{"@context":"https://schema.org","@type":"NewsArticle","generatedAt":"2026-07-23T06:40:50.084Z","headline":"RecGPT-V3 技术报告：状态化混合模态推荐系统在淘宝部署，IPV 提升 1.28%、GPU 成本降低 52.4%","description":"淘宝部署的 RecGPT-V3 采用状态化行为建模、混合模态基础模型（融合文本标签与 Semantic IDs）和潜在意图推理，将用户建模计算量降低 55.8%，输出 token 成本下降。在\"猜你喜欢\"信息流的大规模在线 A/B 测试中，IPV 提升 1.28%、CTR 提升 1.00%、TC 提升 1.97%、GMV 提升 3.97%，端到端服务资源消耗降低 52.4%。","url":"https://www.aioga.com/news/cmrsmh6lu0aj2biwm4bfip4jb/","mainEntityOfPage":"https://www.aioga.com/news/cmrsmh6lu0aj2biwm4bfip4jb/","datePublished":"2026-07-17T00:00:00.000Z","dateModified":"2026-07-17T00:00:00.000Z","inLanguage":"zh-CN","publisher":{"@type":"NewsMediaOrganization","name":"Aioga","url":"https://www.aioga.com"},"citation":["https://arxiv.org/abs/2607.15591","https://aihot.virxact.com/items/cmrsmh6lu0aj2biwm4bfip4jb"],"canonicalUrl":"https://www.aioga.com/news/cmrsmh6lu0aj2biwm4bfip4jb/","directAnswer":{"@type":"Answer","text":"RecGPT-V3 已在淘宝部署，通过状态化行为建模、融合文本标签与 Semantic IDs 的混合模态基础模型及潜在意图推理，降低用户建模计算量与输出 token 成本。摘要称，其在线 A/B 测试中 IPV、CTR、TC、GMV 分别提升 1.28%、1.00%、1.97%、3.97%。","url":"https://www.aioga.com/news/cmrsmh6lu0aj2biwm4bfip4jb/","dateCreated":"2026-07-17T00:00:00.000Z","author":{"@type":"Organization","@id":"https://www.aioga.com/authors/aioga-editorial/#editorial-team","name":"Aioga Editorial Team","url":"https://www.aioga.com/authors/aioga-editorial/"}},"evidence":[{"@type":"CreativeWork","name":"arXiv source article","url":"https://arxiv.org/abs/2607.15591","datePublished":"2026-07-17T00:00:00.000Z","provider":{"@type":"Organization","name":"arXiv","url":"https://arxiv.org/abs/2607.15591"}},{"@type":"CreativeWork","name":"AIHot archive record","url":"https://aihot.virxact.com/items/cmrsmh6lu0aj2biwm4bfip4jb","datePublished":"2026-07-17T00:00:00.000Z","provider":{"@type":"Organization","name":"AIHot","url":"https://aihot.virxact.com/items/cmrsmh6lu0aj2biwm4bfip4jb"}}],"aggregationSource":"HuggingFace Daily Papers（社区热门论文）","originalPublisher":{"name":"arXiv","url":"https://arxiv.org/abs/2607.15591"},"article":{"id":"cmrsmh6lu0aj2biwm4bfip4jb","slug":"cmrsmh6lu0aj2biwm4bfip4jb","url":"https://www.aioga.com/news/cmrsmh6lu0aj2biwm4bfip4jb/","title":"RecGPT-V3 技术报告：状态化混合模态推荐系统在淘宝部署，IPV 提升 1.28%、GPU 成本降低 52.4%","title_en":"RecGPT-V3 Technical Report","summary":"淘宝部署的 RecGPT-V3 采用状态化行为建模、混合模态基础模型（融合文本标签与 Semantic IDs）和潜在意图推理，将用户建模计算量降低 55.8%，输出 token 成本下降。在\"猜你喜欢\"信息流的大规模在线 A/B 测试中，IPV 提升 1.28%、CTR 提升 1.00%、TC 提升 1.97%、GMV 提升 3.97%，端到端服务资源消耗降低 52.4%。","source":"HuggingFace Daily Papers（社区热门论文）","sourceUrl":"https://arxiv.org/abs/2607.15591","aiHotUrl":"https://aihot.virxact.com/items/cmrsmh6lu0aj2biwm4bfip4jb","publishedAt":"2026-07-17T00:00:00.000Z","category":"论文研究","score":81,"selected":true,"articleBody":["arXivLabs is a framework that allows collaborators to develop and share new arXiv features directly on our website.","Have an idea for a project that will add value for arXiv's community? Learn more about arXivLabs ：https://info.arxiv.org/labs/index.html."],"articleImages":[{"sourceUrl":"https://arxiv.org/static/base/1.0.1/images/funders/simons-foundation.png","alt":"Simons Foundation","afterParagraph":1,"url":"/media/articles/cmrsmh6lu0aj2biwm4bfip4jb/e2d7f38d62f5ca91.png"},{"sourceUrl":"https://arxiv.org/static/base/1.0.1/images/funders/simons-foundation-international.png","alt":"Simons Foundation International","afterParagraph":1,"url":"/media/articles/cmrsmh6lu0aj2biwm4bfip4jb/1d56e29c5557cbdc.png"},{"sourceUrl":"https://arxiv.org/static/base/1.0.1/images/funders/schmidt-sciences.png","alt":"Schmidt Sciences","afterParagraph":1,"url":"/media/articles/cmrsmh6lu0aj2biwm4bfip4jb/8e18212b8219c104.png"}],"mediaStatus":"ok","articleBodyZh":["arXivLabs 是一个框架，允许合作者直接在我们的网站上开发和分享新的 arXiv 功能。","有一个可以为 arXiv 社区增加价值的项目想法吗？了解更多关于 arXivLabs 的信息：https://info.arxiv.org/labs/index.html。"],"translationStatus":"translated","bodyOrigin":"source-page","editorial":{"summary":"RecGPT-V3 已在淘宝部署，通过状态化行为建模、融合文本标签与 Semantic IDs 的混合模态基础模型及潜在意图推理，降低用户建模计算量与输出 token 成本。摘要称，其在线 A/B 测试中 IPV、CTR、TC、GMV 分别提升 1.28%、1.00%、1.97%、3.97%。","background":"该材料来自 HuggingFace Daily Papers 的社区热门论文，分类为论文研究。标题将其描述为状态化混合模态推荐系统；摘要说明测试场景为淘宝“猜你喜欢”信息流，但正文摘录仅介绍 arXivLabs，未补充论文方法或实验细节。","viewpoint":"Aioga 判断，这项工作的关注点不仅是推荐指标变化，还包括推理成本控制：摘要同时报告用户建模计算量降低 55.8%、端到端服务资源消耗降低 52.4%。不过，现有材料不足以核验实验周期、样本范围及统计显著性。","implications":"Aioga 判断，若论文完整结果与摘要一致，状态化建模与混合模态表示可能为大型推荐系统提供效果和资源消耗并行优化的参考。值得关注的是，标题所称“GPU 成本降低”与摘要中的“服务资源消耗降低”表述并不完全相同。","nextStep":"下一步应查阅论文全文，重点核对 IPV、CTR、TC 与 GMV 的定义，在线 A/B 测试设计、实验周期、流量范围和显著性结果，以及 52.4% 降幅对应的资源口径，从而判断这些结果的适用边界与可复现程度。","evidenceRefs":["title","summary","articleBody","source"],"status":"published","aiGenerated":true,"autoApproved":true,"generatedBy":"aioga-editorial:gpt-5.6-sol","reviewedBy":"aioga-editorial-review:gpt-5.6-sol","generatedAt":"2026-07-22T18:14:44.510Z","sourceHash":"2ff87e943751787c","review":{"approved":true,"groundedness":96,"clarity":94,"duplicationRisk":18,"blockingIssues":[],"notes":["候选内容准确区分了标题中的“GPU 成本降低”与摘要中的“端到端服务资源消耗降低”，避免将两种口径直接等同。","对实验周期、样本范围和统计显著性无法核验的说明符合现有材料边界。","summary 与 viewpoint 均列出部分指标和成本数据，存在轻微信息重复，但不影响可读性。"]},"validation":{"passed":true,"mode":"ai-auto","revisions":0,"checks":["schema","length","source-attribution","low-source-overlap","no-html","independent-ai-review"]}},"tags":["论文研究","HuggingFace Daily Papers（社区热门论文）"],"translations":{"zh-CN":{"title":"RecGPT-V3 技术报告：状态化混合模态推荐系统在淘宝部署，IPV 提升 1.28%、GPU 成本降低 52.4%","summary":"淘宝部署的 RecGPT-V3 采用状态化行为建模、混合模态基础模型（融合文本标签与 Semantic IDs）和潜在意图推理，将用户建模计算量降低 55.8%，输出 token 成本下降。在\"猜你喜欢\"信息流的大规模在线 A/B 测试中，IPV 提升 1.28%、CTR 提升 1.00%、TC 提升 1.97%、GMV 提升 3.97%，端到端服务资源消耗降低 52.4%。","category":"论文研究","source":"arXiv","aggregationSource":"HuggingFace Daily Papers（社区热门论文）","pageTitle":"RecGPT-V3 技术报告：状态化混合模态推荐系统在淘宝部署，IPV 提升 1.28%、GPU 成本降低 52.4% - Aioga AI资讯","description":"淘宝部署的 RecGPT-V3 采用状态化行为建模、混合模态基础模型（融合文本标签与 Semantic IDs）和潜在意图推理，将用户建模计算量降低 55.8%，输出 token 成本下降。在\"猜你喜欢\"信息流的大规模在线 A/B 测试中，IPV 提升 1.28%、CTR 提升 1.00%、TC 提升 1.97%、GMV 提升 3.97%，端到端服务资源消耗...","url":"https://www.aioga.com/news/cmrsmh6lu0aj2biwm4bfip4jb/"},"en":{"title":"RecGPT-V3 Technical Report: Deployment of a Stateful Hybrid Modal Recommendation System on Taobao, IPV Increased by 1.28%, GPU Cost Reduced by 52.4%","summary":"Taobao's deployed RecGPT-V3 uses stateful behavior modeling, a hybrid-modal base model (integrating text tags and Semantic IDs), and latent intent reasoning, reducing user modeling computation by 55.8% and output token cost. In large-scale online A/B testing of the 'You May Like' feed, IPV increased by 1.28%, CTR increased by 1.00%, TC increased by 1.97%, GMV increased by 3.97%, and end-to-end service resource consumption decreased by 52.4%.","category":"Research","source":"HuggingFace Daily Papers（社区热门论文）","aggregationSource":"HuggingFace Daily Papers（社区热门论文）","pageTitle":"RecGPT-V3 Technical Report: Deployment of a Stateful Hybrid Modal Recommendation System on Taobao, IPV Increased by 1.28%, GPU Cost Reduced by 52.4% - Aioga AI News","description":"Taobao's deployed RecGPT-V3 uses stateful behavior modeling, a hybrid-modal base model (integrating text tags and Semantic IDs), and latent intent reasoning, reducing user modeling...","url":"https://www.aioga.com/en/news/cmrsmh6lu0aj2biwm4bfip4jb/","contentTranslated":true,"sourceHash":"2a214efaf376579a","translatedAt":"2026-07-22T16:45:57.879Z"},"ja":{"title":"RecGPT-V3 技術報告：状態化ハイブリッドモーダル推薦システムの淘宝への展開、IPV 1.28%向上、GPU コスト 52.4%削減","summary":"淘宝が展開する RecGPT-V3 は、状態化行動モデリング、ハイブリッドモーダル基礎モデル（テキストラベルとセマンティックIDの融合）、潜在意図推論を採用し、ユーザーモデリングの計算量を 55.8% 削減し、出力トークンのコストを低下させました。「あなたへのおすすめ」フィードの大規模オンライン A/B テストでは、IPV が 1.28% 向上、CTR が 1.00% 向上、TC が 1.97% 向上、GMV が 3.97% 向上し、エンドツーエンドのサービスリソース消費を 52.4% 低減しました。","category":"論文研究","source":"HuggingFace Daily Papers（社区热门论文）","aggregationSource":"HuggingFace Daily Papers（社区热门论文）","pageTitle":"RecGPT-V3 技術報告：状態化ハイブリッドモーダル推薦システムの淘宝への展開、IPV 1.28%向上、GPU コスト 52.4%削減 - Aioga AIニュース","description":"淘宝が展開する RecGPT-V3 は、状態化行動モデリング、ハイブリッドモーダル基礎モデル（テキストラベルとセマンティックIDの融合）、潜在意図推論を採用し、ユーザーモデリングの計算量を 55.8% 削減し、出力トークンのコストを低下させました。「あなたへのおすすめ」フィードの大規模オンライン A/B テストでは、IPV が 1.28% 向上、CTR が...","url":"https://www.aioga.com/ja/news/cmrsmh6lu0aj2biwm4bfip4jb/","contentTranslated":true,"sourceHash":"2a214efaf376579a","translatedAt":"2026-07-22T16:46:09.165Z"},"ko":{"title":"RecGPT-V3 기술 보고서: 상태화 혼합 모달 추천 시스템을 타오바오에 배치, IPV 1.28% 증가, GPU 비용 52.4% 감소","summary":"타오바오에 배포된 RecGPT-V3는 상태 기반 행동 모델링, 혼합 모달 기초 모델(텍스트 태그와 Semantic IDs 결합) 및 잠재 의도 추론을 채택하여 사용자 모델링 계산량을 55.8% 감소시키고, 출력 토큰 비용을 낮췄습니다. '당신이 좋아할 만한' 정보 피드의 대규모 온라인 A/B 테스트에서 IPV는 1.28%, CTR은 1.00%, TC는 1.97%, GMV는 3.97% 상승했으며, 엔드투엔드 서비스 자원 소모는 52.4% 감소했습니다.","category":"연구","source":"HuggingFace Daily Papers（社区热门论文）","aggregationSource":"HuggingFace Daily Papers（社区热门论文）","pageTitle":"RecGPT-V3 기술 보고서: 상태화 혼합 모달 추천 시스템을 타오바오에 배치, IPV 1.28% 증가, GPU 비용 52.4% 감소 - Aioga AI 뉴스","description":"타오바오에 배포된 RecGPT-V3는 상태 기반 행동 모델링, 혼합 모달 기초 모델(텍스트 태그와 Semantic IDs 결합) 및 잠재 의도 추론을 채택하여 사용자 모델링 계산량을 55.8% 감소시키고, 출력 토큰 비용을 낮췄습니다. '당신이 좋아할 만한' 정보 피드의 대규모 온라인 A/B 테스트에서 IPV는 1.28%...","url":"https://www.aioga.com/ko/news/cmrsmh6lu0aj2biwm4bfip4jb/","contentTranslated":true,"sourceHash":"2a214efaf376579a","translatedAt":"2026-07-22T16:46:57.298Z"},"es":{"title":"Informe técnico de RecGPT-V3: Despliegue del sistema de recomendación multimodal estatuado en Taobao, aumento del IPV en 1,28%, reducción del costo de GPU en 52,4%","summary":"RecGPT-V3, desplegado por Taobao, utiliza modelado de comportamiento con estado, un modelo base multimodal híbrido (que integra etiquetas de texto e IDs semánticas) y razonamiento de intención latente, reduciendo la carga computacional del modelado de usuarios en un 55,8% y disminuyendo el costo de los tokens de salida. En pruebas A/B en línea a gran escala del flujo de información 'Te podría gustar', el IPV aumentó un 1,28%, el CTR un 1,00%, el TC un 1,97% y el GMV un 3,97%, mientras que el consumo de recursos del servicio de extremo a extremo se redujo en un 52,4%.","category":"Investigación","source":"HuggingFace Daily Papers（社区热门论文）","aggregationSource":"HuggingFace Daily Papers（社区热门论文）","pageTitle":"Informe técnico de RecGPT-V3: Despliegue del sistema de recomendación multimodal estatuado en Taobao, aumento del IPV en 1,28%, reducción del costo de GPU en 52,4% - Aioga Noticias de IA","description":"RecGPT-V3, desplegado por Taobao, utiliza modelado de comportamiento con estado, un modelo base multimodal híbrido (que integra etiquetas de texto e IDs semánticas) y razonamiento...","url":"https://www.aioga.com/es/news/cmrsmh6lu0aj2biwm4bfip4jb/","contentTranslated":true,"sourceHash":"2a214efaf376579a","translatedAt":"2026-07-22T16:46:55.839Z"},"fr":{"title":"Rapport technique RecGPT-V3 : Déploiement du système de recommandation multimodal à états chez Taobao, augmentation de l'IPV de 1,28 % et réduction des coûts GPU de 52,4 %","summary":"RecGPT-V3 déployé par Taobao utilise la modélisation du comportement état-par-état, un modèle de base multimodal mixte (fusionnant des étiquettes textuelles et des IDs sémantiques) et le raisonnement sur les intentions latentes, réduisant le volume de calcul du modèle utilisateur de 55,8 % et le coût des tokens de sortie. Dans les tests A/B en ligne à grande échelle du flux d'informations \"Vous pourriez aimer\", l'IPV a augmenté de 1,28 %, le CTR de 1,00 %, le TC de 1,97 %, le GMV de 3,97 %, et la consommation de ressources des services de bout en bout a diminué de 52,4 %.","category":"Recherche","source":"HuggingFace Daily Papers（社区热门论文）","aggregationSource":"HuggingFace Daily Papers（社区热门论文）","pageTitle":"Rapport technique RecGPT-V3 : Déploiement du système de recommandation multimodal à états chez Taobao, augmentation de l'IPV de 1,28 % et réduction des coûts GPU de 52,4 % - Aioga Actualités IA","description":"RecGPT-V3 déployé par Taobao utilise la modélisation du comportement état-par-état, un modèle de base multimodal mixte (fusionnant des étiquettes textuelles et des IDs sémantiques)...","url":"https://www.aioga.com/fr/news/cmrsmh6lu0aj2biwm4bfip4jb/","contentTranslated":true,"sourceHash":"2a214efaf376579a","translatedAt":"2026-07-22T16:47:47.067Z"},"de":{"title":"RecGPT-V3 Technischer Bericht: Zustandsbasierte hybride Modalität Empfehlungssystemeinsatz bei Taobao, IPV-Steigerung um 1,28 %, GPU-Kostenreduktion um 52,4 %","summary":"RecGPT-V3, das von Taobao eingesetzt wird, verwendet zustandsbasierte Verhaltensmodellierung, hybride multimodale Grundmodelle (die Text-Tags und semantische IDs kombinieren) und latente Intentionslogik, wodurch der Rechenaufwand für die Benutzermodellierung um 55,8 % reduziert und die Token-Ausgabekosten gesenkt werden. In groß angelegten Online-A/B-Tests im \"Für Sie empfohlen\"-Feed stiegen IPV um 1,28 %, CTR um 1,00 %, TC um 1,97 % und GMV um 3,97 %, während der Gesamtressourcenverbrauch des End-to-End-Dienstes um 52,4 % gesenkt wurde.","category":"论文研究","source":"HuggingFace Daily Papers（社区热门论文）","aggregationSource":"HuggingFace Daily Papers（社区热门论文）","pageTitle":"RecGPT-V3 Technischer Bericht: Zustandsbasierte hybride Modalität Empfehlungssystemeinsatz bei Taobao, IPV-Steigerung um 1,28 %, GPU-Kostenreduktion um 52,4 % - Aioga KI-News","description":"RecGPT-V3, das von Taobao eingesetzt wird, verwendet zustandsbasierte Verhaltensmodellierung, hybride multimodale Grundmodelle (die Text-Tags und semantische IDs kombinieren) und l...","url":"https://www.aioga.com/de/news/cmrsmh6lu0aj2biwm4bfip4jb/","contentTranslated":true,"sourceHash":"2a214efaf376579a","translatedAt":"2026-07-22T16:47:47.183Z"},"pt-BR":{"title":"Relatório Técnico RecGPT-V3: Sistema de Recomendação Multimodal Estatal Implantado no Taobao, Aumento de 1,28% no IPV, Redução de 52,4% no Custo de GPU","summary":"O RecGPT-V3 implantado pelo Taobao utiliza modelagem de comportamento com estado, modelo base multimodal híbrido (integrando etiquetas de texto e IDs semânticos) e inferência de intenção latente, reduzindo a carga de cálculo do modelo de usuário em 55,8% e diminuindo o custo de tokens de saída. Nos testes online em grande escala tipo A/B no feed de \"Você pode gostar\", o IPV aumentou 1,28%, o CTR aumentou 1,00%, o TC aumentou 1,97%, o GMV aumentou 3,97% e o consumo de recursos do serviço de ponta a ponta caiu 52,4%.","category":"论文研究","source":"HuggingFace Daily Papers（社区热门论文）","aggregationSource":"HuggingFace Daily Papers（社区热门论文）","pageTitle":"Relatório Técnico RecGPT-V3: Sistema de Recomendação Multimodal Estatal Implantado no Taobao, Aumento de 1,28% no IPV, Redução de 52,4% no Custo de GPU - Aioga Notícias de IA","description":"O RecGPT-V3 implantado pelo Taobao utiliza modelagem de comportamento com estado, modelo base multimodal híbrido (integrando etiquetas de texto e IDs semânticos) e inferência de in...","url":"https://www.aioga.com/pt-BR/news/cmrsmh6lu0aj2biwm4bfip4jb/","contentTranslated":true,"sourceHash":"2a214efaf376579a","translatedAt":"2026-07-22T16:48:33.152Z"},"ru":{"title":"Технический отчет RecGPT-V3: Развертывание рекомендательной системы с комбинированной модальностью состояния на Taobao, рост IPV на 1,28%, снижение затрат на GPU на 52,4%","summary":"RecGPT-V3, развернутый на Taobao, использует моделирование поведенческих состояний, гибридную модальную базовую модель (объединяющую текстовые метки и Semantic IDs) и вывод скрытых намерений, что снижает вычислительную нагрузку на моделирование пользователя на 55,8% и снижает стоимость output token. В масштабном онлайн A/B тестировании в информационном потоке \"Вам может понравиться\" IPV увеличился на 1,28%, CTR — на 1,00%, TC — на 1,97%, GMV — на 3,97%, а потребление ресурсов на всем сервисе снизилось на 52,4%.","category":"论文研究","source":"HuggingFace Daily Papers（社区热门论文）","aggregationSource":"HuggingFace Daily Papers（社区热门论文）","pageTitle":"Технический отчет RecGPT-V3: Развертывание рекомендательной системы с комбинированной модальностью состояния на Taobao, рост IPV на 1,28%, снижение затрат на GPU на 52,4% - Aioga Новости ИИ","description":"RecGPT-V3, развернутый на Taobao, использует моделирование поведенческих состояний, гибридную модальную базовую модель (объединяющую текстовые метки и Semantic IDs) и вывод скрытых...","url":"https://www.aioga.com/ru/news/cmrsmh6lu0aj2biwm4bfip4jb/","contentTranslated":true,"sourceHash":"2a214efaf376579a","translatedAt":"2026-07-22T16:48:33.052Z"},"ar":{"title":"تقرير تقني RecGPT-V3: نشر نظام التوصية متعدد النماذج المدمج ذو الحالة في Taobao، زيادة IPV بنسبة 1.28٪، وانخفاض تكلفة GPU بنسبة 52.4٪","summary":"تعتمد RecGPT-V3 المنشورة على Taobao على نمذجة السلوك الحالة، ونموذج أساسي متعدد الوسائط المدمج (يجمع بين تسميات النصوص وSemantic IDs) واستدلال النية الكامنة، مما يقلل من حجم الحساب اللازم لنمذجة المستخدم بنسبة 55.8%، ويخفض تكلفة مخرجات الرموز. في اختبار A/B واسع النطاق عبر الإنترنت لتدفق المعلومات \"قد يعجبك\"، ارتفع IPV بنسبة 1.28%، وارتفع CTR بنسبة 1.00%، وارتفع TC بنسبة 1.97%، وزاد GMV بنسبة 3.97%، بينما انخفض استهلاك موارد الخدمة الشاملة بنسبة 52.4%.","category":"论文研究","source":"HuggingFace Daily Papers（社区热门论文）","aggregationSource":"HuggingFace Daily Papers（社区热门论文）","pageTitle":"تقرير تقني RecGPT-V3: نشر نظام التوصية متعدد النماذج المدمج ذو الحالة في Taobao، زيادة IPV بنسبة 1.28٪، وانخفاض تكلفة GPU بنسبة 52.4٪ - Aioga أخبار الذكاء الاصطناعي","description":"تعتمد RecGPT-V3 المنشورة على Taobao على نمذجة السلوك الحالة، ونموذج أساسي متعدد الوسائط المدمج (يجمع بين تسميات النصوص وSemantic IDs) واستدلال النية الكامنة، مما يقلل من حجم الحساب...","url":"https://www.aioga.com/ar/news/cmrsmh6lu0aj2biwm4bfip4jb/","contentTranslated":true,"sourceHash":"2a214efaf376579a","translatedAt":"2026-07-22T16:49:22.431Z"},"hi":{"title":"RecGPT-V3 तकनीकी रिपोर्ट: टाओबाओ में स्थिति-संगत मिश्रित मोड सिफारिश प्रणाली की तैनाती, IPV 1.28% बढ़ा, GPU लागत 52.4% कम हुई","summary":"Taobao द्वारा तैनात RecGPT-V3 स्टेटफुल व्यवहार मॉडलिंग, हाइब्रिड मोडाल बेस मॉडल (टेक्स्ट टैग और सेमांटिक ID को मर्ज करके) और संभावित इरादे की तर्कशीलता का उपयोग करता है, जिससे उपयोगकर्ता मॉडलिंग की गणना 55.8% घट जाती है और आउटपुट टोकन लागत कम हो जाती है। \"आपको पसंद आ सकता है\" इनफोस्ट्रीम के बड़े पैमाने पर ऑनलाइन A/B टेस्ट में, IPV 1.28%, CTR 1.00%, TC 1.97%, GMV 3.97% बढ़ गया, और एंड-टू-एंड सेवा संसाधन की खपत 52.4% कम हो गई।","category":"论文研究","source":"HuggingFace Daily Papers（社区热门论文）","aggregationSource":"HuggingFace Daily Papers（社区热门论文）","pageTitle":"RecGPT-V3 तकनीकी रिपोर्ट: टाओबाओ में स्थिति-संगत मिश्रित मोड सिफारिश प्रणाली की तैनाती, IPV 1.28% बढ़ा, GPU लागत 52.4% कम हुई - Aioga AI समाचार","description":"Taobao द्वारा तैनात RecGPT-V3 स्टेटफुल व्यवहार मॉडलिंग, हाइब्रिड मोडाल बेस मॉडल (टेक्स्ट टैग और सेमांटिक ID को मर्ज करके) और संभावित इरादे की तर्कशीलता का उपयोग करता है, जिससे उपयो...","url":"https://www.aioga.com/hi/news/cmrsmh6lu0aj2biwm4bfip4jb/","contentTranslated":true,"sourceHash":"2a214efaf376579a","translatedAt":"2026-07-22T16:49:21.474Z"},"it":{"title":"Relazione tecnica RecGPT-V3: sistema di raccomandazione multimodale stato-centrato implementato su Taobao, aumento dell'IPV del 1,28%, riduzione dei costi GPU del 52,4%","summary":"RecGPT-V3, implementato da Taobao, utilizza la modellazione del comportamento stato-dipendente, un modello di base multimodale ibrido (che integra etichette testuali e Semantic ID) e il ragionamento sulle intenzioni latenti, riducendo il carico computazionale del modello utente del 55,8% e il costo dei token in output. Nei test A/B online su larga scala del feed \"Ti potrebbe piacere\", l'IPV è aumentato dell'1,28%, il CTR dell'1,00%, il TC dell'1,97%, il GMV del 3,97%, mentre il consumo di risorse del servizio end-to-end è diminuito del 52,4%.","category":"论文研究","source":"HuggingFace Daily Papers（社区热门论文）","aggregationSource":"HuggingFace Daily Papers（社区热门论文）","pageTitle":"Relazione tecnica RecGPT-V3: sistema di raccomandazione multimodale stato-centrato implementato su Taobao, aumento dell'IPV del 1,28%, riduzione dei costi GPU del 52,4% - Aioga Notizie IA","description":"RecGPT-V3, implementato da Taobao, utilizza la modellazione del comportamento stato-dipendente, un modello di base multimodale ibrido (che integra etichette testuali e Semantic ID)...","url":"https://www.aioga.com/it/news/cmrsmh6lu0aj2biwm4bfip4jb/","contentTranslated":true,"sourceHash":"2a214efaf376579a","translatedAt":"2026-07-22T16:50:13.541Z"},"nl":{"title":"RecGPT-V3 technisch rapport: Gesta­teerde hybride multimodale aanbevelingssystemen geïmplementeerd op Taobao, IPV verhoogd met 1,28%, GPU-kosten verlaagd met 52,4%","summary":"RecGPT-V3, ingezet door Taobao, maakt gebruik van statusgebaseerde gedragsmodellering, een hybride multimodale basismodel (dat tekstlabels en semantische ID's combineert) en latent intentieredenken, waarmee de berekeningsbelasting van gebruikersmodellen met 55,8% wordt verminderd en de outputtokenkosten dalen. In grootschalige online A/B-tests van de 'Aan jou aanbevolen'-informatiestroom steeg IPV met 1,28%, CTR met 1,00%, TC met 1,97% en GMV met 3,97%, terwijl het verbruik van end-to-end serverbronnen met 52,4% werd verminderd.","category":"论文研究","source":"HuggingFace Daily Papers（社区热门论文）","aggregationSource":"HuggingFace Daily Papers（社区热门论文）","pageTitle":"RecGPT-V3 technisch rapport: Gesta­teerde hybride multimodale aanbevelingssystemen geïmplementeerd op Taobao, IPV verhoogd met 1,28%, GPU-kosten verlaagd met 52,4% - Aioga AI-nieuws","description":"RecGPT-V3, ingezet door Taobao, maakt gebruik van statusgebaseerde gedragsmodellering, een hybride multimodale basismodel (dat tekstlabels en semantische ID's combineert) en latent...","url":"https://www.aioga.com/nl/news/cmrsmh6lu0aj2biwm4bfip4jb/","contentTranslated":true,"sourceHash":"2a214efaf376579a","translatedAt":"2026-07-22T16:50:08.059Z"},"tr":{"title":"RecGPT-V3 Teknik Raporu: Durum Odaklı Hibrit Modlu Öneri Sistemi Taobao'da Devreye Alındı, IPV %1,28 Arttı, GPU Maliyeti %52,4 Azaldı","summary":"Taobao tarafından dağıtılan RecGPT-V3, durum tabanlı davranış modellemesi, karma modlu temel model (metin etiketleri ve Semantic ID’lerin birleşimi) ve gizli niyet çıkarımını kullanarak, kullanıcı modelleme hesaplama yükünü %55,8 oranında azaltmış ve token maliyetini düşürmüştür. \"Sana hoş gelebilir\" bilgi akışının büyük ölçekli çevrimiçi A/B testlerinde, IPV %1,28, CTR %1,00, TC %1,97, GMV %3,97 artarken, uçtan uca hizmet kaynak tüketimi %52,4 azalmıştır.","category":"论文研究","source":"HuggingFace Daily Papers（社区热门论文）","aggregationSource":"HuggingFace Daily Papers（社区热门论文）","pageTitle":"RecGPT-V3 Teknik Raporu: Durum Odaklı Hibrit Modlu Öneri Sistemi Taobao'da Devreye Alındı, IPV %1,28 Arttı, GPU Maliyeti %52,4 Azaldı - Aioga AI Haberleri","description":"Taobao tarafından dağıtılan RecGPT-V3, durum tabanlı davranış modellemesi, karma modlu temel model (metin etiketleri ve Semantic ID’lerin birleşimi) ve gizli niyet çıkarımını kulla...","url":"https://www.aioga.com/tr/news/cmrsmh6lu0aj2biwm4bfip4jb/","contentTranslated":true,"sourceHash":"2a214efaf376579a","translatedAt":"2026-07-22T16:51:09.619Z"},"vi":{"title":"Báo cáo kỹ thuật RecGPT-V3: Hệ thống gợi ý đa phương thức có trạng thái được triển khai trên Taobao, tăng 1,28% IPV, giảm 52,4% chi phí GPU","summary":"RecGPT-V3 do Taobao triển khai sử dụng mô hình hóa hành vi có trạng thái, mô hình cơ bản đa mô thức (kết hợp nhãn văn bản và Semantic IDs) và suy luận ý định tiềm ẩn, làm giảm 55,8% khối lượng tính toán mô hình người dùng và giảm chi phí token đầu ra. Trong thử nghiệm A/B trực tuyến quy mô lớn của luồng thông tin \"Có thể bạn thích\", IPV tăng 1,28%, CTR tăng 1,00%, TC tăng 1,97%, GMV tăng 3,97%, mức tiêu thụ tài nguyên dịch vụ đầu cuối giảm 52,4%.","category":"论文研究","source":"HuggingFace Daily Papers（社区热门论文）","aggregationSource":"HuggingFace Daily Papers（社区热门论文）","pageTitle":"Báo cáo kỹ thuật RecGPT-V3: Hệ thống gợi ý đa phương thức có trạng thái được triển khai trên Taobao, tăng 1,28% IPV, giảm 52,4% chi phí GPU - Tin tức AI Aioga","description":"RecGPT-V3 do Taobao triển khai sử dụng mô hình hóa hành vi có trạng thái, mô hình cơ bản đa mô thức (kết hợp nhãn văn bản và Semantic IDs) và suy luận ý định tiềm ẩn, làm giảm 55,8...","url":"https://www.aioga.com/vi/news/cmrsmh6lu0aj2biwm4bfip4jb/","contentTranslated":true,"sourceHash":"2a214efaf376579a","translatedAt":"2026-07-22T16:51:05.878Z"},"id":{"title":"Laporan Teknis RecGPT-V3: Sistem Rekomendasi Multimodal Terstatus yang Diterapkan di Taobao, Peningkatan IPV 1,28%, Penurunan Biaya GPU 52,4%","summary":"RecGPT-V3 yang diterapkan oleh Taobao menggunakan pemodelan perilaku berstatus, model dasar multimodal campuran (menggabungkan label teks dan Semantic IDs) serta inferensi niat tersembunyi, mengurangi beban komputasi pemodelan pengguna sebesar 55,8% dan menurunkan biaya token keluaran. Dalam uji A/B online skala besar pada aliran informasi 'Rekomendasi untuk Anda', IPV meningkat 1,28%, CTR meningkat 1,00%, TC meningkat 1,97%, GMV meningkat 3,97%, dan konsumsi sumber daya layanan end-to-end turun 52,4%.","category":"论文研究","source":"HuggingFace Daily Papers（社区热门论文）","aggregationSource":"HuggingFace Daily Papers（社区热门论文）","pageTitle":"Laporan Teknis RecGPT-V3: Sistem Rekomendasi Multimodal Terstatus yang Diterapkan di Taobao, Peningkatan IPV 1,28%, Penurunan Biaya GPU 52,4% - Berita AI Aioga","description":"RecGPT-V3 yang diterapkan oleh Taobao menggunakan pemodelan perilaku berstatus, model dasar multimodal campuran (menggabungkan label teks dan Semantic IDs) serta inferensi niat ter...","url":"https://www.aioga.com/id/news/cmrsmh6lu0aj2biwm4bfip4jb/","contentTranslated":true,"sourceHash":"2a214efaf376579a","translatedAt":"2026-07-22T16:51:54.501Z"},"th":{"title":"รายงานทางเทคนิค RecGPT-V3: ระบบแนะนำแบบหลายโหมดสภาพสถานะถูกนำไปใช้ใน Taobao, เพิ่ม IPV 1.28%, ลดค่าใช้จ่าย GPU 52.4%","summary":"RecGPT-V3 ของ Taobao ที่ปรับใช้ ใช้การสร้างแบบจำลองพฤติกรรมเป็นสถานะ, โมเดลพื้นฐานแบบมัลติฟังก์ชันผสม (รวมป้ายข้อความและ Semantic IDs) และการอนุมานความตั้งใจแฝง ทำให้ลดปริมาณการคำนวณการสร้างแบบจำลองผู้ใช้ลง 55.8% และลดต้นทุน token ในกระบวนการส่งออก ในการทดสอบ A/B ออนไลน์ขนาดใหญ่ของฟีดข้อมูล \"猜你喜欢\" (คุณอาจชอบ) IPV เพิ่มขึ้น 1.28%, CTR เพิ่มขึ้น 1.00%, TC เพิ่มขึ้น 1.97%, GMV เพิ่มขึ้น 3.97%, และการใช้ทรัพยากรบริการแบบ end-to-end ลดลง 52.4%.","category":"论文研究","source":"HuggingFace Daily Papers（社区热门论文）","aggregationSource":"HuggingFace Daily Papers（社区热门论文）","pageTitle":"รายงานทางเทคนิค RecGPT-V3: ระบบแนะนำแบบหลายโหมดสภาพสถานะถูกนำไปใช้ใน Taobao, เพิ่ม IPV 1.28%, ลดค่าใช้จ่าย GPU 52.4% - ข่าว AI Aioga","description":"RecGPT-V3 ของ Taobao ที่ปรับใช้ ใช้การสร้างแบบจำลองพฤติกรรมเป็นสถานะ, โมเดลพื้นฐานแบบมัลติฟังก์ชันผสม (รวมป้ายข้อความและ Semantic IDs) และการอนุมานความตั้งใจแฝง ทำให้ลดปริมาณการคำน...","url":"https://www.aioga.com/th/news/cmrsmh6lu0aj2biwm4bfip4jb/","contentTranslated":true,"sourceHash":"2a214efaf376579a","translatedAt":"2026-07-22T16:52:03.582Z"},"pl":{"title":"Raport techniczny RecGPT-V3: System rekomendacji o mieszanej modalności ze stanem wdrożony na Taobao, zwiększenie IPV o 1,28%, obniżenie kosztów GPU o 52,4%","summary":"RecGPT-V3 wdrożony przez Taobao wykorzystuje modelowanie zachowań w stanie, hybrydowy model modalny (łączący etykiety tekstowe i identyfikatory semantyczne) oraz wnioskowanie o ukrytych intencjach, zmniejszając obciążenie obliczeniowe modelowania użytkownika o 55,8% i obniżając koszt tokenów wyjściowych. W dużej skali testu online A/B w strumieniu informacji \"Może Ci się spodobać\", IPV wzrosło o 1,28%, CTR o 1,00%, TC o 1,97%, GMV o 3,97%, a zużycie zasobów usługi end-to-end zmniejszyło się o 52,4%.","category":"论文研究","source":"HuggingFace Daily Papers（社区热门论文）","aggregationSource":"HuggingFace Daily Papers（社区热门论文）","pageTitle":"Raport techniczny RecGPT-V3: System rekomendacji o mieszanej modalności ze stanem wdrożony na Taobao, zwiększenie IPV o 1,28%, obniżenie kosztów GPU o 52,4% - Aioga Wiadomości AI","description":"RecGPT-V3 wdrożony przez Taobao wykorzystuje modelowanie zachowań w stanie, hybrydowy model modalny (łączący etykiety tekstowe i identyfikatory semantyczne) oraz wnioskowanie o ukr...","url":"https://www.aioga.com/pl/news/cmrsmh6lu0aj2biwm4bfip4jb/","contentTranslated":true,"sourceHash":"2a214efaf376579a","translatedAt":"2026-07-22T16:52:57.180Z"}}}}