AI RESEARCH
Meet Dynamic Individual Preferences: Resolving Conflicting Human Value with Paired Fine-Tuning
arXiv CS.CL
•
ArXi:2604.12479v1 Announce Type: new Recent advances in large language models (LLMs) have significantly improved the alignment of models with general human preferences. However, a major challenge remains in adapting LLMs to individual preferences, which are not only diverse but also dynamic. In this paper, we