T-POP: Test-Time Personalization with Online Preference Feedback
This addresses the challenge of personalizing LLMs for new users without extensive data or resources, though it is incremental as it builds on existing test-time and bandit methods.
The paper tackles the cold-start problem in personalizing large language models for new users by introducing T-POP, a test-time algorithm that uses online pairwise feedback to steer decoding without fine-tuning, achieving rapid and data-efficient personalization with significant performance improvements over baselines.
Personalizing large language models (LLMs) to individual user preferences is a critical step beyond generating generically helpful responses. However, current personalization methods are ill-suited for new users, as they typically require either slow, resource-intensive fine-tuning or a substantial amount of pre-existing user data, creating a significant cold-start problem. To address this challenge, we introduce a new paradigm for real-time personalization by learning from online pairwise preference feedback collected during text generation. We propose T-POP (Test-Time Personalization with Online Preference Feedback}), a novel algorithm that synergistically combines test-time alignment with dueling bandits. Without updating the LLM parameters, T-POP steers the decoding process of a frozen LLM by learning a reward function online that captures user preferences. By leveraging dueling bandits, T-POP intelligently queries the user to efficiently balance between exploring their preferences and exploiting the learned knowledge to generate personalized text. Extensive experiments demonstrate that T-POP achieves rapid and data-efficient personalization, significantly outperforming existing baselines and showing consistent improvement with more user interactions.