Researchers Evaluate Explanations for IRS Risk
Researchers and IRS collaborators empirically evaluate explainability techniques using real, line-by-line IRS audits in a peer-reviewed article accepted 02/18/2026. They decompose aggregate tax under-reporting into constituent line-item misreporting, compare local explanation models with direct constituent-risk estimation and rule-based heuristics, and find local explanations depend on model quality and often fail to recover true risks. The findings advise caution using explanations in high-stakes settings.
Scoring Rationale
IRS-backed empirical evaluation drives high score; focus on tax-audit setting may limit broader generalizability despite rigorous methods.
Practice interview problems based on real data
1,500+ SQL & Python problems across 15 industry datasets — the exact type of data you work with.
Try 250 free problems
