Hallucination-Reduction Reasoning Engine
An Epistemically Grounded Reliability Framework for Auditable LLM Reasoning. Large language models don't just hallucinate facts — they fail at the level of claim-evidence alignment. This paper presents HRRE, a framework that addresses that problem through adversarial multi-agent critique, claim-level formalization, and a judge model that evaluates reasoning auditability rather than surface fluency.