Guardado en:
Detalles Bibliográficos
Autores principales: Mohammadzadeh, Shahrad, Guerra, Juan David, Bonizzato, Marco, Rabbany, Reihaneh, Farnadi, Golnoosh
Formato: Preprint
Publicado: 2024
Materias:
Acceso en línea:https://arxiv.org/abs/2410.15460
Etiquetas: Agregar Etiqueta
Sin Etiquetas, Sea el primero en etiquetar este registro!
_version_ 1866908578480128000
author Mohammadzadeh, Shahrad
Guerra, Juan David
Bonizzato, Marco
Rabbany, Reihaneh
Farnadi, Golnoosh
author_facet Mohammadzadeh, Shahrad
Guerra, Juan David
Bonizzato, Marco
Rabbany, Reihaneh
Farnadi, Golnoosh
contents As large language models (LLMs) become increasingly prevalent, concerns about their reliability, particularly due to hallucinations - factually inaccurate or irrelevant outputs - have grown. Our research investigates the relationship between the uncertainty in training dynamics and the emergence of hallucinations. Using models from the Pythia suite and several hallucination detection metrics, we analyze hallucination trends and identify significant variance during training. To address this, we propose Sensitivity Dropout (SenD), a novel training protocol designed to reduce hallucination variance during training by deterministically dropping embedding indices with significant variability. In addition, we develop an unsupervised hallucination detection metric, Efficient EigenScore (EES), which approximates the traditional EigenScore in 2x speed. This metric is integrated into our training protocol, allowing SenD to be both computationally scalable and effective at reducing hallucination variance. SenD improves test-time reliability of Pythia and Meta's Llama models by up to 17% and enhances factual accuracy in Wikipedia, Medical, Legal, and Coding domains without affecting downstream task performance.
format Preprint
id arxiv_https___arxiv_org_abs_2410_15460
institution arXiv
publishDate 2024
record_format arxiv
spellingShingle Hallucination Detox: Sensitivity Dropout (SenD) for Large Language Model Training
Mohammadzadeh, Shahrad
Guerra, Juan David
Bonizzato, Marco
Rabbany, Reihaneh
Farnadi, Golnoosh
Artificial Intelligence
Computation and Language
Spectral Theory
As large language models (LLMs) become increasingly prevalent, concerns about their reliability, particularly due to hallucinations - factually inaccurate or irrelevant outputs - have grown. Our research investigates the relationship between the uncertainty in training dynamics and the emergence of hallucinations. Using models from the Pythia suite and several hallucination detection metrics, we analyze hallucination trends and identify significant variance during training. To address this, we propose Sensitivity Dropout (SenD), a novel training protocol designed to reduce hallucination variance during training by deterministically dropping embedding indices with significant variability. In addition, we develop an unsupervised hallucination detection metric, Efficient EigenScore (EES), which approximates the traditional EigenScore in 2x speed. This metric is integrated into our training protocol, allowing SenD to be both computationally scalable and effective at reducing hallucination variance. SenD improves test-time reliability of Pythia and Meta's Llama models by up to 17% and enhances factual accuracy in Wikipedia, Medical, Legal, and Coding domains without affecting downstream task performance.
title Hallucination Detox: Sensitivity Dropout (SenD) for Large Language Model Training
topic Artificial Intelligence
Computation and Language
Spectral Theory
url https://arxiv.org/abs/2410.15460