AACL-IJCNLP 2026 Tutorial

This tutorial provides a structured overview of code-switching for multilingual large language models (LLMs), connecting linguistic theory, NLP methodology, and human-centered perspectives. Code-switching is a natural communicative practice among multilingual speakers and is increasingly central to LLM interactions, where users bring mixed-language practices to prompts and models themselves may produce mixed-language outputs. The tutorial will cover theoretical foundations from sociolinguistics, code-switching behavior in human-AI interaction, datasets and evaluation benchmarks, synthetic data generation, multilingual LLM training and inference, mechanistic interpretability of unintended language switching, speech and multimodal code-switching, and societal impacts. By synthesizing work across NLP, linguistics, HCI, speech processing, and multilingual communities, the tutorial aims to give participants a coherent view of the field and identify open challenges for building multilingual LLMs that better reflect how multilingual users actually communicate.

• A survey of Code-Switching: Linguistic and Social Perspectives for Language Technologies (Dogğruöz et al., 2021)

• The Decades Progress on Code-Switching Research in NLP: A Systematic Survey on Trends and Challenges (Winata et al., 2023)

• Multilingual Large Language Models Are Not (Yet) Code-Switchers (Zhang et al., 2023)

• SwitchLingua: The First Large-Scale Multilingual and Multi-Ethnic Code-Switching Dataset (Xie et al., 2025)

• Investigating and Scaling up Code-Switching for Multilingual Language Model Pre-Training (Wang et al., 2025)

• Language Mixing in Reasoning Language Models: Patterns, Impact, and Internal Causes (Wang et al., 2025)

• Mechanistic Understanding and Mitigation of Language Confusion in English-Centric Large Language Models (Nie et al., 2025)

• CoSTA: Code-Switched Speech Translation using Aligned Speech-Text Interleaving (Shankar et al., 2025)

• Toward a Multilingual Conversational Agent: Challenges and Expectations of Code-mixing Multilingual Users (Choi et al., 2023)