SLM Meets LLM: Balancing Latency, Interpretability and Consistency in Hallucination Detection

Mengya Hu, Rui Xu, Deren Lei, Yaxi Li, Mingyu Wang +3 more
8/22/2024
cs.CLcs.AIcs.LG

No official code found. Best implementation: microsoft/ConstrainedReasoner (13 stars, MIT, last commit 2024-08-26).

Abstract

Large language models (LLMs) are highly capable but face latency challenges in real-time applications, such as conducting online hallucination detection. To overcome this issue, we propose a novel framework that leverages a small language model (SLM) classifier for initial detection, followed by a LLM as constrained reasoner to generate detailed explanations for detected hallucinated content. This study optimizes the real-time interpretable hallucination detection by introducing effective prompting techniques that align LLM-generated explanations with SLM decisions. Empirical experiment results demonstrate its effectiveness, thereby enhancing the overall user experience.

View on arXivView PDF

Code Implementations(1)

microsoft/ConstrainedReasonerCommunity100%
MIT
132Shell, PythonJul 19, 20242 years agoMIT

Similar Papers

HuDEx: Integrating Hallucination Detection and Explainability for Enhancing the Reliability of LLM responses

91% match2025

Hallucination Detection with the Internal Layers of LLMs

91% match2025

Unsupervised Real-Time Hallucination Detection based on the Internal States of Large Language Models

91% match1 repo652024

InterrogateLLM: Zero-Resource Hallucination Detection in LLM-Generated Answers

91% match1 repo112024

The Two Sides of the Coin: Hallucination Generation and Detection with LLMs as Evaluators for LLMs

90% match2024

Can LLMs Detect Intrinsic Hallucinations in Paraphrasing and Machine Translation?

90% match2025

Cite this paper

@article{hu2024meets,
  title  = {SLM Meets LLM: Balancing Latency, Interpretability and Consistency in Hallucination Detection},
  author = {Mengya Hu and Rui Xu and Deren Lei and Yaxi Li and Mingyu Wang and Emily Ching and Eslam Kamal and Alex Deng},
  year   = {2024},
  eprint = {2408.12748},
  archivePrefix = {arXiv},
  url    = {http://arxiv.org/abs/2408.12748v1}
}

Add a badge to your README

Shows whether official code is linked for this paper.

Code of Paper badge
[![Code of Paper](https://codeofpaper.com/badge/2408.12748.svg)](https://codeofpaper.com/paper/2408.12748)

Discussion