Facilitating large language model Russian adaptation with Learned Embedding Propagation

TL;DR

This paper introduces Learned Embedding Propagation (LEP), a cost-effective method to adapt large language models to Russian without extensive instruction-tuning, maintaining high performance and reducing data requirements.

Contribution

LEP enables direct embedding of new language knowledge into existing models, bypassing traditional instruction-tuning and lowering data and computational costs.

Findings

LEP achieves performance comparable to instruction-tuning methods.

LEP improves Russian language adaptation in LLMs with minimal data.

Self-calibration further enhances task-solving capabilities.

Abstract

Rapid advancements of large language model (LLM) technologies led to the introduction of powerful open-source instruction-tuned LLMs that have the same text generation quality as the state-of-the-art counterparts such as GPT-4. While the emergence of such models accelerates the adoption of LLM technologies in sensitive-information environments the authors of such models don not disclose the training data necessary for replication of the results thus making the achievements model-exclusive. Since those open-source models are also multilingual this in turn reduces the benefits of training a language specific LLMs as improved inference computation efficiency becomes the only guaranteed advantage of such costly procedure. More cost-efficient options such as vocabulary extension and subsequent continued pre-training are also inhibited by the lack of access to high-quality instruction-tuning data since it is the major factor behind the resulting LLM task-solving capabilities. To address the limitations and cut the costs of the language adaptation pipeline we propose Learned Embedding Propagation (LEP). Unlike existing approaches our method has lower training data size requirements due to minimal impact on existing LLM knowledge which we reinforce using novel ad-hoc embedding propagation procedure that allows to skip the instruction-tuning step and instead implant the new language knowledge directly into any existing instruct-tuned variant. We evaluated four Russian vocabulary adaptations for LLaMa-3-8B and Mistral-7B, showing that LEP is competitive with traditional instruction-tuning methods, achieving performance comparable to OpenChat 3.5 and LLaMa-3-8B-Instruct, with further improvements via self-calibration and continued tuning enhancing task-solving capabilities.