Translate Hugging Face blog post: Granite Embedding Multilingual R2: Open Apache 2.0 Multilingual Embeddings with 32K Context — Best Sub-100M Retrieval Quality

[Jwaminju]

Source: https://huggingface.co/blog/ibm-granite/granite-embedding-multilingual-r2

This PR adds a Korean translation draft for granite-embedding-multilingual-r2.

Downstream handoff:

• SEO review should use the translation-flow manifest.
• Quality review should use the translation-flow manifest.

[github-actions]

PR Preview Action v1.8.1
🚀 View preview at https://hugging-face-krew.github.io/pr-preview/pr-132/
Built to branch gh-pages at 2026-05-18 12:31 UTC. Preview will be ready when the GitHub Pages deployment is complete.

[github-actions]

PR Preview Action v1.8.1
🚀 View preview at https://hugging-face-krew.github.io/pr-preview/pr-132/
Built to branch gh-pages at 2026-07-28 13:59 UTC. Preview will be ready when the GitHub Pages deployment is complete.

[github-actions]

HF Agent Review

Gate	Result
Quality	✅ Pass
SEO	✅ Pass

Head SHA: dfc15a7566f60c77d247e5e3530411e1041e050a

Quality report — ✅ Pass

Quality Report

• Status: review_required
• Quality Score: 84.0
• Hard failures: 0
• Issues: 7
• Source available: True
• Source changed: False
• Source segments: 8
• Target segments: 81

Scorecard

Dimension	Score
adequacy	55.0
technical_accuracy	100.0
completeness	80.0
terminology	100.0
fluency	100.0
publishing_integrity	100.0
style_locale	60.0

Metrics

• qe_metric: heuristic
• cache_hits: 0
• cache_misses: 0
• warning: Source was fetched as HTML text; structural hard gates, segment coverage, and segment metrics were skipped.

MQM Judge

• Enabled: True
• Provider: openai
• Model: gpt-5.6-luna
• Reasoning effort: none
• Prompt: /home/runner/work/hugging-face-krew.github.io/hugging-face-krew.github.io/workflow/skills/quality/judges/mqm_prompt.md
• Requested segments: 0
• Evaluated segments: 0
• MQM errors: 0
• Cache hits: 0
• Cache misses: 0
• warning: MQM judge skipped: no source/target segment alignment is available.
• warning: MQM judge skipped: source was fetched as HTML text, not comparable Markdown segments.

Style Guide

• Enabled: True
• Guide: /home/runner/work/hugging-face-krew.github.io/hugging-face-krew.github.io/workflow/skills/quality/style/hf-blog-ko-translation-guide.md
• Policy: /home/runner/work/hugging-face-krew.github.io/hugging-face-krew.github.io/workflow/skills/quality/configs/style_policy.yml
• Style score: 60.0
• Rule hits: {'list_consistency': 1, 'modal_strength': 5}

Style Guide Findings

Rule	Severity	Segment	Current	Suggested
list_consistency	minor		phrase, sentence, sentence, sentence, sentence, phrase, sentence, sentence, sentence, sentence, sentence, sentence, phrase, phrase, sentence, phrase, sentence, sentence, sentence, sentence, phrase, phrase	Use either sentence-style endings or phrase-style endings consistently within one list.
modal_strength	major	h_001	Granite Embedding Multilingual R2: 32K 다국어 임베딩	Preserve the strength of may using: 수 있습니다, 일 수 있습니다.
modal_strength	major	h_001	Granite Embedding Multilingual R2: 32K 다국어 임베딩	Preserve the strength of can using: 수 있습니다.
modal_strength	major	h_001	Granite Embedding Multilingual R2: 32K 다국어 임베딩	Preserve the strength of should using: 좋습니다, 해야 합니다.
modal_strength	major	h_001	Granite Embedding Multilingual R2: 32K 다국어 임베딩	Preserve the strength of up to using: 최대.
modal_strength	major	p_008	두 모델 모두 200+개 언어를 지원하고, 52개 언어 및 프로그래밍 코드에 대해 향상된 검색 품질을 제공합니다. 또한 최대 32,768 토큰의 컨텍스트를 처리할 수 있으며(이전 모델 대비 64배 증가), Apache 2.0 라이선스로 배포됩니다. 기본적으로 sentence-transformers와 transformers에서 바로 작동하며, 과제별 특별한 지시가 필요 없고 LangChain, LlamaIndex, Haystack, Milvus에서 모델 이름 한 줄의 변경으로 드롭인 대체로 사용할 수 있습니다. 현재 영어 전용 기본을 사용하는 프레임워크의 경우 한 줄 변경으로 커뮤니티의 모든 사용자가 200+개 언어를 지원할 수 있습니다 — API 변경, 새로운 의존성, 또는 endpoint 코드 변경이 필요 없습니다. 두 모델 모두 CPU 최적화 추론용 ONNX 및 OpenVINO 가중치를 함께 제공합니다.	Preserve the strength of should using: 좋습니다, 해야 합니다.

Issues

QL-001 style_locale / minor

• Message: List mixes sentence-style and phrase-style endings.
• Target: phrase, sentence, sentence, sentence, sentence, phrase, sentence, sentence, sentence, sentence, sentence, sentence, phrase, phrase, sentence, phrase, sentence, sentence, sentence, sentence, phrase, phrase
• Suggested fix: Use either sentence-style endings or phrase-style endings consistently within one list.
• Reason: The style guide requires consistent list item endings.

QL-002 accuracy / major

• Message: Modal or certainty strength may have changed.
• Source: may
• Target: Granite Embedding Multilingual R2: 32K 다국어 임베딩
• Suggested fix: Preserve the strength of may using: 수 있습니다, 일 수 있습니다.
• Reason: The style guide requires preserving may/can/should/must/up to/in some cases/not always strength.

QL-003 accuracy / major

• Message: Modal or certainty strength may have changed.
• Source: can
• Target: Granite Embedding Multilingual R2: 32K 다국어 임베딩
• Suggested fix: Preserve the strength of can using: 수 있습니다.
• Reason: The style guide requires preserving may/can/should/must/up to/in some cases/not always strength.

QL-004 accuracy / major

• Message: Modal or certainty strength may have changed.
• Source: should
• Target: Granite Embedding Multilingual R2: 32K 다국어 임베딩
• Suggested fix: Preserve the strength of should using: 좋습니다, 해야 합니다.
• Reason: The style guide requires preserving may/can/should/must/up to/in some cases/not always strength.

QL-005 accuracy / major

• Message: Modal or certainty strength may have changed.
• Source: up to
• Target: Granite Embedding Multilingual R2: 32K 다국어 임베딩
• Suggested fix: Preserve the strength of up to using: 최대.
• Reason: The style guide requires preserving may/can/should/must/up to/in some cases/not always strength.

QL-006 accuracy / major

• Message: Modal or certainty strength may have changed.
• Source: should
• Target: 두 모델 모두 200+개 언어를 지원하고, 52개 언어 및 프로그래밍 코드에 대해 향상된 검색 품질을 제공합니다. 또한 최대 32,768 토큰의 컨텍스트를 처리할 수 있으며(이전 모델 대비 64배 증가), Apache 2.0 라이선스로 배포됩니다. 기본적으로 sentence-transformers와 transformers에서 바로 작동하며, 과제별 특별한 지시가 필요 없고 LangChain, LlamaIndex, Haystack, Milvus에서 모델 이름 한 줄의 변경으로 드롭인 대체로 사용할 수 있습니다. 현재 영어 전용 기본을 사용하는 프레임워크의 경우 한 줄 변경으로 커뮤니티의 모든 사용자가 200+개 언어를 지원할 수 있습니다 — API 변경, 새로운 의존성, 또는 endpoint 코드 변경이 필요 없습니다. 두 모델 모두 CPU 최적화 추론용 ONNX 및 OpenVINO 가중치를 함께 제공합니다.
• Suggested fix: Preserve the strength of should using: 좋습니다, 해야 합니다.
• Reason: The style guide requires preserving may/can/should/must/up to/in some cases/not always strength.

QL-007 accuracy / major

• Message: Semantic adequacy evaluation is incomplete.
• Suggested fix: Complete MQM evaluation for every aligned segment before auto-passing.
• Reason: Heuristic QE only checks surface signals and cannot establish semantic equivalence.

SEO report — ✅ Pass

SEO Eval Report

Gate: ✅ PASS — deterministic AND rubric

• File: ../target/_posts/2026-05-15-granite-embedding-multilingual-r2.md
• Source: —
• Primary keyword: (none — D5 skipped)
• Mode: file

Gate

• Status: PASS
• Blockers: ✅ pass
• Deterministic REQUIRED (D1–D7): ✅ pass
• Rubric (R1–R6): ✅ pass (mean None, min None)

Blockers

✅ body_not_empty: Body is not empty
✅ robots_indexable: Robots is indexable
✅ internal_links_resolve: All internal links resolve
✅ local_images_resolve: All local images resolve

Required checks (gated)

✅ heading_hierarchy: Heading hierarchy: Valid

OpenAI rubric checks

✅ semantic_metadata: PASS (required) — Semantic metadata across title, description, rendered H1, headings, and opening text are aligned; no contradictions detected.
✅ alt_semantics: PASS (review) — —

Advisory checks (not gated)

✅ opening_summary: Opening 3 paragraphs: 104 words (recommend ≥50 for GEO)
✅ h1_count: Markdown H1 count: 1 (review against rendered layout)
✅ citations: Citations/statistics: 19 (recommend ≥1 for GEO)
✅ question_headings: Scannable H2/H3 (question or keyword): 2 (2 question, 0 keyword)
✅ internal_links: Internal links: 12 (recommend 2-3)
✅ word_count: Word count: 2669 (recommend ≥300)
ℹ️ primary_keyword: No primary_keyword in manifest — keyword check skipped
ℹ️ no_images: No images found (optional)

Signals (evidence — not directly gated)

• Frontmatter: title 46 chars, description 81 chars, author present True
• Title text: Granite Embedding Multilingual R2: 32K 다국어 임베딩
• Description text: IBM Granite Embedding Multilingual R2의 32K 컨텍스트, 다국어 검색 성능, Matryoshka 지원을 정리합니다.
• Opening text: * TOC
  {:toc}

_이 글은 Hugging Face 블로그의 [Granite Embedding Multilingual R2: Open Apache 2.0 Multilingual Embeddings with 32K Context — Best Sub-100M Retrieval Quality](https://huggingface.co/blog/ibm-granite/granite-embedding-multil

• Opening: first paragraph 267 chars, first 3 paragraphs 610 chars
• Headings: markdown H1 1, rendered effective H1 2, layout title H1 True
• Links: total 30, external 18, internal 0, citation signals 19
• Images: total 0, empty alt 0, filename-like alt 0, missing local files 0

Semantic review packet

• Title: Granite Embedding Multilingual R2: 32K 다국어 임베딩
• Description: IBM Granite Embedding Multilingual R2의 32K 컨텍스트, 다국어 검색 성능, Matryoshka 지원을 정리합니다.
• Rendered H1 candidates: Granite Embedding Multilingual R2: 32K 다국어 임베딩, Granite Embedding Multilingual R2: 32K 다국어 임베딩
• Opening: * TOC
  {:toc}

이 글은 Hugging Face 블로그의 Granite Embedding Multilingual R2: Open Apache 2.0 Multilingual Embeddings with 32K Context — Best Sub-100M Retrieval Quality를 한국어로 번역한 글입니다.

• Canonical/permalink: —
• Instruction: Compare title, description, rendered H1, and opening text for meaning consistency. This packet is evidence only; it does not decide pass/fail.

Frontmatter (advisory — written by metadata step, not gated)

✅ title: Title: 46 chars (recommend ≤60)
✅ description: Description: 81 chars (semantic quality reviewed separately)
✅ image: OG image: assets/images/blog/posts/2026-05-15-granite-embedding-multilingual-r2/thumbnail.png
✅ categories: Categories: 2 (recommend 2-3)
✅ author: Author: dailybot

SEO metadata suggestion — PARTIAL

This is a suggestion. SEO is applied only when the post frontmatter is updated.
To apply safe fields from a partial suggestion, leave a trusted PR comment: metadata apply.

• Auto apply: False
• Requires human: True
• Mode: frontmatter_only
• Reason: metadata candidate needs policy decisions or missing title/description

Candidate

• title: Granite Embedding Multilingual R2: 32K 다국어 임베딩
• description: IBM Granite Embedding Multilingual R2의 32K 컨텍스트, 다국어 검색 성능, Matryoshka 지원을 정리합니다.
• categories: ['Translation', 'HuggingFace']
• image: assets/images/blog/posts/2026-05-15-granite-embedding-multilingual-r2/thumbnail.png

Needs policy decision

• target_url
• source_url
• canonical_policy
• translation_indexing
• target_locale
• source_locale

Warnings

• 본 메타데이터는 frontmatter 정보를 기반으로 하며, 해당 글이 Hugging Face 블로그의 글을 한국어로 번역한 콘텐츠임을 나타냅니다.
• source_url이 비어 있어 원문 URL이 확인되지 않습니다.