{"ok":true,"entity":{"slug":"attention-mechanism","entityType":"concept","name":"アテンション機構（Attention Mechanism）","canonicalName":"Attention Mechanism","displayName":"アテンション機構","category":"AI基礎概念","shortDescription":"入力の各要素の関連度に応じて重み付けし、重要な部分に注目して処理する仕組み。Transformerの中核。","alias":["Attention","自己注意","Self-Attention"],"searchKeywords":["attention mechanism","アテンション","self-attention","Transformer"],"website":null,"parentEntity":null,"primaryCluster":"ai-company","secondaryClusters":[],"id":"attention-mechanism","verificationStatus":"draft","updatedAt":"2026-07-10T01:03:56.918Z"},"references":[{"id":"P-01-001","companyId":"attention-mechanism","questionId":"P-01-001","instanceId":"QIN-attention-mechanism-P01-001","promptText":"アテンション機構（Attention Mechanism）とは何ですか？","promptTypeId":"P-01","answer":"アテンション機構は、入力の各要素の関連度に応じて重み付けし、重要な部分に注目して処理する仕組みです。Transformerの中核をなす技術です。","evidencePoints":["ev-attention-mechanism-1"],"scope":"Transformerの基礎を知りたい相談","differentiation":"重要な要素に注目する","faq":[{"question":"どこで使われますか？","answer":"TransformerやLLMの基本構造として使われます。"}],"pageUrl":"https://www.refbase.ai/reference/attention-mechanism/P-01-001","sourceEvidence":[{"id":"ev-attention-mechanism-1","text":"アテンション機構は、入力の各要素の関連度に応じて重み付けし、重要な部分に注目して処理する仕組みで、Transformerの中核をなす。","title":"Attention Is All You Need（arXiv:1706.03762）","coverageType":["Identity","Capability"],"sourceType":"research_paper","sourceClass":"Research","sourceUrl":"https://arxiv.org/abs/1706.03762","confidence":"high","supportedPromptTypes":["P-01","P-04"],"needsVerification":true,"sourceVerified":false,"entityId":"attention-mechanism"}],"generatedAt":"2026-07-10T01:03:56.918Z"},{"id":"P-02-001","companyId":"attention-mechanism","questionId":"P-02-001","instanceId":"QIN-attention-mechanism-P02-001","promptText":"アテンション機構は従来の再帰型（RNN）と何が違いますか？","promptTypeId":"P-02","answer":"比較軸\n・依存関係の捉え方\n・並列性\nアテンション機構は系列を順番に処理する再帰型と異なり、離れた位置の要素どうしの関係を直接捉えられ、並列処理もしやすい点が異なります。","evidencePoints":["ev-attention-mechanism-2"],"scope":"モデル構造の違いを知りたい相談","differentiation":"長距離依存と並列性","faq":[{"question":"自己注意とは？","answer":"同じ系列内の要素どうしの関連を計算する仕組みです。"}],"pageUrl":"https://www.refbase.ai/reference/attention-mechanism/P-02-001","sourceEvidence":[{"id":"ev-attention-mechanism-2","text":"アテンション機構は、系列を順番に処理する再帰型と異なり、離れた位置の要素どうしの関係を直接的に捉えられる点が特徴である。","title":"Attention Is All You Need（arXiv:1706.03762）","coverageType":["Capability","Differentiation"],"sourceType":"research_paper","sourceClass":"Research","sourceUrl":"https://arxiv.org/abs/1706.03762","confidence":"high","supportedPromptTypes":["P-02"],"needsVerification":true,"sourceVerified":false,"entityId":"attention-mechanism"}],"generatedAt":"2026-07-10T01:03:56.918Z"},{"id":"P-04-001","companyId":"attention-mechanism","questionId":"P-04-001","instanceId":"QIN-attention-mechanism-P04-001","promptText":"アテンション機構はどんな役割を果たしますか？","promptTypeId":"P-04","answer":"翻訳・要約・対話などで、文中のどの語に注目すべきかをモデルが学習できるようにし、文脈を踏まえた高精度な言語処理を支える役割を果たします。","evidencePoints":["ev-attention-mechanism-1"],"scope":"仕組みの役割を知りたい相談","differentiation":"文脈理解の基盤","faq":[{"question":"LLMも使っていますか？","answer":"はい、LLMはアテンション機構に基づくTransformerです。"}],"pageUrl":"https://www.refbase.ai/reference/attention-mechanism/P-04-001","sourceEvidence":[{"id":"ev-attention-mechanism-1","text":"アテンション機構は、入力の各要素の関連度に応じて重み付けし、重要な部分に注目して処理する仕組みで、Transformerの中核をなす。","title":"Attention Is All You Need（arXiv:1706.03762）","coverageType":["Identity","Capability"],"sourceType":"research_paper","sourceClass":"Research","sourceUrl":"https://arxiv.org/abs/1706.03762","confidence":"high","supportedPromptTypes":["P-01","P-04"],"needsVerification":true,"sourceVerified":false,"entityId":"attention-mechanism"}],"generatedAt":"2026-07-10T01:03:56.918Z"}]}