
Incorporating community-specific context during the preference tuning of language models leads to more tailored responses that better reflect the norms and values of those communities.


coremsg

personalizing-language-models-with-community-preferences-a-study-using-reddit-data


Personalizing Language Models with Community Preferences: A Study Using Reddit Data


title_rewrite


This paper proposes a novel method for efficiently personalizing large language models (LLMs) to individual user preferences using small, locally trainable "preference agents" that generate natural language rules to guide the LLM's output without requiring resource-intensive fine-tuning.


aligning-large-language-models-to-individual-preferences-using-preference-agents


Aligning Large Language Models to Individual Preferences Using Preference Agents
