Papers by Daniel Frees
SynthesizeMe! Inducing Persona-Guided Prompts for Personalized Reward Models in LLMs (2025.acl-long)
Copied to clipboard
| Challenge: | Recent calls for pluralistic alignment of Large Language Models encourage adapting models to diverse user preferences. |
| Approach: | They propose a method to induce synthetic user personas from user interactions for personalized reward modeling. |
| Outcome: | The proposed approach improves LLM-as-a-judge accuracy by 4.4% on Chatbot Arena. |