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

arXiv: https://arxiv.org/abs/2512.01797 · HTML: https://mlsanity.com/paper/2512.01797/

## Key facts

- Cited by 22 papers in the map (data through 14 Oct 2026).
- First posted on arXiv 1 Dec 2025.

## Built on by (most cited first, 22 of 22)

- [Locate, Steer, and Improve: A Practical Survey of Actionable Mechanistic Interpretability in Large Language Models](https://mlsanity.com/paper/2601.14004/) (2026)
- [Hallucination Cascade: Analyzing Error Propagation in Multi-Agent LLM Systems](https://mlsanity.com/paper/2606.07937/) (2026)
- [Sparse Reward Subsystem in Large Language Models](https://mlsanity.com/paper/2602.00986/) (2026)
- [Steering the Verifiability of Multimodal AI Hallucinations](https://mlsanity.com/paper/2604.06714/) (2026)
- [A Few Neurons Reveal When LLMs Misuse Tools: Sparse Detection and Selective Steering for Reliable Tool Use](https://mlsanity.com/paper/2608.00218/) (2026)
- [KARL: Mitigating Hallucinations in LLMs via Knowledge-Boundary-Aware Reinforcement Learning](https://mlsanity.com/paper/2604.22779/) (2026)
- [How Do LLMs and VLMs Understand Viewpoint Rotation Without Vision? An Interpretability Study](https://mlsanity.com/paper/2604.15294/) (2026)
- [Do Hallucination Neurons Generalize? Evidence from Cross-Domain Transfer in LLMs](https://mlsanity.com/paper/2604.19765/) (2026)
- [Understanding Why Language Models Hallucinate: Testing Reasoning Against Priors](https://mlsanity.com/paper/2607.00447/) (2026)
- [Rewriting or Reweighting? A Geometric Account in Language Models](https://mlsanity.com/paper/2608.01835/) (2026)
- [The Hallucination Signal Is a Mean Shift: Why Simple Probes Suffice](https://mlsanity.com/paper/2608.28930/) (2026)
- [Adaptive Activation Cancellation for Hallucination Mitigation in Large Language Models](https://mlsanity.com/paper/2603.10195/) (2026)
- [Geometric Properties of the Voronoi Tessellation in Latent Semantic Manifolds of Large Language Models](https://mlsanity.com/paper/2604.06767/) (2026)
- [Targeted Interpretable Safety Neuron Enhancement for Multilingual Vision-Language Large Models](https://mlsanity.com/paper/2604.08881/) (2026)
- [CausalGaze: Unveiling Hallucinations via Counterfactual Graph Intervention in Large Language Models](https://mlsanity.com/paper/2604.11087/) (2026)
- [Where Fake Citations Are Made: Tracing Field-Level Hallucination to Specific Neurons in LLMs](https://mlsanity.com/paper/2604.18880/) (2026)
- [Max-pooling Network Revisited: Analyzing the Role of Semantic Probability in Multiple Instance Learning for Hallucination Detection](https://mlsanity.com/paper/2605.08863/) (2026)
- [Readable but Not Controllable: Neuron-Level Evidence for Medical LLM Hallucination](https://mlsanity.com/paper/2607.00158/) (2026)
- [One Anchor for All: Unified Multilingual and Multimodal Safety Alignment for LVLMs](https://mlsanity.com/paper/2607.27917/) (2026)
- [NeuroActiSep: Detecting Factual Hallucinations from Feed-Forward Neurons in a Single Pass](https://mlsanity.com/paper/2609.14448/) (2026)
- [Hallucination Neurons and Where to Find Them: An Investigation into the existence of Hallucination Neurons](https://mlsanity.com/paper/2609.29781/) (2026)
- [RESCUE: Repairing Language Model Errors to Sparse Circuits via Reinforcement Learning](https://mlsanity.com/paper/2609.36813/) (2026)

## Abstract (arXiv)

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.