Salvato in:
Dettagli Bibliografici
Autori principali: Poey, Ian, Liu, Jiajun, Zhong, Qishuai, Chenailler, Adrien
Natura: Preprint
Pubblicazione: 2024
Soggetti:
Accesso online:https://arxiv.org/abs/2411.03920
Tags: Aggiungi Tag
Nessun Tag, puoi essere il primo ad aggiungerne!!
_version_ 1866912107396595712
author Poey, Ian
Liu, Jiajun
Zhong, Qishuai
Chenailler, Adrien
author_facet Poey, Ian
Liu, Jiajun
Zhong, Qishuai
Chenailler, Adrien
contents Real-time detection of out-of-context LLM outputs is crucial for enterprises looking to safely adopt RAG applications. In this work, we train lightweight models to discriminate LLM-generated text that is semantically out-of-context from retrieved text documents. We preprocess a combination of summarisation and semantic textual similarity datasets to construct training data using minimal resources. We find that DeBERTa is not only the best-performing model under this pipeline, but it is also fast and does not require additional text preprocessing or feature engineering. While emerging work demonstrates that generative LLMs can also be fine-tuned and used in complex data pipelines to achieve state-of-the-art performance, we note that speed and resource limits are important considerations for on-premise deployment.
format Preprint
id arxiv_https___arxiv_org_abs_2411_03920
institution arXiv
publishDate 2024
record_format arxiv
spellingShingle RAGulator: Lightweight Out-of-Context Detectors for Grounded Text Generation
Poey, Ian
Liu, Jiajun
Zhong, Qishuai
Chenailler, Adrien
Computation and Language
Real-time detection of out-of-context LLM outputs is crucial for enterprises looking to safely adopt RAG applications. In this work, we train lightweight models to discriminate LLM-generated text that is semantically out-of-context from retrieved text documents. We preprocess a combination of summarisation and semantic textual similarity datasets to construct training data using minimal resources. We find that DeBERTa is not only the best-performing model under this pipeline, but it is also fast and does not require additional text preprocessing or feature engineering. While emerging work demonstrates that generative LLMs can also be fine-tuned and used in complex data pipelines to achieve state-of-the-art performance, we note that speed and resource limits are important considerations for on-premise deployment.
title RAGulator: Lightweight Out-of-Context Detectors for Grounded Text Generation
topic Computation and Language
url https://arxiv.org/abs/2411.03920