Logo Lanfrica
  • Home
  • Atlas
  • Insights
  • Docs
  • Sign in

© 2026 Lanfrica. All rights reserved. All copyrights of the resources shown on the Lanfrica website belong to the original copyright holders, unless explicitly stated otherwise.

Cross-lingual Relation Extraction with Large Language Models: Zero-Shot, Few-Shot, and Fine-Tuned Evaluation on Romanian

Domain:

natural language processing

Record type:

datasetpaper
Creator:
VasApoTomPas
Publisher:
arXiv
Host:avatar
Relation extraction (RE) for low-resource languages is typically constrained by the lack of annotated corpora. We investigate the feasibility of cross-lingual RE for Romanian by combining automatic dataset translation with large language model (LLM) inference. We translate the SemEval-2010 Task 8 benchmark from English to Romanian using an LLM-based translation pipeline and evaluate Gemma 4 31B under zero-shot, few-shot, and QLoRA fine-tuned configurations, against four encoder baselines spanning 125M to 560M parameters: XLM- RoBERTa (base and large), Romanian BERT, and RoBERT- large. We assess two task formulations: relation classification with marked entities and end-to-end extraction. Our results show that Romanian incurs a 3 to 5 percentage point (pp) drop relative to English in prompt-only settings, that few-shot prompting provides marginal gains over zero-shot, and that QLoRA fine-tuning improves macro F1-Score by more than 22 percentage points in both languages while reducing the cross-lingual gap from 3.3 to 1.4pp. The encoder baselines come within 1-4pp of QLoRA Gemma on Romanian despite being 50-250 times smaller, with monolingual Romanian BERT at 125M parameters matching multilingual XLM-R at 278M. The case for using a 31B model for single-task RE on Romanian is therefore weak in deployment scenarios where compute matters. We release the translated dataset, evaluation code, and trained models.

Visit

doi.orgarxiv.org

Tasks

information extraction

Tags

Computation and Language (cs.CL)Artificial Intelligence (cs.AI)FOS: Computer and information sciences

Licenses

arXiv.org perpetual, non-exclusive licensehttp://arxiv.org/licenses/nonexclusive-distrib/1.0/

Similar

BUFFET: Benchmarking Large Language Models for Few-shot Cross-lingual TransferZero-Shot Cross-Lingual Reranking with Large Language Models for Low-Resource LanguagesComparative Performance of Teacher-Student Cross-Lingual NER Models Versus Zero-Shot and Few-Shot Approaches on XLENTFew-Shot Cross-Lingual Transfer for Prompting Large Language Models in Low-Resource LanguagesImproving Zero-Shot Cross-Lingual Hate Speech Detection with Pseudo-Label Fine-Tuning of Transformer Language ModelsTranslation and Fusion Improves Zero-shot Cross-lingual Information Extraction

BUFFET: Benchmarking Large Language Models for Few-shot Cross-lingual Transfer

Despite remarkable advancements in few-shot generalization in natural language processing, most mode

Zero-Shot Cross-Lingual Reranking with Large Language Models for Low-Resource Languages

Large language models (LLMs) have shown impressive zero-shot capabilities in various document rerank

Comparative Performance of Teacher-Student Cross-Lingual NER Models Versus Zero-Shot and Few-Shot Approaches on XLENT

Identifying user intents and their corresponding slots is the first step in the utterance interpreta

Few-Shot Cross-Lingual Transfer for Prompting Large Language Models in Low-Resource Languages

Large pre-trained language models (PLMs) are at the forefront of advances in Natural Language Proces

Improving Zero-Shot Cross-Lingual Hate Speech Detection with Pseudo-Label Fine-Tuning of Transformer Language Models

Hate speech has proliferated on social media platforms in recent years. While this has been the focu

Translation and Fusion Improves Zero-shot Cross-lingual Information Extraction

Large language models (LLMs) combined with instruction tuning have shown significant progress in inf