Benchmark analysis of black-box local explanation methods

Benchmark analysis of black-box local explanation methods

The publication (Bodria et al., 2023) provides a comprehensive benchmarking and survey of explanation methods for black-box models, categorizing them according to the type of explanation returned and the supported input data formats. It qualitatively contrasts the visual appearance of explanations produced by the most representative explainers and quantitatively benchmarks a subset of the most robust and widely adopted ones across a repertoire of metrics. A continuously updated companion website collects new explainers as they appear, offering a compass for researchers and practitioners navigating the XAI landscape.

Reference

Bodria, F., Giannotti, F., Guidotti, R., Naretto, F., Pedreschi, D., & Rinzivillo, S. (2023). Benchmarking and survey of explanation methods for black box models. Data Mining and Knowledge Discovery, 37(5), 1719–1778. https://doi.org/10.1007/s10618-023-00933-9