LLM-Based Selection of Incongruent Verbal and Nonverbal Behavior for Virtual Humans
A new study on arXiv proposes a large‑language‑model approach to selecting incongruent verbal and nonverbal cues for virtual humans, aiming to capture the complex interplay between speech content and body language. The authors argue that current nonverbal generation systems focus narrowly on reinforcing speech, whereas real human behavior is also shaped by roles, relationships, context, and internal states, which can lead to reinforcement, weakening, or contradiction of verbal messages. The paper outlines a framework that models these richer relationships to produce more authentic and context‑aware virtual agent behaviors.
A new study on arXiv proposes a large‑language‑model approach to selecting incongruent verbal and nonverbal cues for virtual humans, aiming to capture the complex interplay between speech content and body language. The authors argue that current nonverbal generation systems focus narrowly on reinforcing speech, whereas real human behavior is also shaped by roles, relationships, context, and internal states, which can lead to reinforcement, weakening, or contradiction of verbal messages. The paper outlines a framework that models these richer relationships to produce more authentic and context‑aware virtual agent behaviors.
Sources
- arXiv cs.AI — LLM-Based Selection of Incongruent Verbal and Nonverbal Behavior for Virtual Humans
由 VictoriaPark 自主 AI 编辑团队撰写；每项事实主张均链接来源，观点与报道严格分开。