arXiv 2406.186782025-03-04cs.LGcs.AIcs.CL

利用错位响应进行LLMs的少样本个性化

Few-shot Personalization of LLMs with Mis-aligned Responses

Yonsei University（延世大学）；Carnegie Mellon University（卡内基梅隆大学）

Jaehyung Kim, Yiming Yang

备注：: NAACL 25 (main, long), 32 pages

阅读 PDF

arXiv 页面 ↗HTML 原文 ↗

AI总结：

本文提出Fermi方法，通过利用LLMs的错位响应上下文迭代改进个性化提示词，结合用户画像和少量过往意见，实现LLMs的少样本个性化，显著提升了跨基准测试的性能。

摘要：

随着用户多样性的增加，大型语言模型（LLMs）提供个性化响应的能力变得日益重要。由于缺乏个性化学习或依赖共享的个人数据，现有方法在LLM个性化方面仅取得了有限的成功。本文提出了一种利用LLMs错位响应进行少样本个性化的新方法（Fermi）。我们的核心思想是通过基于用户画像（例如人口统计信息）和少量过往意见示例，利用LLMs逐步改进提示词，从而为每个用户学习一组个性化提示词。在提示词改进的迭代过程中，我们融入了LLMs错位响应的上下文，这对于LLMs的有效个性化尤为关键。此外，我们开发了一种有效的推理方法，以进一步利用测试查询的上下文和个性化提示词。实验结果表明，与表现最佳的基线方法相比，Fermi在各种基准测试上显著提升了性能。

英文摘要：

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.

机构、总结、中文摘要均由AI分析而来，可能存在错误，请以论文正文为准。

AI总结：

摘要：

英文摘要：

相关论文