Summary of On the Way to Llm Personalization: Learning to Remember User Conversations, by Lucie Charlotte Magister et al.
On the Way to LLM Personalization: Learning to Remember User Conversations
by Lucie Charlotte Magister, Katherine Metcalf, Yizhe Zhang, Maartje ter Hoeve
First submitted to arxiv on: 20 Nov 2024
Categories
- Main: Computation and Language (cs.CL)
- Secondary: Machine Learning (cs.LG)
GrooveSquid.com Paper Summaries
GrooveSquid.com’s goal is to make artificial intelligence research accessible by summarizing AI papers in simpler terms. Each summary below covers the same AI paper, written at different levels of difficulty. The medium difficulty and low difficulty versions are original summaries written by GrooveSquid.com, while the high difficulty version is the paper’s original abstract. Feel free to learn from the version that suits you best!
| Summary difficulty | Written by | Summary |
|---|---|---|
| High | Paper authors | High Difficulty Summary Read the original abstract here |
| Medium | GrooveSquid.com (original content) | Medium Difficulty Summary This paper explores the limitation of Large Language Models (LLMs) in personalizing responses to user preferences and behaviors through knowledge injection. Prior work has focused on style transfer or incorporating small factoids about users, but this challenge remains open. The authors propose PLUM, a pipeline that up-samples conversations as question-answer pairs and fine-tunes a low-rank adaptation adapter with weighted cross entropy loss. This approach enables personalized conversations while respecting real-world constraints such as sequential conversations and parameter-efficient settings. The proposed method achieves an accuracy of 81.5% on 100 conversations, outperforming baselines like RAG. |
| Low | GrooveSquid.com (original content) | Low Difficulty Summary Imagine a super smart computer that can help us with many tasks. However, it’s not perfect because it doesn’t really understand what we want to talk about or who we are. This paper tries to solve this problem by teaching the computer to learn from past conversations and use that knowledge to have more personalized chats in the future. The authors created a new way to do this using something called PLUM, which helps the computer understand how to adapt to different people’s styles and preferences. Even in its early stages, this approach is quite effective, with an accuracy rate of 81.5% on conversations. |
Keywords
» Artificial intelligence » Cross entropy » Low rank adaptation » Parameter efficient » Rag » Style transfer
