vix.ing · top · new · best · stats · spec

H-Neurons: On the Existence, Impact, and Origin of Hallucination-Associated Neurons in LLMs

2025/12/01 by Cheng Gao, Huimin Chen, Gao, Cheng +9 · 14 voices
Medicine · Computer Science · Engineering · #Schizophrenia research and treatment #Adversarial Robustness in Machine Learning #Ferroelectric and Negative Capacitance Devices

paper · pdf · doi:10.48550/arxiv.2512.01797

Abstract

Large language models (LLMs) frequently generate hallucinations -- plausible but factually incorrect outputs -- undermining their reliability. While prior work has examined hallucinations from macroscopic perspectives such as training data and objectives, the underlying neuron-level mechanisms remain largely unexplored. In this paper, we conduct a systematic investigation into hallucination-associated neurons (H-Neurons) in LLMs from three perspectives: identification, behavioral impact, and origins. Regarding their identification, we demonstrate that a remarkably sparse subset of neurons (less than 0.1% of total neurons) can reliably predict hallucination occurrences, with strong generalization across diverse scenarios. In terms of behavioral impact, controlled interventions reveal that these neurons are causally linked to over-compliance behaviors. Concerning their origins, we trace these neurons back to the pre-trained base models and find that these neurons remain predictive for hallucination detection, indicating they emerge during pre-training. Our findings bridge macroscopic behavioral patterns with microscopic neural mechanisms, offering insights for developing more reliable LLMs.

Cited by

Discussions

Related