Pretraining Language Models with LoRA and Artificial Languages
Identifikátory výsledku
Kód výsledku v IS VaVaI
<a href="https://www.isvavai.cz/riv?ss=detail&h=RIV%2F00216208%3A11320%2F25%3A10511610" target="_blank" >RIV/00216208:11320/25:10511610 - isvavai.cz</a>
Výsledek na webu
<a href="https://aclanthology.org/2025.babylm-main.37/" target="_blank" >https://aclanthology.org/2025.babylm-main.37/</a>
DOI - Digital Object Identifier
—
Alternativní jazyky
Jazyk výsledku
angličtina
Název v původním jazyce
Pretraining Language Models with LoRA and Artificial Languages
Popis výsledku v původním jazyce
Large language models (LLMs) require a substantial amount of training data, which contrasts with the data-efficient learning observed in humans. In our submission to the BabyLM Challenge, we address this disparity by proposing a parameter-efficient pretraining approach for language acquisition from limited data. Our approach involves initializing the model with token embeddings trained by a shallow model, followed by tuning the non-embedding parameters with non-linguistic data to introduce structural biases. Then, we freeze the resulting model and pretrain it on the 10M-token BabyLM corpus using LoRA adapters. Experiments on small corpora demonstrate that our approach improves upon classic pretraining of the entire model.
Název v anglickém jazyce
Pretraining Language Models with LoRA and Artificial Languages
Popis výsledku anglicky
Large language models (LLMs) require a substantial amount of training data, which contrasts with the data-efficient learning observed in humans. In our submission to the BabyLM Challenge, we address this disparity by proposing a parameter-efficient pretraining approach for language acquisition from limited data. Our approach involves initializing the model with token embeddings trained by a shallow model, followed by tuning the non-embedding parameters with non-linguistic data to introduce structural biases. Then, we freeze the resulting model and pretrain it on the 10M-token BabyLM corpus using LoRA adapters. Experiments on small corpora demonstrate that our approach improves upon classic pretraining of the entire model.
Klasifikace
Druh
O - Ostatní výsledky
CEP obor
—
OECD FORD obor
10201 - Computer sciences, information science, bioinformathics (hardware development to be 2.2, social aspect to be 5.8)
Návaznosti výsledku
Projekt
—
Návaznosti
R - Projekt Ramcoveho programu EK
Ostatní
Rok uplatnění
2025
Kód důvěrnosti údajů
S - Úplné a pravdivé údaje o projektu nepodléhají ochraně podle zvláštních právních předpisů