research
Research on social pragmatics, humor, and human–AI interaction.
My research focuses on how language models interpret social meaning and how people create, communicate, and participate alongside AI systems. Here are two recent papers.
EMNLP 2026 · Main Conference · Oral
You Really Didn't Get That? Benchmarking Social Pragmatic Inference for Indirect and Playful Chinese Online Comments
Can a language model tell what a short, playful comment is doing in an online conversation? We built 4,735 diagnostic questions from real Chinese social media interactions. Across eight models, average accuracy was 68.7%, compared with 90.8% for human readers.
Findings of ACL 2026
Multi-Agent Comedy Club: Investigating Community Discussion Effects on LLM Humor Generation
How does comedy change when AI writers work in a community, responding to other performers and audience feedback? We investigate how multi-agent discussion shapes humor writing, including its social and creative qualities.