From Passive Response to Proactive Correction: Enhancing LLM Robustness Against Input Fact Perturbations

arXiv:2608.25894v1 Announce Type: new Abstract: 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.

Related

• Resonant Context Anchoring: Decoupling Attention Routing and Signal Gain at Inference Time
• Knowing When to Quit: A Principled Framework for Dynamic Abstention in LLM Reasoning
• Language Bias under Conflicting Information in Multilingual LLMs
• Mitigating Context-Memory Conflicts in LLMs through Dynamic Cognitive Reconciliation Decoding

Source: arXiv cs.CL | 2026-08-27