Few-shot Personalization of LLMs with Mis-aligned Responses

Fuente: arXiv
Saved in:
Bibliographic Details
Main Authors: Kim, Jaehyung, Yang, Yiming
Format: Preprint
Published: 2024
Subjects:
Online Access:
Tags: Add Tag
No Tags, Be the first to tag this record!
_version_ 1866916641645789184
author Kim, Jaehyung
Yang, Yiming
author_facet Kim, Jaehyung
Yang, Yiming
contents As the diversity of users increases, the capability of providing personalized responses by large language models (LLMs) has become increasingly important. Existing approaches have only limited successes in LLM personalization, due to the absence of personalized learning or the reliance on shared personal data. This paper proposes a new approach for a few-shot personalization of LLMs with their mis-aligned responses (Fermi). Our key idea is to learn a set of personalized prompts for each user by progressively improving the prompts using LLMs, based on user profile (e.g., demographic information) and a few examples of previous opinions. During an iterative process of prompt improvement, we incorporate the contexts of mis-aligned responses by LLMs, which are especially crucial for the effective personalization of LLMs. In addition, we develop an effective inference method to further leverage the context of the test query and the personalized prompts. Our experimental results demonstrate that Fermi significantly improves performance across various benchmarks, compared to best-performing baselines.
format Preprint
id arxiv_https___arxiv_org_abs_2406_18678
institution arXiv
publishDate 2024
record_format arxiv
spellingShingle Few-shot Personalization of LLMs with Mis-aligned Responses
Kim, Jaehyung
Yang, Yiming
Machine Learning
Artificial Intelligence
Computation and Language
As the diversity of users increases, the capability of providing personalized responses by large language models (LLMs) has become increasingly important. Existing approaches have only limited successes in LLM personalization, due to the absence of personalized learning or the reliance on shared personal data. This paper proposes a new approach for a few-shot personalization of LLMs with their mis-aligned responses (Fermi). Our key idea is to learn a set of personalized prompts for each user by progressively improving the prompts using LLMs, based on user profile (e.g., demographic information) and a few examples of previous opinions. During an iterative process of prompt improvement, we incorporate the contexts of mis-aligned responses by LLMs, which are especially crucial for the effective personalization of LLMs. In addition, we develop an effective inference method to further leverage the context of the test query and the personalized prompts. Our experimental results demonstrate that Fermi significantly improves performance across various benchmarks, compared to best-performing baselines.
title Few-shot Personalization of LLMs with Mis-aligned Responses
topic Machine Learning
Artificial Intelligence
Computation and Language
url https://arxiv.org/abs/2406.18678