TorchXAI is a lightweight PyTorch toolkit for evaluating machine learning models using explainability techniques. It offers efficient implementations of explainability metrics that integrate seamlessly with the Captum ecosystem, with a focus on batch computation and task/data-agnostic evaluation to make scalable XAI evaluation easy.

Why TorchXAI?

Designed for XAI evaluation — gives you ready-to-use metrics to quantify explanation quality (for example completeness and other axiomatic metrics).

Captum-compatible — works alongside Captum explainers so you can compute metrics on attributions you already produce.

Batch & scalable — implementations aim to be efficient for dataset-scale evaluation so you can compare explainers across many inputs.

Who is this for?

• ML researchers and engineers who want quantitative ways to evaluate explanation methods.
• People already using Captum who want plug-and-play metrics for large datasets.
• Anyone comparing multiple explainers and needing standardized evaluation metrics.