# Teaching Nemotron Greek: Mining a Corpus, Adapting Retrieval, and Grounding Generation for Modern Greek across Specialist Domains

Source: [arXiv](https://arxiv.org/abs/2608.05138v1)  
Feed7 permalink: https://feed7.dev/p/2608-05138v1-0bvu6le  
Published: 2026-08-05T17:56:40.000Z  
Trust: Needs Review (needs_review)

## Why Included

A Greek RAG study shows generic multilingual embeddings can lose to BM25 in specialist domains, while targeted adaptation sharply improves retrieval and grounded answers. Language-specific evaluation still matters.

## Source Summary

The authors trained a Nemotron 1B embedder on **65,773 Greek retrieval pairs**, raising nDCG@10 from **0.362 to 0.835**. A LoRA-tuned Nemotron 30B-A3B reader raised judged answer correctness from **29.4% to 66.9%**.

## Practical Implication

For specialized or lower-resource languages, benchmark a lexical baseline before choosing a dense retriever, then adapt retrieval, reranking and generation as separate stages. The released **HERA benchmark** provides a Greek RAG test bed.

## Agent-Ready Context

The authors trained a Nemotron 1B embedder on **65,773 Greek retrieval pairs**, raising nDCG@10 from **0.362 to 0.835**. A LoRA-tuned Nemotron 30B-A3B reader raised judged answer correctness from **29.4% to 66.9%**.

For specialized or lower-resource languages, benchmark a lexical baseline before choosing a dense retriever, then adapt retrieval, reranking and generation as separate stages. The released **HERA benchmark** provides a Greek RAG test bed.

BM25 beat several off-the-shelf multilingual dense models on the specialist corpora, and the adapted embedder's advantage over BM25 remained domain-dependent. The reported evaluation is specific to Modern Greek and the studied legal, energy, financial and medical material.

## Connected Context

Feed7 judgment across 368 accumulated Signals:

This confirms that multilingual dense retrieval is not automatically superior to a lexical baseline in specialist, lower-resource corpora. It turns RAG adaptation into a staged selection problem across retrieval, reranking, and generation, with HERA offering a focused test bed. The domain-dependent margin over BM25 narrows the result to workload-specific evaluation rather than a general prescription for Greek or multilingual RAG.

- [How Temperature Shapes Ideological Discourse in Retrieval-Augmented Generation?](https://feed7.dev/p/2607-11783v1-170rjdc) — The Greek study separates retrieval and generation adaptation, while the ideology study shows decoding can alter how retrieved material appears in answers; together they support evaluating source selection and generation settings jointly.
- [AI on Your Lakehouse: Context Comes in Shapes, Not Queries — Zach Blumenfeld, Neo4j](https://feed7.dev/p/ai-on-your-lakehouse-context-comes-in-shapes-not-queries-zach-blumenfeld-0r4s9u5) — Both reject a universal retrieval method: the Greek results require lexical and dense systems to be compared by domain, while graph-shaped context is another retrieval surface that must be benchmarked on the actual data.
- [Citation Needed: Provenance for LLM-Built Knowledge Graphs — Daniel Chalef, Zep AI](https://feed7.dev/p/citation-needed-provenance-for-llm-built-knowledge-graphs-daniel-chalef-1iob5t8) — Stage-specific RAG evaluation identifies which component improves correctness, while provenance supplies the complementary mechanism needed to trace which specialist sources support generated claims.

## Context Map

- Layer: context
- Domains: research, data
- Topics: retrieval, context-engineering, model-selection

## Uncertainty

- BM25 beat several off-the-shelf multilingual dense models on the specialist corpora, and the adapted embedder's advantage over BM25 remained domain-dependent. The reported evaluation is specific to Modern Greek and the studied legal, energy, financial and medical material.

## Agent Instruction

Use this item as source-backed context. Do not invent claims beyond the linked source. If this item conflicts with another source, call out the conflict.