🎉 SAC Highlights at EMNLP 2025 🎉

I’m honored to share that I presented my first paper at EMNLP2025, the very conference where I first read the paper that sparked my journey into AI research: “SODA: Million-scale Dialogue Distillation with Social Commonsense Contextualization.” Starting from the concepts introduced in SODA, I eventually delved into the research area of social norm dialogue, which became the foundation of my current work. It was truly meaningful to come full circle and present at the same conference that first inspired me.

Our paper, “NormGenesis: Multicultural Dialogue Generation via Exemplar-Guided Social Norm Modeling and Violation Recovery,” was nominated for Outstanding Paper, SAC Highlights, and Best Resource Paper, and we are thrilled to have ultimately won the 🏆SAC Highlights Award!

In this work, we addressed the challenges of modeling social norms and improving linguistic quality in low-resource languages, an area often overlooked in social dialogue research. We achieved this by proposing a novel Exemplar-Guided Refinement method.

Crucially, we are the first to introduce the “Violation-to-Resolution” (V2R) dialogue type, which allows conversational systems to model the process of social recovery after a norm violation, leading to more realistic and robust real-world interactions. You can read our full paper here: 👉 https://arxiv.org/abs/2509.18395 We plan to release the code and dataset as soon as possible. Stay tuned for updates!

Also, a big thanks to my co-author Jangho Choi and advisor Jihie Kim.