Teaching Nemotron Greek: Mining a Corpus, Adapting Retrieval, and Grounding Generation for Modern Greek across Specialist Domains
2026-08-05 • Artificial Intelligence
Artificial IntelligenceComputation and Language
AI summaryⓘ
The authors adapted NVIDIA's Nemotron retrieval system to work with Modern Greek, which was previously unsupported despite its importance in various fields. They created a Greek dataset, trained retrieval and reading models, and developed a new benchmark called HERA to evaluate performance. Their results showed that simple BM25 methods initially outperformed general multilingual models, but fine-tuned Nemotron models performed much better on Greek texts. They also improved answer accuracy and citation quality for Greek question answering. Finally, they released their models and benchmark to help others build better Greek language retrieval systems.
Modern GreekNemotron retrievalBM25nDCG@10rerankerreader fine-tuningretrieval-augmented generationcross-encoderLoRA-tuningbenchmark
Authors
Ayoub Kirouane, Christos Petrocheilos
Abstract
Modern Greek is absent from NVIDIA's Nemotron retrieval models and from major multilingual retrieval benchmarks, despite being important for retrieval-augmented generation (RAG) in legal, energy, financial, and medical applications. We present an end-to-end adaptation of the Nemotron retrieval stack for Modern Greek, including corpus mining, synthetic supervision, retrieval model training, reranker adaptation, reader fine-tuning, and a new benchmark called HERA. Our study shows that a parameter-free BM25 baseline outperforms several off-the-shelf multilingual dense retrieval models on specialist Greek corpora. After fine-tuning on 65,773 Greek retrieval pairs, a Nemotron 1B embedder improves nDCG@10 from 0.362 to 0.835 and substantially outperforms its unadapted counterpart. The learned language competence transfers to general-domain Greek, although the advantage over BM25 remains domain-dependent. We further adapt a cross-encoder reranker and demonstrate consistent improvements across specialist domains. Finally, we LoRA-tune a Nemotron 30B-A3B mixture-of-experts reader for grounded generation, increasing judged answer correctness from 29.4% to 66.9% while significantly improving faithfulness and citation quality. We also introduce HERA, the first large-scale Greek benchmark for retrieval-augmented generation, and release our adapted models and benchmark to support future research on Greek-language RAG systems.