A Neuro-Symbolic Benchmark Suite for Concept Quality and Reasoning Shortcuts

TL;DR

This paper introduces rsbench, a benchmark suite for evaluating how well models learn and reason with concepts, addressing the problem of reasoning shortcuts that undermine interpretability and trustworthiness.

Contribution

The paper presents rsbench, a customizable benchmark with metrics and verification procedures to systematically assess concept quality and reasoning shortcuts in neural and neuro-symbolic models.

Findings

High-quality concept learning remains challenging for current models.

RSs significantly affect model reasoning and interpretability.

rsbench provides a systematic way to evaluate and mitigate RSs.

Abstract

The advent of powerful neural classifiers has increased interest in problems that require both learning and reasoning. These problems are critical for understanding important properties of models, such as trustworthiness, generalization, interpretability, and compliance to safety and structural constraints. However, recent research observed that tasks requiring both learning and reasoning on background knowledge often suffer from reasoning shortcuts (RSs): predictors can solve the downstream reasoning task without associating the correct concepts to the high-dimensional data. To address this issue, we introduce rsbench, a comprehensive benchmark suite designed to systematically evaluate the impact of RSs on models by providing easy access to highly customizable tasks affected by RSs. Furthermore, rsbench implements common metrics for evaluating concept quality and introduces novel formal verification procedures for assessing the presence of RSs in learning tasks. Using rsbench, we highlight that obtaining high quality concepts in both purely neural and neuro-symbolic models is a far-from-solved problem. rsbench is available at: https://unitn-sml.github.io/rsbench.