Search papers, labs, and topics across Lattice.
This paper introduces DEDUCE, a three-stage framework designed to enhance the robustness of large language models (LLMs) against fact perturbations in user inputs. By detecting errors, devising correction strategies, and correcting misconceptions, DEDUCE transforms LLMs from passive responders into proactive error correctors. Experiments show that this approach significantly improves accuracy and error correction across various LLM families, validated using the newly introduced MisFactQA dataset and other benchmarks.
LLMs can now proactively correct user input errors, significantly boosting their accuracy and reliability in real-world applications.
Large language models (LLMs) frequently produce confident yet factually incorrect responses when user inputs contain misleading premises, a phenomenon we attribute to fact perturbations in the input. Existing approaches to hallucination mitigation typically assume reliable user inputs, overlooking how such factual errors can actively mislead model reasoning. To address this vulnerability, we propose DEDUCE, a three-stage framework that transforms LLMs from passive responders into proactive error correctors. DEDUCE operates in three stages: (1) detect errors through fine-grained fact extraction and verification; (2) devise correction strategies via multi perspective deliberation; and (3) correct misconceptions while delivering reliable answers. We also present MisFactQA, a dataset containing factual errors of varying degrees, and propose new metrics for evaluating model robustness. Experiments on TruthfulQA, FalseQA, and our MisFactQA benchmark demonstrate that DEDUCE significantly improves both accuracy and error correction capability. Consistent gains across Qwen, LLaMA, and Gemma families confirm its effectiveness and scalability.
