geometry of feature superposition. Because features are encoded in overlapping representations, fine-tuning that amplifies a target feature also unintentionally strengthens nearby harmful features in accordance with their similarity. We give a simple gradient-level derivation of this effect and empirically test it in multiple LLMs (Gemma-2 2B/9B/27B, LLaMA-3.1 8B, GPT-OSS 20B). Using sparse autoencoders (SAEs), we identify features tied to misalignment-inducing data and to harmful behaviors, and show that they are geometrically closer to each other than features derived from non-inducing data. This trend generalizes across domains (e.g., health, career, legal advice). Finally, we show that a geometry-aware approach, filtering training samples closest to toxic features, reduces misalignment by 34.5%, substantially outperforming random removal and achieving comparable or slightly lower misalignment than LLM-as-a-judge-based filtering. Our study links emergent misalignment to feature superposition, providing a basis for understanding and mitigating this phenomenon.

Beyond Explainable AI (XAI): An Overdue Paradigm Shift and Post-XAI Research Directions

Saleh Afroogh, Syed Ishtiaque Ahmed, Petra Ahrweiler +46

This study provides a cross-disciplinary examination of Explainable Artificial Intelligence (XAI) approaches-focusing on deep neural networks (DNNs) and large language models (LLMs)-and identifies empirical and conceptual limitations in current XAI. We discuss critical symptoms that stem from deeper root causes (i.e., two paradoxes, two conceptual confusions, and five false assumptions). These fundamental problems within the current XAI research field reveal three insights: experimentally, XAI exhibits significant flaws; conceptually, it is paradoxical; and pragmatically, further attempts to reform the paradoxical XAI might exacerbate its confusion-demanding fundamental shifts and new research directions. To move beyond XAI's limitations, we propose a four-pronged synthesized paradigm shift toward reliable and certified AI development. These four components include: verification-focused Interactive AI (IAI) to establish scientific community protocols for certifying AI system performance rather than attempting post-hoc explanations, AI Epistemology for rigorous scientific foundations, User-Sensible AI to create context-aware systems tailored to specific user communities, and Model-Centered Interpretability for faithful technical analysis-together offering comprehensive post-XAI research directions.

Explainable AI: Context-Aware Layer-Wise Integrated Gradients for Explaining Transformer Models

Melkamu Abay Mersha, Jugal Kalita

Transformer models achieve state-of-the-art performance across domains and tasks, yet their deeply layered representations make their predictions difficult to interpret. Existing explainability methods rely on final-layer attributions, capture either local token-level attributions or global attention patterns without unification, and lack context-awareness of inter-token dependencies and structural components. They also fail to capture how relevance evolves across layers and how structural components shape decision-making. To address these limitations, we proposed the \textbf{Context-Aware Layer-wise Integrated Gradients (CA-LIG) Framework}, a unified hierarchical attribution framework that computes layer-wise Integrated Gradients within each Transformer block and fuses these token-level attributions with class-specific attention gradients. This integration yields signed, context-sensitive attribution maps that capture supportive and opposing evidence while tracing the hierarchical flow of relevance through the Transformer layers. We evaluate the CA-LIG Framework across diverse tasks, domains, and transformer model families, including sentiment analysis and long and multi-class document classification with BERT, hate speech detection in a low-resource language setting with XLM-R and AfroLM, and image classification with Masked Autoencoder vision Transformer model. Across all tasks and architectures, CA-LIG provides more faithful attributions, shows stronger sensitivity to contextual dependencies, and produces clearer, more semantically coherent visualizations than established explainability methods. These results indicate that CA-LIG provides a more comprehensive, context-aware, and reliable explanation of Transformer decision-making, advancing both the practical interpretability and conceptual understanding of deep neural models.

In geophysics, hydrocarbon prospect risking involves assessing the risks associated with hydrocarbon exploration by integrating data from various sources. Machine learning-based classifiers trained on tabular data have been recently used to make faster decisions on these prospects. The lack of transparency in the decision-making processes of such models has led to the emergence of explainable AI (XAI). LIME and SHAP are two such examples of these XAI methods which try to generate explanations of a particular decision by ranking the input features in terms of importance. However, explanations of the same scenario generated by these two different explanation strategies have shown to disagree or be different, particularly for complex data. This is because the definitions of "importance" and "relevance" differ for different explanation strategies. Thus, grounding these ranked features using theoretically backed causal ideas of necessity and sufficiency can prove to be a more reliable and robust way to improve the trustworthiness of the concerned explanation strategies.We propose a unified framework to generate counterfactuals as well as quantify necessity and sufficiency and use these to perform a robustness evaluation of the explanations provided by LIME and SHAP on high dimensional structured prospect risking data. This robustness test gives us deeper insights into the models capabilities to handle erronous data and which XAI module works best in pair with which model for our dataset for hydorcarbon indication.

Prior work suggests that language models, while trained on next token prediction, show implicit planning behavior: they may select the next token in preparation to a predicted future token, such as a likely rhyming word, as supported by a prior qualitative study of Claude 3.5 Haiku using a cross-layer transcoder. We propose much simpler techniques for assessing implicit planning in language models. With case studies on rhyme poetry generation and question answering, we demonstrate that our methodology easily scales to many models. Across models, we find that the generated rhyme (e.g. "-ight") or answer to a question ("whale") can be manipulated by steering at the end of the preceding line with a vector, affecting the generation of intermediate tokens leading up to the rhyme or answer word. We show that implicit planning is a universal mechanism, present in smaller models than previously thought, starting from 1B parameters. Our methodology offers a widely applicable direct way to study implicit planning abilities of LLMs. More broadly, understanding planning abilities of language models can inform decisions in AI safety and control.

Explainable Artificial Intelligence (XAI) techniques, such as Gradient-weighted Class Activation Mapping (Grad-CAM), have become indispensable for visualizing the reasoning process of deep neural networks in medical image analysis. Despite their popularity, the faithfulness and reliability of these heatmap-based explanations remain under scrutiny. This study critically investigates whether Grad-CAM truly represents the internal decision-making of deep models trained for lung cancer image classification. Using the publicly available IQ-OTH/NCCD dataset, we evaluate five representative architectures: ResNet-50, ResNet-101, DenseNet-161, EfficientNet-B0, and ViT-Base-Patch16-224, to explore model-dependent variations in Grad-CAM interpretability. We introduce a quantitative evaluation framework that combines localization accuracy, perturbation-based faithfulness, and explanation consistency to assess Grad-CAM reliability across architectures. Experimental findings reveal that while Grad-CAM effectively highlights salient tumor regions in most convolutional networks, its interpretive fidelity significantly degrades for Vision Transformer models due to non-local attention behavior. Furthermore, cross-model comparisons indicate substantial variability in saliency localization, implying that Grad-CAM explanations may not always correspond to the true diagnostic evidence used by the networks. This work exposes critical limitations of current saliency-based XAI approaches in medical imaging and emphasizes the need for model-aware interpretability methods that are both computationally sound and clinically meaningful. Our findings aim to inspire a more cautious and rigorous adoption of visual explanation tools in medical AI, urging the community to rethink what it truly means to "trust" a model's explanation.

Big Data has become central to modern applications in finance, insurance, and cybersecurity, enabling machine learning systems to perform large-scale risk assessments and fraud detection. However, the increasing dependence on automated analytics introduces important concerns about transparency, regulatory compliance, and trust. This paper examines how explainable artificial intelligence (XAI) can be integrated into Big Data analytics pipelines for fraud detection and risk management. We review key Big Data characteristics and survey major analytical tools, including distributed storage systems, streaming platforms, and advanced fraud detection models such as anomaly detectors, graph-based approaches, and ensemble classifiers. We also present a structured review of widely used XAI methods, including LIME, SHAP, counterfactual explanations, and attention mechanisms, and analyze their strengths and limitations when deployed at scale. Based on these findings, we identify key research gaps related to scalability, real-time processing, and explainability for graph and temporal models. To address these challenges, we outline a conceptual framework that integrates scalable Big Data infrastructure with context-aware explanation mechanisms and human feedback. The paper concludes with open research directions in scalable XAI, privacy-aware explanations, and standardized evaluation methods for explainable fraud detection systems.