shap

This paper proposes a framework for sentence-level interpretability of rubric-based scoring, comparing SHAP and LLM-generated rationales. It finds that fine-tuned pretrained language models outperform LLMs in prediction accuracy, and SHAP provides more faithful and transferable explanations.

This study develops an XGBoost classifier using SHAP explainability on eight clinical biomarkers from the ADNI dataset to achieve three-class Alzheimer's disease detection (normal cognition, MCI, AD), reaching a macro AUC of 0.982 and Cohen's kappa of 0.909 on the held-out test set. SHAP analysis identifies CDR Global as the dominant predictor for NC and MCI, while CDR-SB and MMSE together drive AD classification.

Proposes a verification-based algorithm to compute provable bounds on exact SHAP values for neural networks, scaling to much larger search spaces than prior exact methods.

This paper presents a machine learning framework using CatBoost and SHAP to predict obstructive coronary artery disease from CT calcium scoring scans, achieving high accuracy by combining calcium-omics and epicardial fat features.

This paper proves that no feature ranking can be simultaneously faithful, stable, and complete under collinearity, characterizing the full attribution design space and providing a formally verified impossibility theorem in explainable AI.

This study applies XGBoost and SHAP analysis to CDC data to identify social determinants driving fentanyl overdose mortality in US counties, highlighting 'silent risk' areas and treatment deserts for early intervention.

This study presents a hybrid predictive framework using CatBoost and SHAP to identify risk factors in tree-involved traffic crashes, highlighting restraint non-use as the most critical predictor of severe injury.

This arXiv preprint introduces GRALIS, a unified mathematical framework using Riesz Representation Theory to formalize and compare linear attribution methods like SHAP, LIME, and Integrated Gradients.

A comparative study evaluating three explainability techniques (Integrated Gradients, Attention Rollout, SHAP) on fine-tuned DistilBERT for sentiment classification, highlighting trade-offs between gradient-based, attention-based, and model-agnostic approaches for LLM interpretability.