Adapting LLMs to Hebrew: Unveiling DictaLM 2.0 with Enhanced Vocabulary and Instruction Capabilities

TL;DR

This paper presents DictaLM 2.0 and DictaLM 2.0-Instruct, innovative Hebrew LLMs trained on a large corpus, with novel adaptation techniques and a new benchmark suite for evaluation in low-resource language settings.

Contribution

Introduction of new training methodologies for adapting pre-trained LLMs to Hebrew, along with the development of DictaLM 2.0 models and a comprehensive Hebrew NLP benchmark.

Findings

Models perform well across diverse Hebrew NLP tasks.

Novel adaptation techniques improve Hebrew language learning in LLMs.

Benchmark provides a new standard for Hebrew NLP evaluation.

Abstract

Training large language models (LLMs) in low-resource languages such as Hebrew poses unique challenges. In this paper, we introduce DictaLM2.0 and DictaLM2.0-Instruct, two LLMs derived from the Mistral model, trained on a substantial corpus of approximately 200 billion tokens in both Hebrew and English. Adapting a pre-trained model to a new language involves specialized techniques that differ significantly from training a model from scratch or further training existing models on well-resourced languages such as English. We outline these novel training methodologies, which facilitate effective learning and adaptation to the linguistic properties of Hebrew. Additionally, we fine-tuned DictaLM2.0-Instruct on a comprehensive instruct dataset to enhance its performance on task-specific instructions. To rigorously evaluate our models, we introduce a new benchmark suite for Hebrew LLM evaluation, covering a diverse set of tasks including Question Answering, Sentiment Analysis, Winograd Schema Challenge, Translation, and Summarization. Our work not only addresses the intricacies of training LLMs in low-resource languages but also proposes a framework that can be leveraged for adapting other LLMs to various non-English languages, contributing to the broader field of multilingual NLP.