Semantics Research Group (Tokyo/Hybrid) 2026/6/5

[jrta...@gmail.com]

意味論研究会のお知らせ

意味論研究会の会合をお知らせします。対面の場所・ズームのアクセス情報は下記の通りです。どなたでも参加できますので，多くの方の参加をお待ちしています。また、会合終了後に食事会を予定しております。ご参加希望の方はご指名と所属をご記入の上、その旨を5月25日（水）までにChris Tancrediにご連絡くださいますようお願い致します。(Chris Tancredi, cdtan...@gmail.com)

なお、通常より1時間遅くの18:00開始となりますので、ご注意ください。

The Semantics Research Group will be having a talk as detailed below. There will also be a dinner after the talk. If you would like to participate in the dinner, please RSVP with Chris Tancredi by Wednesday, May 25th. Make sure to include your name and affiliation. (Chris Tancredi,  cdtan...@gmail.com)

Note that the meeting will start one hour later than usual at 18:00. 

Date: June 5 (Friday) Time: 6:00 pm (NOTE THE TIME) 

Place: Keio University, Mita campus, South Annex building, 7th floor conference room https://www.keio.ac.jp/en/maps/mita.html 

Online option: https://keio-univ.zoom.us/j/82972175560?pwd=YlRuaGt0NUFWWmI1ajhrNm1vQ3ZVZz09 

Speaker: Lisa Bylinina (Utrecht University) 

Title: Pathways to human-like behaviour of language models 

Abstract: I will approach the (L)LM "black box" as an empirical laboratory for studying natural language. First, we will look at the emergence of soft, context-sensitive constraints through a case study on adjective order preferences (AOPs) and track how models generalise beyond simple n-gram distributions in training data. Moving from observation to intervention, we introduce targeted manipulations of the learning environment through filtered corpus training and controlled extralinguistic grounding. Ultimately, the goal is to explore what the artificial ontogeny of these models can teach us about human language.