Challenge Submission
Submission of models and results via HuggingFace and evaluation pipeline for the BabyLM Challenge, following the 10-epoch compute limit.
NLP/CORE Unranked
BabyLM Challenge 2026
The BabyLM Challenge at EMNLP 2026
The BabyLM Challenge 2026 is a shared task and workshop co-located with EMNLP 2026, aiming to advance sample-efficient language model pretraining using developmentally plausible, human-scale data budgets. It focuses on simulating child-like language acquisition through constrained training corpora and evaluates models on English, Dutch, and Chinese, while promoting accessible, university-budget research in cognitive modeling and low-resource NLP.
Key deadlines
Verified September 9, 2026Full paper
July 21, 2026
AoE
Conference timeline
Submission and decisions
Full paperKey deadline
July 21, 2026 · AoE
Paper fit
Contribution paths
A strong submission should clearly identify its contribution and evaluate it appropriately.
Workshop Paper
Paper submissions (4-8 pages using CoNLL template) with hyperparameter form, for non-challenge research, submitted via OpenReview.
ARR Commitment
Commitment of papers already submitted to ACL Rolling Review with reviews, to be considered for the BabyLM workshop.
Research areas in scope
01
Training Tracks
Strict (100M-word detoxified dataset)Strict-Small (10M-word detoxified dataset)MultiLingual (BabyBabelLM, focusing on English, Dutch, and Chinese)
Policies worth checking twice
- Submissions must include a HuggingFace link to the model and a download link to results generated via collect_results.py.
- Challenge entries are limited to 10 epochs over training data; workshop-only papers are exempt.
- Multimodal and Interaction tracks are no longer standalone; paired image-text data and teacher-model feedback are permitted within Strict/Strict-Small tracks.
- Submissions must use the CoNLL template (4-8 pages) and include a hyperparameter form.
- Challenge submissions must use the dedicated Challenge OpenReview portal; workshop papers must use the Workshop portal; ARR submissions require prior ACL Rolling Review submission.
Official sources
Compiled from the official call for papers. The organizers’ pages remain authoritative.
Last verified September 9, 2026