Bilingualism and Language Modeling
(BilingLM 2027)

Important Dates

Overview

BilingLM 2027 aims to bridge human bilingualism and multilingual language models. We bring together researchers in natural language processing, linguistics, cognitive neuroscience, and education to explore learning, representation, and language use across humans and machines.

As language models become increasingly multilingual, questions about how they acquire, represent, and generate multiple languages echo long-standing questions in human bilingualism. Do models show critical periods for learning a second language? Do they develop shared semantic spaces or language-specific mechanisms? Which learning dynamics are shared by humans and language models, and where do they differ?

Our inaugural, one-day workshop provides a forum for interdisciplinary research and discussion: how can language models support bilingualism research, and how can findings about human bilingualism inform language modeling?

Topics of Interest

We welcome research on the following themes, including but not limited to:

• Language Acquisition and Learning. Human bilingual and second-language acquisition; cognitively plausible training of multilingual language models; bilingual BabyLM approaches; learning from limited, child-directed data; and comparisons of learning trajectories and critical period effects.
• Bilingual Representations and Mechanisms. Shared and language-specific representations; cross-lingual knowledge alignment; mechanistic interpretability; and comparisons between language model representations and human brain activity during bilingual processing.
• Code-Switching and Language Use. Patterns and mechanisms of code-switching in humans and language models; evaluation on code-switched benchmarks; processing code-mixed text; and generating naturalistic, grammatical code-switched language.
• Applications and Societal Impact. Language-model-assisted bilingualism research and corpus annotation; bilingual education and language assessment; and sociolinguistic questions about language ideologies, dialect prejudice, and multilingual training data.

We particularly encourage work on low-resource language pairs, diverse language families, and bilingual populations underrepresented in current research.

Invited Speakers

To be announced

Organizers

• 

  Hai Hu

  The Hong Kong Polytechnic University

• 

  Siyuan Song

  Princeton University

• 

  Linyang He

  Columbia University

• 

  Ercong Nie

  Shanghai Jiao Tong University

• 

  Ziyin (Maggie) Mai

  The Chinese University of Hong Kong

• 

  Luan Li

  Shanghai Jiao Tong University

• 

  Rui Wang

  Shanghai Jiao Tong University

• 

  Renfen Hu

  Beijing Normal University

• 

  Shaonan Wang

  The Hong Kong Polytechnic University

• 

  Yunhao Zhang

  Institute of Automation,
  Chinese Academy of Sciences

• 

  Adel Chaouch-Orozco

  City University of Hong Kong

• 

  Philipp Wicke

  Zander Labs
  Brandenburg University of Technology Cottbus

• 

  Catherine Arnett

  EleutherAI

• 

  Sercan Karakaş

  University of Chicago

Program Committee

To be announced

Workshop Information

BilingLM 2027 will be held in conjunction with COLING 2027 in Macau, China. Venue details, the call for papers, and the program will be announced here.

For conference information, please visit the COLING 2027 website.

Contact

For questions about the workshop, please contact:

• Hai Hu: hai.hu@polyu.edu.hk
• Siyuan Song: ss1280@princeton.edu