A Removal Based Approach to Improve LLM Faithfulness at Test-Time
A Removal Based Approach to Improve LLM Faithfulness at Test-Time
一种基于移除法的测试时大模型忠实度提升方案
Abstract: Large language models (LLMs) are increasingly used for consequential decisions, making their explanations an important tool for auditing model behavior. Unfortunately, these explanations can be unfaithful, failing to reflect the actual reasoning underlying the model’s decisions.
摘要： 大型语言模型（LLMs）正越来越多地被用于关键决策，这使得它们的解释成为审计模型行为的重要工具。遗憾的是，这些解释往往缺乏忠实度，无法反映模型决策背后的真实推理过程。
We consider a setting in which an LLM provides both an answer and an explanation in response to a question. We identify two distinct dimensions of unfaithful explanations: incompleteness, meaning that the explanation omits factors that influence the answer, and unsoundness, meaning that the explanation cites factors that did not influence the model’s answer.
我们考虑了一种场景，即大模型在回答问题时同时提供答案和解释。我们确定了不忠实解释的两个不同维度：不完整性（incompleteness），即解释遗漏了影响答案的因素；以及不合理性（unsoundness），即解释引用了并未影响模型答案的因素。
Existing approaches to improving LLM faithfulness include training-time methods, which require access to model weights and extensive computational resources, and test-time methods that largely focus on addressing unsoundness. We introduce a test-time approach that directly targets incompleteness.
现有的提升大模型忠实度的方法包括训练时方法（需要访问模型权重并消耗大量计算资源）和主要侧重于解决不合理性的测试时方法。我们引入了一种直接针对不完整性的测试时方法。
We remove from the input the concepts not credited in the model’s explanation and re-query the model on the reduced input. This eliminates unmentioned influences while preserving the influence of mentioned concepts.
我们从输入中移除那些未在模型解释中提及的概念，并使用精简后的输入重新查询模型。这消除了未提及的影响因素，同时保留了已提及概念的影响。
Across two datasets, multiple model families, and two independent faithfulness metrics, our approach improves explanation faithfulness compared to both standard prompting and prompting to encourage faithfulness. Our method is model-agnostic and can be applied at inference time without modifying model parameters, providing a flexible mechanism for reducing hidden influences and improving the reliability and safety of LLM-assisted decision making.
在两个数据集、多个模型系列和两个独立的忠实度指标测试中，与标准提示词（standard prompting）和鼓励忠实度的提示词相比，我们的方法均提升了解释的忠实度。我们的方法与模型无关，可以在推理阶段应用而无需修改模型参数，为减少隐藏影响因素以及提高大模型辅助决策的可靠性和安全性提供了一种灵活的机制。