LLM-CAS: Dynamic Neuron Perturbation for Real-Time Hallucination Correction

TL;DR

LLM-CAS introduces a hierarchical reinforcement learning framework that dynamically perturbs neurons during inference to correct hallucinations in large language models, significantly enhancing factual accuracy without permanent model changes.

Contribution

This work presents a novel, adaptive neuron perturbation method using reinforcement learning for real-time hallucination correction in LLMs, outperforming static and heuristic dynamic approaches.

Findings

Achieves nearly 11% improvement on StoryCloze accuracy

Gains over 2.7 points on TriviaQA accuracy

Outperforms prior static and dynamic correction methods

Abstract

Large language models (LLMs) often generate hallucinated content that lacks factual or contextual grounding, limiting their reliability in critical applications. Existing approaches such as supervised fine-tuning and reinforcement learning from human feedback are data intensive and computationally expensive, while static parameter editing methods struggle with context dependent errors and catastrophic forgetting. We propose LLM-CAS, a framework that formulates real-time hallucination correction as a hierarchical reinforcement learning problem. LLM-CAS trains an agent to learn a policy that dynamically selects temporary neuron perturbations during inference based on the current context. Unlike prior dynamic approaches that rely on heuristic or predefined adjustments, this policy driven mechanism enables adaptive and fine grained correction without permanent parameter modification. Experiments across multiple language models demonstrate that LLM-CAS consistently improves factual accuracy, achieving gains of 10.98 percentage points on StoryCloze, 2.71 points on TriviaQA, and 2.06 points on the MC1 score of TruthfulQA. These results outperform both static editing methods such as ITI and CAA and the dynamic SADI framework. Overall, LLM-CAS provides an efficient and context aware solution for improving the reliability of LLMs, with promising potential for future multimodal extensions.