Seungju Han

wade3han AT snu.ac.kr
Last update - 23/06/06

Bio

Collaborating Researcher @ AI2 (Mosaic Team), Seattle.
Undergraduate @ SNU (Electrical & Computer Engineering). Seoul.
Previously, ML Engineer @ Hyperconnect (acquired by Match Group for $1.7B). Seoul.

These days, my primary focus lies in the exploration of the path leading to AGI. My specific interests encompasses the following topics: (1) Effectively teaching machine using signals from real-world beyond textual data (e.g., with CHAMPAGNE), (2) tackling the alignment problem, and (3) addressing the challenges associated with long-context inputs.

At Hyperconnect, I contributed to solving real-world problems (e.g., with PC Softmax / LADE) using AI, including building social chatbot product (e.g., with DRESS, PDP, CORGE, G2R, and Attentron.) to solve people’s loneliness.

Personal

I like workouts in the gym and singing music. I participated in building a product (GYMWORK) to assist people with their workout routines!

Curriculum Vitae

news

Mar 20, 2023 My first work done in collaboration with AI2 is online!
Oct 6, 2022 One paper, which I participated as a first author, is accepted to EMNLP 22 Findings!
Sep 20, 2022 I am really grateful to start research collaboration with AI2 hosted by Youngjae and advised by Yejin!
Jul 1, 2022 I am so excited that I am going to the University of Washington as an exchange student during the autumn quarter of 2022!
Apr 8, 2022 One paper, which I participated as a first author, is accepted to NAACL 2022!

selected publications

  1. preprint
    CHAMPAGNE: Learning Real-world Conversation from Large-Scale Web Videos
    Han, Seungju, Hessel, Jack, Dziri, Nouha, Choi, Yejin, and Yu, Youngjae
    2023
    PDF Blog Code
  2. EMNLP
    Measuring and Improving Semantic Diversity of Dialogue Generation
    Han, Seungju, Kim, Beomsu, and Chang, Buru
    In Conference on Empirical Methods in Natural Language Processing (Findings) 2022
    PDF Code
  3. NAACL
    Meet Your Favorite Character: Open-domain Chatbot Mimicking Fictional Characters with only a Few Utterances
    Han, Seungju*, Kim, Beomsu*, Yoo, Jin Yong*, Seo, Seokjun, Kim, Sangbum, Erdenee, Enkhbayar, and Chang, Buru
    In Conference of the North American Chapter of the Association for Computational Linguistics (Short) 2022
    PDF Code
  4. ACL Workshop
    Understanding and Improving the Exemplar-based Generation for Open-domain Conversation
    Han, Seungju*, Kim, Beomsu*, Seo, Seokjun*, Erdenee, Enkhbayar*, and Chang, Buru
    In 4th NLP for Conversational AI (Oral, Outstanding Paper), Annual Meeting of the Association for Computational Linguistics 2022
    PDF Code
  5. EMNLP
    Distilling the Knowledge of Large-scale Generative Models into Retrieval Models for Efficient Open-domain Conversation
    Kim, Beomsu*, Seo, Seokjun*,  Han, Seungju*, Enkhbayar, Erdenee*, and Chang, Buru
    In Conference on Empirical Methods in Natural Language Processing (Findings) 2021
    PDF Code