Hallucination Across the Reasoning Lifecycle: Interface Visibility, Causal Evidence, and Release Control in Large Reasoning Models

2026-10-04

AI Summary

This study addresses the hallucination problem arising from the propagation of reasoning errors into subsequent decisions and memory. Based on a large-scale review of 312 publications, this work proposes the UIPCA taxonomy framework, which integrates memory management, tool invocation, and training feedback techniques, with a focus on evidence visibility, causal verification, and release control. The primary contribution lies in systematically linking diagnosis, verification, repair, and persistent state control for the first time, thereby establishing a closed-loop system spanning from diagnosis to remediation. Furthermore, this research reveals the limitations of existing intervention strategies and formulates evidence-based selective release and continuous state control approaches.

Abstract

Reasoning errors can propagate into later decisions and memory. This survey synthesizes 312 papers and first-party reports on text-based reasoning hallucinations around three questions: what evidence is observable, what study designs establish, and which corrective actions the evidence supports. UIPCA records unsupported premises (U), invalid inferences (I), dependent reuse (P), visible answer-trace consistency (C), and action-policy failures (A). Across 58 reviewed sources, no comparison establishes that a specified intervention improves reasoning while reducing factual reliability under matched conditions. The synthesis connects diagnosis to verification, repair, selective release, and persistent-state control across memory, tools, and training feedback.

Problem

• Reasoning Hallucination
• Large Reasoning Models
• Error Propagation
• Factual Reliability
• Persistent-state Control

Innovation

• Reasoning Hallucination
• UIPCA Framework
• Large Reasoning Models
• Release Control
• Persistent-state Control