Dango: A Strictly L1-Only Large Language Model for Studying Second Language Acquisition

2026-06-17 · Source: Computation and Language · Field: Technology & Digital — Artificial Intelligence & Machine Learning, Social Sciences & Behavioral Studies · Depth: Expert, quick

Summary

Dango is a 1.8B-parameter large language model designed for controlled studies of L1-to-L2 (Japanese-to-English) transfer in second language acquisition (SLA). This model addresses a key challenge in scaling such models: L2 contamination within the "monolingual" pretraining corpus used for L1 acquisition. To mitigate this, the researchers propose a filtering method that reduces premature exposure to English while preserving realistic, minimal exposure. Subsequently, Dango is fine-tuned on LLM-generated L2-learning lessons to simulate the L2 acquisition process. Evaluations confirm that Dango develops human-like L2 production patterns, outperforming both unfiltered and standard multilingual baselines. The model, data, and code are released to facilitate reproducible computational SLA research and learner-facing applications.

Key takeaway

For research scientists studying second language acquisition with large language models, Dango offers a robust, controlled environment. You should consider integrating this 1.8B-parameter L1-only model to mitigate L2 contamination issues and achieve more human-like L2 production patterns in your simulations. Its released data and code facilitate reproducible studies, allowing you to explore L1-to-L2 transfer dynamics more accurately.

Key insights

Dango, a 1.8B-parameter LLM, enables controlled study of L1-to-L2 transfer in SLA by mitigating L2 contamination and simulating acquisition.

Principles

• L2 contamination is a key challenge in scaling L1-only LLMs.
• Controlled L2 exposure is crucial for realistic SLA simulation.
• LLM-generated lessons can simulate L2 acquisition processes.

Method

A filtering method reduces L2 contamination in L1 pretraining corpora, followed by fine-tuning on LLM-generated L2-learning lessons to simulate second language acquisition.

In practice

• Use Dango for reproducible computational SLA research.
• Develop learner-facing L2 acquisition applications.
• Apply filtering to create L1-only pretraining datasets.

Topics

• Large Language Models
• Second Language Acquisition
• L1-to-L2 Transfer
• Data Contamination
• Computational Linguistics
• Japanese-to-English

Best for: AI Scientist, Research Scientist

Related on AIssential

• Dango: A Strictly L1-Only Large Language Model for Studying Second Language Acquisition — Dango is a 1.8B-parameter L1-only LLM designed for controlled L1-to-L2 transfer studies in second language acquisition.
• Linguistic Issues in Computational Linguistics for Non-English Languages — English-centric LLM design creates systematic linguistic weaknesses for non-English languages across multiple levels.
• Zero-Shot to Full-Resource: Cross-lingual Transfer Strategies for Aspect-Based Sentiment Analysis — Cross-lingual transfer strategies for ABSA vary in effectiveness based on model architecture and resource availability.
• LLM Parameters for Math Across Languages: Shared or Separate? — LLM mathematical reasoning involves partially shared and partially language-specific parameters, with English dominating parameter sets.
• The F1 of Formula One: Applicability of Pre-trained NER Models to Brazilian TV Interview Transcripts — LLMs offer superior linguistic sensitivity for NER in noisy interview transcripts compared to statistical methods.
• Textual Inference in Portuguese: Comparing Language Models — LLMs struggle with logic-sensitive semantic reasoning in Portuguese, highlighting the need for specialized benchmarks.

Open in AIssential →

Editorial summary, takeaway, and curation by AIssential. Original article published by Computation and Language.