Unbiased Reasoning for Knowledge-Intensive Tasks in Large Language Models via Conditional Front-Door Adjustment

TL;DR

This paper introduces Conditional Front-Door Prompting, a causal framework that reduces internal bias in large language models for knowledge-intensive tasks, leading to improved accuracy and robustness.

Contribution

It proposes a novel causal prompting method that estimates unbiased query-answer relationships by constructing counterfactual external knowledge, improving reasoning in LLMs.

Findings

CFD-Prompting outperforms existing methods in accuracy.

It enhances robustness across multiple datasets.

The approach reduces internal bias in LLM reasoning.

Abstract

Large Language Models (LLMs) have shown impressive capabilities in natural language processing but still struggle to perform well on knowledge-intensive tasks that require deep reasoning and the integration of external knowledge. Although methods such as Retrieval-Augmented Generation (RAG) and Chain-of-Thought (CoT) have been proposed to enhance LLMs with external knowledge, they still suffer from internal bias in LLMs, which often leads to incorrect answers. In this paper, we propose a novel causal prompting framework, Conditional Front-Door Prompting (CFD-Prompting), which enables the unbiased estimation of the causal effect between the query and the answer, conditional on external knowledge, while mitigating internal bias. By constructing counterfactual external knowledge, our framework simulates how the query behaves under varying contexts, addressing the challenge that the query is fixed and is not amenable to direct causal intervention. Compared to the standard front-door adjustment, the conditional variant operates under weaker assumptions, enhancing both robustness and generalisability of the reasoning process. Extensive experiments across multiple LLMs and benchmark datasets demonstrate that CFD-Prompting significantly outperforms existing baselines in both accuracy and robustness.