babylm-2026-mtl-bigckpt
A 213M-parameter GPT-2-style causal language model submitted to the multilingual track of the BabyLM 2026 Challenge (English, Dutch, Chinese), trained from scratch within the 100M English-equivalent-word budget.
This is the overall submission. It obtained a leaderboard-reported Multilingual Average of 50.68.
Model details
| Architecture | GPT-2-style decoder with multi-token prediction (MTP) |
| Parameters | 212,790,272 (213M) |
| Layers / heads / hidden | 14 / 16 / 1024 |
| Context length | 512 |
| Tokenizer | byte-level BPE, 32,000 vocab |
| Training objective | next-token + MTP auxiliary heads ($k{=}4$, ramped $1\to4$ over epochs; only the first head is used at inference) |
| Optimizer | AdamW, peak LR $4.5\times10^{-4}$, 10 epochs |
| Track | BabyLM 2026 multilingual |
| Languages | English, Dutch, Chinese |
Training data
Trained within the fixed multilingual budget on a combination of:
- the official BabyBabelLM corpora for English, Dutch, and Chinese;
- parallel-corpus text from Tatoeba and OPUS-100;
- task-family-targeted synthetic text generated offline by an approved teacher model (Qwen3.5-4B).
Published training data: 1ghub/babylm-2026-mtl-data
Note on budget. The track measures the budget in bytes: the 100M-word tier is 543.00 MB, and a corpus qualifies while its UTF-8 size divided by the language premium stays at or below that. We use the coefficients published in the organizers' table: English 1.0, Dutch 1.0516, and simplified Chinese 0.93597, the script BabyBabelLM uses. (The 0.9894 listed in the call for papers is the coefficient that table gives for traditional Chinese.) Under this rule the training corpus is 527.7 MB of English-equivalent bytes, 97.2% of the cap.
Results
Development-set scores (percentages):
| Metric | Score |
|---|---|
| Official Multilingual Average | 50.68 |
| Zero-shot (16-task mean, ZS16) | 57.82 |
| Fine-tuning (22-task mean, FT22) | 41.25 |
| ZS — English / Dutch / Chinese | 59.89 / 59.77 / 53.41 |
| FT — English / Dutch / Chinese | 43.86 / 39.37 / 40.61 |
Intended use and limitations
Research model for data-efficient multilingual language modeling under the BabyLM setting. It is a small model trained on a developmentally plausible amount of text and is not intended for production use. Scores are on development sets; evaluation reuse across selection stages introduces selection bias.
Teacher model
Synthetic training text was generated (not distilled) by Qwen3.5-4B, an approved teacher for the challenge. Generation prompts used task-level descriptions.
Citation
Companion paper: Language Allocation and Targeted Data Design for a Multilingual BabyLM (BabyLM 2026 Workshop, EMNLP 2026).