Getting Sick After Seeing a Doctor? Diagnosing and Mitigating Knowledge Conflicts in Event Temporal Reasoning

TL;DR

This paper addresses knowledge conflicts in event temporal reasoning by detecting biases and proposing a counterfactual data augmentation method to improve model accuracy and reduce hallucinations.

Contribution

It introduces bias indicators for detecting conflicts and a CDA approach to mitigate knowledge conflicts in event temporal reasoning models.

Findings

Models suffer from knowledge conflicts affecting accuracy.

CDA reduces hallucinations and improves reasoning performance.

Bias indicators effectively identify conflicting examples.

Abstract

Event temporal reasoning aims at identifying the temporal relations between two or more events from narratives. However, knowledge conflicts arise when there is a mismatch between the actual temporal relations of events in the context and the prior knowledge or biases learned by the model. In this paper, we propose to detect knowledge-conflict examples in event temporal reasoning using bias indicators, which include event relation prior bias, tense bias, narrative bias, and dependency bias. We define conflict examples as those where event relations are opposite to biased or prior relations. To mitigate event-related knowledge conflicts, we introduce a Counterfactual Data Augmentation (CDA) based method that can be applied to both Pre-trained Language Models (PLMs) and Large Language Models (LLMs) either as additional training data or demonstrations for In-Context Learning. Experiments suggest both PLMs and LLMs suffer from knowledge conflicts in event temporal reasoning, and CDA has the potential for reducing hallucination and improving model performance.