Disentangling Deception and Hallucination Failures in LLMs

2026-02-16

AI Summary

Large language models (LLMs) often conflate knowledge absence with erroneous outputs in factual question answering, failing to distinguish between hallucination—generating false content due to genuine knowledge gaps—and deception—intentionally misrepresenting known facts. This work proposes a knowledge-behavior decoupling framework that isolates these fundamentally distinct failure modes by constructing an entity-centric controlled environment to systematically analyze four behavioral scenarios. Integrating representational disentanglement analysis, sparse interpretability methods, and activation steering at inference time, the study provides the first mechanistic distinction between hallucination and deception within LLMs. Crucially, it demonstrates that behavioral outputs can be successfully modulated without erasing underlying knowledge, thereby validating the separability of these failure modes and offering a novel pathway for understanding and intervening in LLM errors.

Technology Category

• Machine Learning: Large Multimodal Models (LMMs)
• Natural Language Processing: (Large) Language Models
• Cognitive Modeling & Cognitive Systems: Simulating Human Behavior

Application Category

• Semantics and Knowledge: Data modeling to support human-machine intelligence, including LLMs agents, intelligent system behavior, explanations, and user-friendly interactions
• User Modeling, Personalization and Recommendation: Large Language Models (LLM) for user modeling and recommendation
• Search and Retrieval-Augmented AI: Search Tool Learning with LLM: Teaching LLMs to invoke search and make use of retrieved information

Abstract

Failures in large language models (LLMs) are often analyzed from a behavioral perspective, where incorrect outputs in factual question answering are commonly associated with missing knowledge. In this work, focusing on entity-based factual queries, we suggest that such a view may conflate different failure mechanisms, and propose an internal, mechanism-oriented perspective that separates Knowledge Existence from Behavior Expression. Under this formulation, hallucination and deception correspond to two qualitatively different failure modes that may appear similar at the output level but differ in their underlying mechanisms. To study this distinction, we construct a controlled environment for entity-centric factual questions in which knowledge is preserved while behavioral expression is selectively altered, enabling systematic analysis of four behavioral cases. We analyze these failure modes through representation separability, sparse interpretability, and inference-time activation steering.

Problem

• hallucination
• deception
• large language models
• failure modes
• factual question answering

Innovation

• disentanglement
• hallucination
• deception
• mechanism-oriented analysis
• activation steering