Logo
ScienceStack
Table of Contents
Fetching ...
Sign In

ABE: A Unified Framework for Robust and Faithful Attribution-Based Explainability

Zhiyu Zhu, Jiayu Zhang, Zhibo Jin, Fang Chen, Jianlong Zhou

TL;DR

ABE delivers a principled, modular framework for attribution-based explainability by unifying Fundamental Attribution Methods under attribution axioms and embedding a Robustness Module to evaluate explanations under adversarial perturbations. It supports diverse tasks, including unimodal and multimodal settings, through reusable components for model interfacing, attribution generation, and validation. The framework is validated against a wide range of models and datasets, showing that axiom-compliant methods (e.g., ISA, AttEXplore) generally provide more faithful explanations than non-axiomatic baselines, while robustness-aware updates improve stability and trustworthiness. By providing extensible APIs, standardized evaluation metrics, and practical PyTorch-based tooling, ABE aims to accelerate research, comparison, and deployment of reliable, transparent AI systems.

Abstract

Attribution algorithms are essential for enhancing the interpretability and trustworthiness of deep learning models by identifying key features driving model decisions. Existing frameworks, such as InterpretDL and OmniXAI, integrate multiple attribution methods but suffer from scalability limitations, high coupling, theoretical constraints, and lack of user-friendly implementations, hindering neural network transparency and interoperability. To address these challenges, we propose Attribution-Based Explainability (ABE), a unified framework that formalizes Fundamental Attribution Methods and integrates state-of-the-art attribution algorithms while ensuring compliance with attribution axioms. ABE enables researchers to develop novel attribution techniques and enhances interpretability through four customizable modules: Robustness, Interpretability, Validation, and Data & Model. This framework provides a scalable, extensible foundation for advancing attribution-based explainability and fostering transparent AI systems. Our code is available at: https://github.com/LMBTough/ABE-XAI.

ABE: A Unified Framework for Robust and Faithful Attribution-Based Explainability

TL;DR

ABE delivers a principled, modular framework for attribution-based explainability by unifying Fundamental Attribution Methods under attribution axioms and embedding a Robustness Module to evaluate explanations under adversarial perturbations. It supports diverse tasks, including unimodal and multimodal settings, through reusable components for model interfacing, attribution generation, and validation. The framework is validated against a wide range of models and datasets, showing that axiom-compliant methods (e.g., ISA, AttEXplore) generally provide more faithful explanations than non-axiomatic baselines, while robustness-aware updates improve stability and trustworthiness. By providing extensible APIs, standardized evaluation metrics, and practical PyTorch-based tooling, ABE aims to accelerate research, comparison, and deployment of reliable, transparent AI systems.

Abstract

Attribution algorithms are essential for enhancing the interpretability and trustworthiness of deep learning models by identifying key features driving model decisions. Existing frameworks, such as InterpretDL and OmniXAI, integrate multiple attribution methods but suffer from scalability limitations, high coupling, theoretical constraints, and lack of user-friendly implementations, hindering neural network transparency and interoperability. To address these challenges, we propose Attribution-Based Explainability (ABE), a unified framework that formalizes Fundamental Attribution Methods and integrates state-of-the-art attribution algorithms while ensuring compliance with attribution axioms. ABE enables researchers to develop novel attribution techniques and enhances interpretability through four customizable modules: Robustness, Interpretability, Validation, and Data & Model. This framework provides a scalable, extensible foundation for advancing attribution-based explainability and fostering transparent AI systems. Our code is available at: https://github.com/LMBTough/ABE-XAI.
Paper Structure (31 sections, 17 equations, 3 figures, 8 tables, 1 algorithm)

This paper contains 31 sections, 17 equations, 3 figures, 8 tables, 1 algorithm.

Table of Contents

  1. Introduction
  2. Related Work
  3. Traditional Interpretability Methods
  4. Attribution Axiom-Based Interpretability Methods
  5. Robustness and Attribution Methods
  6. Interpretability Frameworks
  7. Method
  8. Problem Definition
  9. Definition of Neural Network
  10. Definition of Interpretability and Attribution
  11. Definition of Attribution Axioms
  12. Definition of Robustness
  13. Module Design
  14. Fundamental Attribution Methods
  15. Proof of Fundamental Attribution Method Satisfying Sensitivity Axiom and Complete Property
  16. ...and 16 more sections

Figures (3)

  • Figure 1: Flowchart of our Framework
  • Figure 2: Flowchart of Fundamental Attribution Methods.
  • Figure 3: Impact of Different Update Methods on Attribution Performance

Theorems & Definitions (1)

  • proof