ExplainReduce : Summarising local explanations via proxies

Lauri Seppäläinen, Mudong Guo, Kai Puolamäki · 2025

Read ExplainReduce : Summarising local explanations via proxies by Lauri Seppäläinen, Mudong Guo, Kai Puolamäki (2025) in the Tower of Babel, a free online library of public-domain works preserved in clean, searchable text. No sign-up. The full text is served as a plain-text file at a predictable URL, where text is present: https://antarctic-labs.com/tower-of-babel/arxiv-2502-10311.txt.

Description

Lauri Seppäläinen, Mudong Guo, and Kai Puolamäki compress thousands of local model explanations into a handful that still tell the truth. Popular explainers describe a black-box model one neighborhood at a time. Reading thousands of such local notes overwhelms any human. The team instead selects a small proxy set of simple models.

Together, the proxies act as a generative global explanation of the closed box. Finding the set is posed as an optimization problem, solved approximately with greedy heuristics. On many problems, as few as five local explanations faithfully mimic the big model. The reduction held its own against other aggregation methods in their tests. ExplainReduce answers a scaling worry in explainable AI: insight need not drown in its own volume. Five good stand-ins can carry the story.

Catalog record

CREATORLauri Seppäläinen, Mudong Guo, Kai Puolamäki
YEAR2025
COLLECTIONPAPERS
FORMATTXT
FILE SIZE59 KB
VERSION1.0
LICENSEhttp://creativecommons.org/licenses/by/4.0/
RIGHTSCC BY
SOURCEarXiv

Source: https://arxiv.org/abs/2502-10311

Browse the full library · About the Tower of Babel