Causal Walk: Debiasing Multi-Hop Fact Verification with Front-Door Adjustment
Congzhi Zhang, Linhai Zhang, Deyu Zhou
TL;DR
Causal Walk addresses spurious biases in multi-hop fact verification by applying front-door adjustment with a mediator representing the reasoning path between input claims and evidence. It decomposes the causal effect into P(L|do(G)) = sum_r P(L|do(r)) P(r|do(G)), estimated via a random-walk-based path sampler and a NWGM-based deconfounded predictor, respectively. The approach is validated on FEVER and PolitiHop variants, including adversarial and symmetric versions, where it demonstrates improved robustness and multi-hop reasoning performance compared to prior graph-based and causal-debiasing methods. The work also contributes new adversarial and symmetric datasets generated with large language models to stress-test debiasing under complex bias patterns, with code and data to be released publicly.
Abstract
Conventional multi-hop fact verification models are prone to rely on spurious correlations from the annotation artifacts, leading to an obvious performance decline on unbiased datasets. Among the various debiasing works, the causal inference-based methods become popular by performing theoretically guaranteed debiasing such as casual intervention or counterfactual reasoning. However, existing causal inference-based debiasing methods, which mainly formulate fact verification as a single-hop reasoning task to tackle shallow bias patterns, cannot deal with the complicated bias patterns hidden in multiple hops of evidence. To address the challenge, we propose Causal Walk, a novel method for debiasing multi-hop fact verification from a causal perspective with front-door adjustment. Specifically, in the structural causal model, the reasoning path between the treatment (the input claim-evidence graph) and the outcome (the veracity label) is introduced as the mediator to block the confounder. With the front-door adjustment, the causal effect between the treatment and the outcome is decomposed into the causal effect between the treatment and the mediator, which is estimated by applying the idea of random walk, and the causal effect between the mediator and the outcome, which is estimated with normalized weighted geometric mean approximation. To investigate the effectiveness of the proposed method, an adversarial multi-hop fact verification dataset and a symmetric multi-hop fact verification dataset are proposed with the help of the large language model. Experimental results show that Causal Walk outperforms some previous debiasing methods on both existing datasets and the newly constructed datasets. Code and data will be released at https://github.com/zcccccz/CausalWalk.