Empower Nested Boolean Logic via Self-Supervised Curriculum Learning

TL;DR

This paper investigates whether language models can handle nested boolean logic, finds they struggle without specialized training, and introduces a curriculum learning method to improve their logical reasoning abilities.

Contribution

The paper proposes a novel self-supervised curriculum learning approach, Curriculum Logical Reasoning (CLR), to enhance language models' ability to understand complex nested boolean logic.

Findings

Pre-trained language models perform poorly on multi-nested boolean logic tasks.

CLR significantly improves models' generalization to complex logical reasoning.

Boolean logic training enhances performance on broader logical tasks.

Abstract

Beyond the great cognitive powers showcased by language models, it is crucial to scrutinize whether their reasoning capabilities stem from strong generalization or merely exposure to relevant data. As opposed to constructing increasingly complex logic, this paper probes into the boolean logic, the root capability of a logical reasoner. We find that any pre-trained language models even including large language models only behave like a random selector in the face of multi-nested boolean logic, a task that humans can handle with ease. To empower language models with this fundamental capability, this paper proposes a new self-supervised learning method \textit{Curriculum Logical Reasoning} (\textsc{Clr}), where we augment the training data with nested boolean logic chain step-by-step, and program the training from simpler logical patterns gradually to harder ones. This new training paradigm allows language models to effectively generalize to much harder and longer-hop logic, which can hardly be learned through naive training. Furthermore, we show that boolean logic is a great foundation for improving the subsequent general logical tasks.