Commonsense Knowledge with Negation: A Resource to Enhance Negation Understanding

2026-04-21

AI Summary

This work addresses the well-documented weakness of large language models in handling negation and the pervasive lack of negated scenarios in existing commonsense knowledge bases. To bridge this gap, the study presents the first systematic effort to integrate negation into commonsense knowledge representation. It introduces an automated method to generate over two million negation-augmented if-then commonsense triples from existing corpora, resulting in two large-scale negated commonsense knowledge bases. By incorporating these resources into the pretraining of large language models, the approach effectively injects negation-aware commonsense knowledge, leading to significant performance gains across multiple negation understanding benchmarks. This contribution fills a critical void in both resources and methodologies for modeling negated commonsense reasoning.

Technology Category

• Knowledge Representation and Reasoning: Common-Sense Reasoning
• Cognitive Modeling & Cognitive Systems: Conceptual Inference and Reasoning
• Reasoning under Uncertainty: Other Foundations of Reasoning under Uncertainty

Application Category

• Semantics and Knowledge: Methods to enhance, augment, integrate or synergize semantic models such as knowledge graphs and LLMs
• Search and Retrieval-Augmented AI: Large language models for search
• Graph Algorithms and Modeling for the Web: Foundation models and LLMs for Web-related graphs

Abstract

Negation is a common and important semantic feature in natural language, yet Large Language Models (LLMs) struggle when negation is involved in natural language understanding tasks. Commonsense knowledge, on the other hand, despite being a well-studied topic, lacks investigations involving negation. In this work, we show that commonsense knowledge with negation is challenging for models to understand. We present a novel approach to automatically augment existing commonsense knowledge corpora with negation, yielding two new corpora containing over 2M triples with if-then relations. In addition, pre-training LLMs on our corpora benefits negation understanding.

Problem

• negation
• commonsense knowledge
• large language models
• natural language understanding

Innovation

• negation
• commonsense knowledge
• knowledge augmentation
• large language models
• if-then relations