MiSS: A Logic-Driven Explanation of Minimal Sufficient Coalitions for Point Cloud Classifiers

2026-07-27

AI Summary

This work addresses the problem of generating efficient, accurate, and minimal interpretable explanations for black-box 3D point cloud classifiers. To this end, it proposes a perturbation-based sufficiency reasoning framework grounded in superpoint abstraction, which decouples candidate generation from black-box verification. For the first time, the approach integrates a weighted MaxSAT solver with a black-box statistical oracle to certify minimally sufficient coalitions of geometric regions without requiring white-box access. The method further introduces an adaptive cardinality lower bound, a provably tight upper bound, and a query-history-informed learning heuristic to substantially accelerate the search process. Experiments on ModelNet40 and ShapeNet with PointNet and PointMLP demonstrate that the proposed technique outperforms baseline methods in both explanation accuracy and coverage while achieving significantly lower computational cost compared to exhaustive search.

Abstract

We present MiSS, a black-box, query-based framework for explaining 3D point cloud classifiers through perturbation-relative sufficiency reasoning. MiSS treats a superpoint partition as an interpretable abstraction layer and asks whether the original prediction can be certified from a minimal coalition of geometric regions under a specified perturbation distribution. Unlike abductive explainers that require Boolean feature spaces or white-box logical encodings of the predictor, MiSS separates candidate proposal from verification: a weighted MaxSAT procedure proposes coalitions using a heuristic adaptive cardinality floor, certified exact-size fallback, a safely tightened upper bound, blocking clauses, and a surrogate acquisition heuristic learned from previous oracle evaluations, while a blackbox statistical oracle decides sufficiency from prediction queries. The system returns a statistically verified sufficient coalition as a binary attribution, with minimum cardinality guaranteed when certified search completes. Experiments on ModelNet40 and ShapeNet with PointNet and PointMLP classifiers show higher precision and coverage than rule-based baselines in most settings, with lower explanation time than exhaustive search.

Problem

• point cloud classification
• sufficient coalition
• explanation
• perturbation
• minimal explanation

Innovation

• perturbation-relative sufficiency
• superpoint abstraction
• weighted MaxSAT
• black-box explanation
• minimal sufficient coalition