Panza: Design and Analysis of a Fully-Local Personalized Text Writing Assistant

Fuente: arXiv
Saved in:
Bibliographic Details
Main Authors: Nicolicioiu, Armand, Iofinova, Eugenia, Jovanovic, Andrej, Kurtic, Eldar, Nikdan, Mahdi, Panferov, Andrei, Markov, Ilia, Shavit, Nir, Alistarh, Dan
Format: Preprint
Published: 2024
Subjects:
Online Access:
Tags: Add Tag
No Tags, Be the first to tag this record!
_version_ 1866929704691302400
author Nicolicioiu, Armand
Iofinova, Eugenia
Jovanovic, Andrej
Kurtic, Eldar
Nikdan, Mahdi
Panferov, Andrei
Markov, Ilia
Shavit, Nir
Alistarh, Dan
author_facet Nicolicioiu, Armand
Iofinova, Eugenia
Jovanovic, Andrej
Kurtic, Eldar
Nikdan, Mahdi
Panferov, Andrei
Markov, Ilia
Shavit, Nir
Alistarh, Dan
contents The availability of powerful open-source large language models (LLMs) opens exciting use-cases, such as using personal data to fine-tune these models to imitate a user's unique writing style. Two key requirements for such assistants are personalization - in the sense that the assistant should recognizably reflect the user's own writing style - and privacy - users may justifiably be wary of uploading extremely personal data, such as their email archive, to a third-party service. In this paper, we present a new design and evaluation for such an automated assistant, for the specific use case of email generation, which we call Panza. Panza's personalization features are based on a combination of fine-tuning using a variant of the Reverse Instructions technique together with Retrieval-Augmented Generation (RAG). We demonstrate that this combination allows us to fine-tune an LLM to reflect a user's writing style using limited data, while executing on extremely limited resources, e.g. on a free Google Colab instance. Our key methodological contribution is the first detailed study of evaluation metrics for this personalized writing task, and of how different choices of system components--the use of RAG and of different fine-tuning approaches-impact the system's performance. Additionally, we demonstrate that very little data - under 100 email samples - are sufficient to create models that convincingly imitate humans. This finding showcases a previously-unknown attack vector in language models - that access to a small number of writing samples can allow a bad actor to cheaply create generative models that imitate a target's writing style. We are releasing the full Panza code as well as three new email datasets licensed for research use at https://github.com/IST-DASLab/PanzaMail.
format Preprint
id arxiv_https___arxiv_org_abs_2407_10994
institution arXiv
publishDate 2024
record_format arxiv
spellingShingle Panza: Design and Analysis of a Fully-Local Personalized Text Writing Assistant
Nicolicioiu, Armand
Iofinova, Eugenia
Jovanovic, Andrej
Kurtic, Eldar
Nikdan, Mahdi
Panferov, Andrei
Markov, Ilia
Shavit, Nir
Alistarh, Dan
Computation and Language
Artificial Intelligence
Human-Computer Interaction
Machine Learning
The availability of powerful open-source large language models (LLMs) opens exciting use-cases, such as using personal data to fine-tune these models to imitate a user's unique writing style. Two key requirements for such assistants are personalization - in the sense that the assistant should recognizably reflect the user's own writing style - and privacy - users may justifiably be wary of uploading extremely personal data, such as their email archive, to a third-party service. In this paper, we present a new design and evaluation for such an automated assistant, for the specific use case of email generation, which we call Panza. Panza's personalization features are based on a combination of fine-tuning using a variant of the Reverse Instructions technique together with Retrieval-Augmented Generation (RAG). We demonstrate that this combination allows us to fine-tune an LLM to reflect a user's writing style using limited data, while executing on extremely limited resources, e.g. on a free Google Colab instance. Our key methodological contribution is the first detailed study of evaluation metrics for this personalized writing task, and of how different choices of system components--the use of RAG and of different fine-tuning approaches-impact the system's performance. Additionally, we demonstrate that very little data - under 100 email samples - are sufficient to create models that convincingly imitate humans. This finding showcases a previously-unknown attack vector in language models - that access to a small number of writing samples can allow a bad actor to cheaply create generative models that imitate a target's writing style. We are releasing the full Panza code as well as three new email datasets licensed for research use at https://github.com/IST-DASLab/PanzaMail.
title Panza: Design and Analysis of a Fully-Local Personalized Text Writing Assistant
topic Computation and Language
Artificial Intelligence
Human-Computer Interaction
Machine Learning
url https://arxiv.org/abs/2407.10994