Pretraining and Benchmarking Modern Encoders for Latvian

TL;DR

This paper pretrains and benchmarks Latvian-specific encoder models using recent transformer architectures, demonstrating competitive performance and releasing resources to advance Latvian NLP research and applications.

Contribution

It introduces Latvian-specific encoder models based on modern architectures and evaluates their performance across multiple benchmarks, filling a gap in low-resource language NLP.

Findings

lv-deberta-base outperforms larger multilingual models

Models are competitive with existing Latvian encoders

Resources are publicly released for further research

Abstract

Encoder-only transformers remain essential for practical NLP tasks. While recent advances in multilingual models have improved cross-lingual capabilities, low-resource languages such as Latvian remain underrepresented in pretraining corpora, and few monolingual Latvian encoders currently exist. We address this gap by pretraining a suite of Latvian-specific encoders based on RoBERTa, DeBERTaV3, and ModernBERT architectures, including long-context variants, and evaluating them across a diverse set of Latvian diagnostic and linguistic benchmarks. Our models are competitive with existing monolingual and multilingual encoders while benefiting from recent architectural and efficiency advances. Our best model, lv-deberta-base (111M parameters), achieves the strongest overall performance, outperforming larger multilingual baselines and prior Latvian-specific encoders. We release all pretrained models and evaluation resources to support further research and practical applications in Latvian NLP.