Saved in:
Bibliographic Details
Main Authors: Sankararaman, Hithesh, Yasin, Mohammed Nasheed, Sorensen, Tanner, Di Bari, Alessandro, Stolcke, Andreas
Format: Preprint
Published: 2024
Subjects:
Online Access:https://arxiv.org/abs/2411.01022
Tags: Add Tag
No Tags, Be the first to tag this record!
_version_ 1866929600107380736
author Sankararaman, Hithesh
Yasin, Mohammed Nasheed
Sorensen, Tanner
Di Bari, Alessandro
Stolcke, Andreas
author_facet Sankararaman, Hithesh
Yasin, Mohammed Nasheed
Sorensen, Tanner
Di Bari, Alessandro
Stolcke, Andreas
contents We present a light-weight approach for detecting nonfactual outputs from retrieval-augmented generation (RAG). Given a context and putative output, we compute a factuality score that can be thresholded to yield a binary decision to check the results of LLM-based question-answering, summarization, or other systems. Unlike factuality checkers that themselves rely on LLMs, we use compact, open-source natural language inference (NLI) models that yield a freely accessible solution with low latency and low cost at run-time, and no need for LLM fine-tuning. The approach also enables downstream mitigation and correction of hallucinations, by tracing them back to specific context chunks. Our experiments show high area under the ROC curve (AUC) across a wide range of relevant open source datasets, indicating the effectiveness of our method for fact-checking RAG output.
format Preprint
id arxiv_https___arxiv_org_abs_2411_01022
institution arXiv
publishDate 2024
record_format arxiv
spellingShingle Provenance: A Light-weight Fact-checker for Retrieval Augmented LLM Generation Output
Sankararaman, Hithesh
Yasin, Mohammed Nasheed
Sorensen, Tanner
Di Bari, Alessandro
Stolcke, Andreas
Computation and Language
We present a light-weight approach for detecting nonfactual outputs from retrieval-augmented generation (RAG). Given a context and putative output, we compute a factuality score that can be thresholded to yield a binary decision to check the results of LLM-based question-answering, summarization, or other systems. Unlike factuality checkers that themselves rely on LLMs, we use compact, open-source natural language inference (NLI) models that yield a freely accessible solution with low latency and low cost at run-time, and no need for LLM fine-tuning. The approach also enables downstream mitigation and correction of hallucinations, by tracing them back to specific context chunks. Our experiments show high area under the ROC curve (AUC) across a wide range of relevant open source datasets, indicating the effectiveness of our method for fact-checking RAG output.
title Provenance: A Light-weight Fact-checker for Retrieval Augmented LLM Generation Output
topic Computation and Language
url https://arxiv.org/abs/2411.01022