通过消除错误信念提升大模型推理可靠性
Rectifying Belief Space via Unlearning to Harness LLMs' Reasoning
- 用双向搜索定位导致错误的答案背后的错误信念
- 通过去学习机制抑制错误信念，强化正确信念
- 在多个数据集上提升推理准确率且不损害整体性能
大语言模型虽具较强推理能力，但仍常生成错误答案。我们假设这类错误源于虚假信念——模型内部认为正确但实际错误的命题。为此，提出一种修正信念空间的方法：在保持真实信念的同时，抑制虚假信念，从而实现更可靠的推理。首先利用前向-后向束搜索（FBBS） prompting 模型生成解释，识别出导致错误或正确答案的信念；随后应用去学习技术，压制虚假信念并增强真实信念，有效修正模型的信念空间。在多个问答数据集和大模型上的实验表明，该方法可纠正先前错误回答的问题，且不影响整体性能。此外，该方法在未见数据上表现出更强泛化能力，说明修正模型信念空间是降低错误、提升可靠性的可行方向。
原文摘要 · Abstract (English)
Large language models (LLMs) can exhibit advanced reasoning yet still generate incorrect answers. We hypothesize that such errors frequently stem from spurious beliefs, propositions the model internally considers true but are incorrect. To address this, we propose a method to rectify the belief space by suppressing these spurious beliefs while simultaneously enhancing true ones, thereby enabling more reliable inferences. Our approach first identifies the beliefs that lead to incorrect or correct answers by prompting the model to generate textual explanations, using our Forward-Backward Beam Search (FBBS). We then apply unlearning to suppress the identified spurious beliefs and enhance the true ones, effectively rectifying the model's belief space. Empirical results on multiple QA datasets and LLMs show that our method corrects previously misanswered questions without harming overall model performance. Furthermore, our approach yields improved generalization on unseen data, suggesting that rectifying a model's belief space is a promising direction for mitigating errors and enhancing overall reliability.
Thank you to arXiv for use of its open access interoperability. PaperDance 不是 arXiv 官方产品；中文卡片由大模型生成，请以原文为准。