Quanda: An Interpretability Toolkit for Training Data Attribution Evaluation and Beyond

TL;DR

This work tackles the lack of standardized evaluation tooling for training data attribution (TDA) methods in deep neural networks. It introduces Quanda, an open-source toolkit that unifies TDA implementations (e.g., SIM, Influence Functions, TracIn, Representer Points, TRAK) and provides a suite of evaluation metrics organized into Ground Truth, Downstream, and Heuristics, plus controlled benchmarks. A modular design with Explainer, Metric, and Benchmark base classes enables on-demand explanations and incremental metric updates; a formal definition of ground-truth LDS uses subsets $ \mathcal{D}' \subset \mathcal{D}$ and a Spearman correlation $\rho$ between predicted and actual outcomes. Empirical results show that Quanda enables systematic, reproducible comparisons and lowers barriers to adopting TDA in practice.

Abstract

In recent years, training data attribution (TDA) methods have emerged as a promising direction for the interpretability of neural networks. While research around TDA is thriving, limited effort has been dedicated to the evaluation of attributions. Similar to the development of evaluation metrics for traditional feature attribution approaches, several standalone metrics have been proposed to evaluate the quality of TDA methods across various contexts. However, the lack of a unified framework that allows for systematic comparison limits trust in TDA methods and stunts their widespread adoption. To address this research gap, we introduce Quanda, a Python toolkit designed to facilitate the evaluation of TDA methods. Beyond offering a comprehensive set of evaluation metrics, Quanda provides a uniform interface for seamless integration with existing TDA implementations across different repositories, thus enabling systematic benchmarking. The toolkit is user-friendly, thoroughly tested, well-documented, and available as an open-source library on PyPi and under https://github.com/dilyabareeva/quanda.

Figures (2)

• Figure 1: Overview of exemplary benchmarking results facilitated by quanda . The figure depicts the training data attribution (TDA) evaluation pipeline. Firstly, each TDA method generates attribution scores for test samples. Subsequently, quanda computes metric scores for each TDA method, assessing various aspects of attribution quality across different evaluation strategies. Each metric provides insight into a specific dimension of the attributor's performance, resulting in a comprehensive analysis of the strengths and weaknesses of each method. For a detailed experimental setup and a discussion of the results, please refer to Appendix \ref{['exp_details']}.
• Figure 2: Illustration of main quanda components. Each block represents a base class. The class fields are listed on the top section, indicated with the icon , while the class methods are listed on the bottom, indicated with . The leading torch is omitted for PyTorch types. The arrows explain the functions of and relations between individual components. Specifically, while Explainer and Metric classes relate indirectly through the generated explanations, the Benchmark class internally utilizes a Metric instance to evaluate the Explainer class, handling the generation of explanations.