H-Neurons: On the Existence, Impact, and Origin of Hallucination-Associated Neurons in LLMs
Authors:
Preprint
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.
Keywords
All Related Versions
This publication has 20 references indexed in Scilit:
- "Do Anything Now": Characterizing and Evaluating In-The-Wild Jailbreak Prompts on Large Language Models2024
- Detecting hallucinations in large language models using semantic entropyNature, 2024
- Calibrated Language Models Must Hallucinate2024
- Survey of Hallucination in Natural Language GenerationACM Computing Surveys, 2023
- Neurons learn by predicting future activityNature Machine Intelligence, 2022
- Cell cycle control in cancerNature Reviews Molecular Cell Biology, 2021
- Natural Questions: A Benchmark for Question Answering ResearchTransactions of the Association for Computational Linguistics, 2019
- Memory formation depends on both synapse-specific modifications of synaptic strength and cell-specific increases in excitabilityNature Neuroscience, 2018
- An overview of the BIOASQ large-scale biomedical semantic indexing and question answering competitionBMC Bioinformatics, 2015
- Synaptic Theory of Working MemoryScience, 2008