Paper on human behavior prediction with LLMs accepted at RO-MAN 2025