Findings of the BabyLM Challenge: Sample-Efficient Pretraining on Developmentally Plausible Corpora

Citations Over TimeTop 10% of 2023 papers

Abstract

Children can acquire language from less than 100 million words of input. Large language models are far less data-efficient: they typically require 3 or 4 orders of magnitude more data and still do not perform as well as humans on many evaluations. These intensive resource demands limit the ability of researchers to train new models and use existing models as developmentally plausible cognitive models. The BabyLM Challenge is a communal effort in which participants compete to optimize language model training on a fixed data budget. Submissions are compared on various evaluation tasks targeting grammatical ability, downstream task performance, and generalization. Participants can submit to up to three tracks with progressively looser data restrictions. From over 30 submissions, we extract concrete recommendations on how best to train data-efficient language models, and on where future efforts should (and perhaps should not) focus. The winning submissions using the LTG-BERT architecture (Samuel et al., 2023) outperformed models trained on trillions of words. Other submissions achieved strong results through training on shorter input sequences or training a student model on a pretrained teacher. Curriculum learning attempts, which accounted for a large number of submissions, were largely unsuccessful, though some showed modest improvements.

Related Papers

• → Distribution of vehicular loads on bridge girders by the FEA using ADINA: modeling, simulation, and comparison(1999)33 cited
• Application of ADINA in Nonlinear Analysis of Reinforced Concrete Structure(2006)
• → Finite element analysis—design aid for a proposed shaft at hoover dam(1987)1 cited
• Computational Check of Bearing Capacity of Aohan Double-arched Bridge With ADINA(2007)
• THE NONLINEAR ANALYSIS OF REINFORCED CONCRETE STRUCTURES BASED ON ADINA(2013)