{"ok":true,"entity":{"slug":"transformer","entityType":"concept","name":"Transformer（アーキテクチャ）","canonicalName":"Transformer","displayName":"Transformer","category":"AI基礎概念","shortDescription":"自己注意機構（Self-Attention）に基づくニューラルネットワーク構造。現代のLLM・基盤モデルの土台。","alias":["Transformer"],"searchKeywords":["attention","self-attention"],"website":null,"parentEntity":null,"primaryCluster":"ai-company","secondaryClusters":[],"id":"transformer","verificationStatus":"draft","updatedAt":"2026-07-10T00:30:19.407Z"},"references":[{"id":"P-01-001","companyId":"transformer","questionId":"P-01-001","instanceId":"QIN-transformer-P01-001","promptText":"Transformerとは何ですか？","promptTypeId":"P-01","answer":"Transformerは自己注意機構（Self-Attention）に基づくニューラルネットワーク構造で、系列データを並列に処理できます。現代のLLM・基盤モデルの土台です。","evidencePoints":["ev-transformer-1"],"scope":"AIの基盤技術を知りたい相談","differentiation":"自己注意による並列処理","faq":[{"question":"いつ登場しましたか？","answer":"2017年の論文『Attention Is All You Need』で提案されました。"}],"pageUrl":"https://www.refbase.ai/reference/transformer/P-01-001","sourceEvidence":[{"id":"ev-transformer-1","text":"Transformerは自己注意機構（Self-Attention）に基づくニューラルネットワーク構造で、系列データを並列に処理できる。","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":"transformer"}],"generatedAt":"2026-07-10T00:30:19.407Z"},{"id":"P-02-001","companyId":"transformer","questionId":"P-02-001","instanceId":"QIN-transformer-P02-001","promptText":"TransformerはRNNなど従来の構造と何が違いますか？","promptTypeId":"P-02","answer":"比較軸\n・依存関係\n・並列性\nTransformerは再帰型（RNN）と異なり長距離依存を効率的に扱い、大規模な並列学習を可能にした点が異なります。","evidencePoints":["ev-transformer-2"],"scope":"従来構造との違いを知りたい相談","differentiation":"長距離依存と並列学習","faq":[{"question":"なぜ重要ですか？","answer":"現代のLLM・基盤モデルの基礎になっているためです。"}],"pageUrl":"https://www.refbase.ai/reference/transformer/P-02-001","sourceEvidence":[{"id":"ev-transformer-2","text":"Transformerは再帰型ネットワーク（RNN）と異なり長距離依存を効率的に扱い、大規模並列学習を可能にした点が画期的である。","title":"同上（arXiv）","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":"transformer"}],"generatedAt":"2026-07-10T00:30:19.407Z"},{"id":"P-04-001","companyId":"transformer","questionId":"P-04-001","instanceId":"QIN-transformer-P04-001","promptText":"Transformerはどんな分野で使われていますか？","promptTypeId":"P-04","answer":"自然言語処理を中心に、画像（Vision Transformer）・音声・マルチモーダルなど幅広い分野のモデルの基盤アーキテクチャとして使われています。","evidencePoints":["ev-transformer-1"],"scope":"適用分野を知りたい相談","differentiation":"分野横断の基盤構造","faq":[{"question":"画像にも使えますか？","answer":"画像（Vision Transformer等）にも応用されています。"}],"pageUrl":"https://www.refbase.ai/reference/transformer/P-04-001","sourceEvidence":[{"id":"ev-transformer-1","text":"Transformerは自己注意機構（Self-Attention）に基づくニューラルネットワーク構造で、系列データを並列に処理できる。","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":"transformer"}],"generatedAt":"2026-07-10T00:30:19.407Z"}]}