Learning Transferable Latent User Preferences for Human-Aligned Decision Making

arXiv:2605. 12682v1 Announce Type: new Abstract: Large language models (LLMs) are increasingly used as reasoning modules in many applications.