Streaming Hallucination Detection in Long Chain-of-Thought Reasoning

2026-01-05

arXiv.org

AI Summary

This work proposes the first streaming hallucination detection framework for long-chain-of-thought (CoT) reasoning, addressing the challenge that hallucinations in such settings are often subtle, propagate across reasoning steps, and are difficult to detect and localize in real time. The approach models hallucination as a dynamically evolving latent state throughout the reasoning process, leveraging step-level judgments as local observations to construct prefix-accumulated signals that trace the global evolution of hallucinatory behavior. By doing so, the method enables real-time, interpretable monitoring of hallucinations in extended CoT sequences and provides fine-grained evidential support for detected anomalies. This significantly enhances both the timeliness and transparency of hallucination detection, offering a principled and practical solution for improving the reliability of complex reasoning systems.

Technology Category

Cognitive Modeling & Cognitive Systems: Conceptual Inference and ReasoningKnowledge Representation and Reasoning: Computational Complexity of ReasoningReasoning under Uncertainty: Sequential Decision Making

Application Category

Search and Retrieval-Augmented AI: Web learning to rank, online learning, and counterfactual learning for rankingEconomics, Online Markets and Human Computation: LLM based quality controls for crowd workSemantics and Knowledge: Data modeling to support human-machine intelligence, including LLMs agents, intelligent system behavior, explanations, and user-friendly interactions

Abstract

Long chain-of-thought (CoT) reasoning improves the performance of large language models, yet hallucinations in such settings often emerge subtly and propagate across reasoning steps. We suggest that hallucination in long CoT reasoning is better understood as an evolving latent state rather than a one-off erroneous event. Accordingly, we treat step-level hallucination judgments as local observations and introduce a cumulative prefix-level hallucination signal that tracks the global evolution of the reasoning state over the entire trajectory. Overall, our approach enables streaming hallucination detection in long CoT reasoning, providing real-time, interpretable evidence.

Problem

• hallucination
• chain-of-thought reasoning
• streaming detection
• large language models
• reasoning trajectory

Innovation

• streaming hallucination detection
• long chain-of-thought reasoning
• latent hallucination state
• cumulative prefix-level signal
• real-time interpretability