{"@context":"https://schema.org","@type":"NewsArticle","generatedAt":"2026-08-24T04:40:54.192Z","headline":"扩展分类流映射（Categorical Flow Maps）规模","description":"连续扩散与流匹配模型有望成为语言建模中自回归方法的有力替代，可解锁加速采样与倾斜等连续模态优势。近期研究通过高斯分布与one-hot编码数据分布间的简单流匹配过程，实现离散数据的连续生成，并借助分类流映射（CFMs）验证了加速采样的可行性，样本质量具有竞争力。 🔗 阅读原文 via AIHOT · https://aihot.virxact.com/items/cmsj4jrne24d5ronkvb40rucl","url":"https://www.aioga.com/news/cmsj4jrne24d5ronkvb40rucl/","mainEntityOfPage":"https://www.aioga.com/news/cmsj4jrne24d5ronkvb40rucl/","datePublished":"2026-08-07T00:00:00.000Z","dateModified":"2026-08-07T00:00:00.000Z","inLanguage":"zh-CN","publisher":{"@type":"NewsMediaOrganization","name":"Aioga","url":"https://www.aioga.com"},"citation":["https://machinelearning.apple.com/research/scaling-categorical-flow-maps","https://aihot.virxact.com/items/cmsj4jrne24d5ronkvb40rucl"],"canonicalUrl":"https://www.aioga.com/news/cmsj4jrne24d5ronkvb40rucl/","directAnswer":{"@type":"Answer","text":"材料摘要称，连续扩散与流匹配模型可能成为语言建模中自回归方法的替代路径；分类流映射通过高斯分布与独热编码数据分布之间的流匹配，实现离散数据的连续生成，并展示加速采样的可行性。","url":"https://www.aioga.com/news/cmsj4jrne24d5ronkvb40rucl/","dateCreated":"2026-08-07T00:00:00.000Z","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":"machinelearning.apple.com source article","url":"https://machinelearning.apple.com/research/scaling-categorical-flow-maps","datePublished":"2026-08-07T00:00:00.000Z","provider":{"@type":"Organization","name":"machinelearning.apple.com","url":"https://machinelearning.apple.com/research/scaling-categorical-flow-maps"}},{"@type":"CreativeWork","name":"AIHot archive record","url":"https://aihot.virxact.com/items/cmsj4jrne24d5ronkvb40rucl","datePublished":"2026-08-07T00:00:00.000Z","provider":{"@type":"Organization","name":"AIHot","url":"https://aihot.virxact.com/items/cmsj4jrne24d5ronkvb40rucl"}}],"aggregationSource":"Apple Machine Learning Research（RSS）","originalPublisher":{"name":"machinelearning.apple.com","url":"https://machinelearning.apple.com/research/scaling-categorical-flow-maps"},"geoDeepAnswer":null,"article":{"id":"cmsj4jrne24d5ronkvb40rucl","slug":"cmsj4jrne24d5ronkvb40rucl","url":"https://www.aioga.com/news/cmsj4jrne24d5ronkvb40rucl/","title":"扩展分类流映射（Categorical Flow Maps）规模","title_en":"","summary":"连续扩散与流匹配模型有望成为语言建模中自回归方法的有力替代，可解锁加速采样与倾斜等连续模态优势。近期研究通过高斯分布与one-hot编码数据分布间的简单流匹配过程，实现离散数据的连续生成，并借助分类流映射（CFMs）验证了加速采样的可行性，样本质量具有竞争力。 🔗 阅读原文 via AIHOT · https://aihot.virxact.com/items/cmsj4jrne24d5ronkvb40rucl","source":"Apple Machine Learning Research（RSS）","sourceUrl":"https://machinelearning.apple.com/research/scaling-categorical-flow-maps","aiHotUrl":"https://aihot.virxact.com/items/cmsj4jrne24d5ronkvb40rucl","publishedAt":"2026-08-07T00:00:00.000Z","category":"行业动态","score":72,"selected":true,"articleBody":["Authors Oscar Davis†**, Anastasiia Filippova, Victor Turrisi, Amitis Shidani, Pierre Ablin, Marco Cuturi, Louis Béthune","View publication：https://arxiv.org/abs/2605.07820","Score Distillation of Flow Matching Models","December 16, 2025 research area Computer Vision：/research/?domain=Computer%20Vision, research area Methods and Algorithms：/research/?domain=Methods%20and%20Algorithms","Diffusion models achieve high-quality image generation but are limited by slow iterative sampling. Distillation methods alleviate this by enabling one- or few-step generation. Flow matching, originally introduced as a distinct framework, has since been shown to be theoretically equivalent to diffusion under Gaussian assumptions, raising the question of whether distillation techniques such as score distillation transfer directly. We provide a…","CAR-Flow: Condition-Aware Reparameterization Aligns Source and Target for Better Flow Matching","November 12, 2025 research area Computer Vision：/research/?domain=Computer%20Vision conference NeurIPS：/research/?event=NeurIPS","Conditional generative modeling aims to learn a conditional data distribution from samples containing data-condition pairs. For this, diffusion and flow-based methods have attained compelling results. These methods use a learned (flow) model to transport an initial standard Gaussian noise that ignores the condition to the conditional data distribution. The model is hence required to learn both mass transport and conditional injection. To ease the…","Our research in machine learning breaks new ground every day."],"articleImages":[{"sourceUrl":"https://mlr.cdn-apple.com/media/Discover_1440x420_2x_9c465d585e.jpg","alt":"Bottom banner","afterParagraph":7,"url":"/media/articles/cmsj4jrne24d5ronkvb40rucl/64c2324784f2cf67.jpg"}],"mediaStatus":"ok","articleBodyZh":["作者 Oscar Davis†**、Anastasiia Filippova、Victor Turrisi、Amitis Shidani、Pierre Ablin、Marco Cuturi、Louis Béthune","查看出版物：https://arxiv.org/abs/2605.07820","流匹配模型的分数蒸馏","2025年12月16日 研究领域 计算机视觉：/research/?domain=Computer Vision，研究领域 方法与算法：/research/?domain=Methods and Algorithms","扩散模型能够生成高质量图像，但受限于慢速迭代采样。蒸馏方法通过实现一步或少步生成缓解了这一问题。流匹配最初被引入作为一个独立框架，但随后已被证明在高斯假设下与扩散理论上等价，因此提出了一个问题：诸如分数蒸馏的蒸馏技术是否可以直接迁移。我们提供了一项…","CAR-Flow：条件感知重参数化对齐源与目标以实现更好的流匹配","2025年11月12日 研究领域 计算机视觉：/research/?domain=Computer Vision 会议 NeurIPS：/research/?event=NeurIPS","条件生成建模旨在从包含数据-条件对的样本中学习条件数据分布。为此，扩散和基于流的方法已经取得了令人信服的成果。这些方法使用学习到的（流）模型将忽略条件的初始标准高斯噪声传输到条件数据分布。因此，模型需要同时学习质量传输和条件注入。为了简化…","我们的机器学习研究每天都在开辟新领域。"],"translationStatus":"translated","bodyOrigin":"source-page","editorial":{"summary":"材料摘要称，连续扩散与流匹配模型可能成为语言建模中自回归方法的替代路径；分类流映射通过高斯分布与独热编码数据分布之间的流匹配，实现离散数据的连续生成，并展示加速采样的可行性。","background":"分类流映射关注如何将面向连续空间的生成方法用于离散数据。现有材料提到加速采样和具有竞争力的样本质量，但正文摘录主要展示其他流匹配研究条目，未提供该论文的方法细节、实验设置或量化结果。","viewpoint":"Aioga 判断，这项工作的关注点在于把连续生成框架扩展至离散建模，并探索减少采样步骤的可能性。不过，当前材料不足以验证其相对自回归方法的实际优势，也无法判断扩展规模、计算成本及适用边界。","implications":"若原论文证实分类流映射在更大规模下仍保持样本质量与采样效率，该方向可能为离散生成模型提供新的技术选择。值得关注的是，现有摘要只表述了可行性与竞争力，不能据此推断其已具备部署优势。","nextStep":"后续应核对原始论文，重点确认模型规模、数据集、评价指标、采样步数、对照基线及计算开销，并查明标题所称“扩展规模”的具体含义。在这些信息补齐前，不宜对性能提升幅度或应用前景作确定结论。","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":"2026-08-12T01:22:27.726Z","sourceHash":"c99fb3d8b0df5959","review":{"approved":true,"groundedness":96,"clarity":93,"duplicationRisk":15,"blockingIssues":[],"notes":["候选内容准确区分了来源事实、编辑判断与条件性推论，没有把潜在优势表述为已证实结论。","“Aioga 判断”属于明确标注的观点，不构成观点冒充事实；如需增强可追溯性，可补充该判断的作者或编辑来源。","正文摘录包含其他流匹配研究条目，候选内容已说明这些材料不提供目标论文的方法细节、实验设置或量化结果。"]},"validation":{"passed":true,"mode":"ai-auto","revisions":0,"checks":["schema","length","source-attribution","low-source-overlap","no-html","independent-ai-review"]}},"tags":["行业动态","Apple Machine Learning Research（RSS）"],"translations":{"zh-CN":{"title":"扩展分类流映射（Categorical Flow Maps）规模","summary":"连续扩散与流匹配模型有望成为语言建模中自回归方法的有力替代，可解锁加速采样与倾斜等连续模态优势。近期研究通过高斯分布与one-hot编码数据分布间的简单流匹配过程，实现离散数据的连续生成，并借助分类流映射（CFMs）验证了加速采样的可行性，样本质量具有竞争力。 🔗 阅读原文 via AIHOT · https://aihot.virxact.com/items/cmsj4jrne24d5ronkvb40rucl","category":"行业动态","source":"machinelearning.apple.com","aggregationSource":"Apple Machine Learning Research（RSS）","pageTitle":"扩展分类流映射（Categorical Flow Maps）规模 - Aioga AI资讯","description":"连续扩散与流匹配模型有望成为语言建模中自回归方法的有力替代，可解锁加速采样与倾斜等连续模态优势。近期研究通过高斯分布与one-hot编码数据分布间的简单流匹配过程，实现离散数据的连续生成，并借助分类流映射（CFMs）验证了加速采样的可行性，样本质量具有竞争力。 🔗 阅读原文 via AIHOT · https://aihot.virxact.com/item...","url":"https://www.aioga.com/news/cmsj4jrne24d5ronkvb40rucl/","articleBody":["作者 Oscar Davis†**、Anastasiia Filippova、Victor Turrisi、Amitis Shidani、Pierre Ablin、Marco Cuturi、Louis Béthune","查看出版物：https://arxiv.org/abs/2605.07820","流匹配模型的分数蒸馏","2025年12月16日 研究领域 计算机视觉：/research/?domain=Computer Vision，研究领域 方法与算法：/research/?domain=Methods and Algorithms","扩散模型能够生成高质量图像，但受限于慢速迭代采样。蒸馏方法通过实现一步或少步生成缓解了这一问题。流匹配最初被引入作为一个独立框架，但随后已被证明在高斯假设下与扩散理论上等价，因此提出了一个问题：诸如分数蒸馏的蒸馏技术是否可以直接迁移。我们提供了一项…","CAR-Flow：条件感知重参数化对齐源与目标以实现更好的流匹配","2025年11月12日 研究领域 计算机视觉：/research/?domain=Computer Vision 会议 NeurIPS：/research/?event=NeurIPS","条件生成建模旨在从包含数据-条件对的样本中学习条件数据分布。为此，扩散和基于流的方法已经取得了令人信服的成果。这些方法使用学习到的（流）模型将忽略条件的初始标准高斯噪声传输到条件数据分布。因此，模型需要同时学习质量传输和条件注入。为了简化…","我们的机器学习研究每天都在开辟新领域。"]},"en":{"title":"Scale of Categorical Flow Maps","summary":"Continuous diffusion and flow-matching models are expected to become strong alternatives to autoregressive methods in language modeling, unlocking the advantages of continuous modalities such as accelerated sampling and skewing. Recent research has achieved continuous generation of discrete data through a simple flow-matching process between Gaussian distributions and one-hot encoded data distributions, and validated the feasibility of accelerated sampling using classification flow mapping (CFMs), with competitive sample quality. 🔗 Read the original via AIHOT · https://aihot.virxact.com/items/cmsj4jrne24d5ronkvb40rucl","category":"Industry","source":"Apple Machine Learning Research（RSS）","aggregationSource":"Apple Machine Learning Research（RSS）","pageTitle":"Scale of Categorical Flow Maps - Aioga AI News","description":"Continuous diffusion and flow-matching models are expected to become strong alternatives to autoregressive methods in language modeling, unlocking the advantages of continuous moda...","url":"https://www.aioga.com/en/news/cmsj4jrne24d5ronkvb40rucl/","contentTranslated":true,"sourceHash":"fdeda6806310cc7d","translatedAt":"2026-08-10T11:16:23.937Z"},"ja":{"title":"カテゴリカルフローマップのスケール","summary":"連続拡散とフローマッチングモデルは、言語モデリングにおける自己回帰法の有力な代替手段となる可能性があり、サンプリングの高速化や連続モードの利点を引き出すことができます。最近の研究では、ガウス分布と one-hot エンコードデータ分布間のシンプルなフローマッチングプロセスを通じて、離散データの連続生成を実現し、分類フローマッピング（CFMs）を用いてサンプリング高速化の実現可能性を検証しており、サンプル品質も競争力があります。 🔗 原文を読む via AIHOT · https://aihot.virxact.com/items/cmsj4jrne24d5ronkvb40rucl","category":"業界動向","source":"Apple Machine Learning Research（RSS）","aggregationSource":"Apple Machine Learning Research（RSS）","pageTitle":"カテゴリカルフローマップのスケール - Aioga AIニュース","description":"連続拡散とフローマッチングモデルは、言語モデリングにおける自己回帰法の有力な代替手段となる可能性があり、サンプリングの高速化や連続モードの利点を引き出すことができます。最近の研究では、ガウス分布と one-hot エンコードデータ分布間のシンプルなフローマッチングプロセスを通じて、離散データの連続生成を実現し、分類フローマッピング（CFMs）を用いてサンプリ...","url":"https://www.aioga.com/ja/news/cmsj4jrne24d5ronkvb40rucl/","contentTranslated":true,"sourceHash":"fdeda6806310cc7d","translatedAt":"2026-08-10T11:16:25.652Z"},"ko":{"title":"범주형 흐름도 규모","summary":"연속 확산 및 흐름 매칭 모델은 언어 모델링에서 자기 회귀(Self-Regressive) 방법의 강력한 대안이 될 가능성이 있으며, 가속 샘플링과 기울기 등 연속 모달 장점을 활용할 수 있습니다. 최근 연구에서는 가우시안 분포와 원-핫(one-hot) 인코딩 데이터 분포 간의 단순한 흐름 매칭 과정을 통해 이산 데이터를 연속적으로 생성할 수 있음을 실현했으며, 분류 흐름 매핑(CFM)을 통해 가속 샘플링의 실현 가능성을 검증하고, 샘플 품질이 경쟁력이 있음을 입증했습니다. 🔗 원문 읽기 via AIHOT · https://aihot.virxact.com/items/cmsj4jrne24d5ronkvb40rucl","category":"업계 동향","source":"Apple Machine Learning Research（RSS）","aggregationSource":"Apple Machine Learning Research（RSS）","pageTitle":"범주형 흐름도 규모 - Aioga AI 뉴스","description":"연속 확산 및 흐름 매칭 모델은 언어 모델링에서 자기 회귀(Self-Regressive) 방법의 강력한 대안이 될 가능성이 있으며, 가속 샘플링과 기울기 등 연속 모달 장점을 활용할 수 있습니다. 최근 연구에서는 가우시안 분포와 원-핫(one-hot) 인코딩 데이터 분포 간의 단순한 흐름 매칭 과정을 통해 이산 데이터를...","url":"https://www.aioga.com/ko/news/cmsj4jrne24d5ronkvb40rucl/","contentTranslated":true,"sourceHash":"fdeda6806310cc7d","translatedAt":"2026-08-10T11:16:41.772Z"},"es":{"title":"Escala de mapas de flujo categóricos","summary":"Los modelos de difusión continua y coincidencia de flujo tienen el potencial de convertirse en una alternativa poderosa a los métodos autorregresivos en el modelado del lenguaje, ya que pueden desbloquear ventajas de modalidad continua como el muestreo acelerado y la inclinación. Investigaciones recientes lograron la generación continua de datos discretos mediante un sencillo proceso de coincidencia de flujo entre la distribución gaussiana y la distribución de datos codificados en one-hot, y demostraron la viabilidad del muestreo acelerado mediante mapas de flujo de clasificación (CFMs), con una calidad de muestra competitiva. 🔗 Leer artículo original via AIHOT · https://aihot.virxact.com/items/cmsj4jrne24d5ronkvb40rucl","category":"Industria","source":"Apple Machine Learning Research（RSS）","aggregationSource":"Apple Machine Learning Research（RSS）","pageTitle":"Escala de mapas de flujo categóricos - Aioga Noticias de IA","description":"Los modelos de difusión continua y coincidencia de flujo tienen el potencial de convertirse en una alternativa poderosa a los métodos autorregresivos en el modelado del lenguaje, y...","url":"https://www.aioga.com/es/news/cmsj4jrne24d5ronkvb40rucl/","contentTranslated":true,"sourceHash":"fdeda6806310cc7d","translatedAt":"2026-08-10T11:16:40.098Z"},"fr":{"title":"Échelle des cartes d’écoulement catégorielles","summary":"Les modèles de diffusion continue et de correspondance de flux devraient devenir une alternative solide aux méthodes autorégressives en modélisation du langage, débloquant des avantages modaux continus tels que l’échantillonnage accéléré et l’assympathie. Des études récentes ont permis de générer en continu des données discrètes grâce à des processus simples de correspondance de flux entre les distributions gaussiennes et les distributions de données encodées en un hot, et ont confirmé la faisabilité d’un échantillonnage accéléré à l’aide de la cartographie des flux de classification (CFM), ce qui a permis d’obtenir une qualité d’échantillon compétitive. 🔗 Lisez l’article original via AIHOT · https://aihot.virxact.com/items/cmsj4jrne24d5ronkvb40rucl","category":"Industrie","source":"Apple Machine Learning Research（RSS）","aggregationSource":"Apple Machine Learning Research（RSS）","pageTitle":"Échelle des cartes d’écoulement catégorielles - Aioga Actualités IA","description":"Les modèles de diffusion continue et de correspondance de flux devraient devenir une alternative solide aux méthodes autorégressives en modélisation du langage, débloquant des avan...","url":"https://www.aioga.com/fr/news/cmsj4jrne24d5ronkvb40rucl/","contentTranslated":true,"sourceHash":"fdeda6806310cc7d","translatedAt":"2026-08-10T11:16:59.119Z"},"de":{"title":"Maßstab kategorischer Flusskarten","summary":"Kontinuierliche Diffusions- und Flussanpassungsmodelle versprechen, eine leistungsstarke Alternative zu autoregressiven Methoden im Sprachmodellieren zu werden und können Vorteile kontinuierlicher Modalitäten wie beschleunigtes Sampling und Schiefen freisetzen. Jüngste Forschungen haben durch einen einfachen Flussanpassungsprozess zwischen der Gauß-Verteilung und der One-Hot-codierten Datenverteilung die kontinuierliche Erzeugung von diskreten Daten realisiert und die Machbarkeit des beschleunigten Samplings mithilfe von Klassifikationsflussabbildungen (CFMs) überprüft. Die Stichprobenqualität ist wettbewerbsfähig. 🔗 Originalartikel lesen via AIHOT · https://aihot.virxact.com/items/cmsj4jrne24d5ronkvb40rucl","category":"行业动态","source":"Apple Machine Learning Research（RSS）","aggregationSource":"Apple Machine Learning Research（RSS）","pageTitle":"Maßstab kategorischer Flusskarten - Aioga KI-News","description":"Kontinuierliche Diffusions- und Flussanpassungsmodelle versprechen, eine leistungsstarke Alternative zu autoregressiven Methoden im Sprachmodellieren zu werden und können Vorteile...","url":"https://www.aioga.com/de/news/cmsj4jrne24d5ronkvb40rucl/","contentTranslated":true,"sourceHash":"fdeda6806310cc7d","translatedAt":"2026-08-10T11:16:56.226Z"},"pt-BR":{"title":"Escala de Mapas de Escoamento Categórico","summary":"Modelos de difusão contínua e correspondência de fluxo têm potencial para se tornar uma alternativa poderosa aos métodos autoregressivos na modelagem de linguagem, podendo desbloquear vantagens contínuas como amostragem acelerada e inclinação. Pesquisas recentes alcançaram a geração contínua de dados discretos através de um simples processo de correspondência de fluxo entre a distribuição normal e a distribuição de dados codificados em one-hot, e validaram a viabilidade da amostragem acelerada por meio do mapeamento de fluxo de classificação (CFMs), com qualidade de amostra competitiva. 🔗 Leia o texto completo via AIHOT · https://aihot.virxact.com/items/cmsj4jrne24d5ronkvb40rucl","category":"行业动态","source":"Apple Machine Learning Research（RSS）","aggregationSource":"Apple Machine Learning Research（RSS）","pageTitle":"Escala de Mapas de Escoamento Categórico - Aioga Notícias de IA","description":"Modelos de difusão contínua e correspondência de fluxo têm potencial para se tornar uma alternativa poderosa aos métodos autoregressivos na modelagem de linguagem, podendo desbloqu...","url":"https://www.aioga.com/pt-BR/news/cmsj4jrne24d5ronkvb40rucl/","contentTranslated":true,"sourceHash":"fdeda6806310cc7d","translatedAt":"2026-08-10T11:17:12.645Z"},"ru":{"title":"Масштаб категориальных карт потоков","summary":"Модель непрерывного диффузионного и потокового соответствия обещает стать мощной альтернативой авторегрессивным методам в языковом моделировании, открывая возможности ускоренной выборки и преимущества непрерывных модальностей, такие как склонность. Недавние исследования показали, что с помощью простого процесса потокового соответствия между гауссовским распределением и распределением данных с one-hot кодированием возможно непрерывное поколение дискретных данных, а также с помощью картирования классификационного потока (CFMs) была подтверждена возможность ускоренной выборки, при этом качество образцов остается конкурентоспособным. 🔗 Читать оригинал через AIHOT · https://aihot.virxact.com/items/cmsj4jrne24d5ronkvb40rucl","category":"行业动态","source":"Apple Machine Learning Research（RSS）","aggregationSource":"Apple Machine Learning Research（RSS）","pageTitle":"Масштаб категориальных карт потоков - Aioga Новости ИИ","description":"Модель непрерывного диффузионного и потокового соответствия обещает стать мощной альтернативой авторегрессивным методам в языковом моделировании, открывая возможности ускоренной вы...","url":"https://www.aioga.com/ru/news/cmsj4jrne24d5ronkvb40rucl/","contentTranslated":true,"sourceHash":"fdeda6806310cc7d","translatedAt":"2026-08-10T11:17:12.670Z"},"ar":{"title":"مقياس خرائط التدفق التصنيفي","summary":"من المتوقع أن يصبح نموذج الانتشار المتواصل ومطابقة التدفق بديلاً قوياً للطريقة التوليدية الذاتية في نمذجة اللغة، ويمكن أن يفتح مزايا النماذج المستمرة مثل التسريع في العينة والتحيز. أظهرت الدراسات الحديثة أنه من خلال عملية بسيطة لمطابقة التدفق بين توزيع بيانات Gaussian والتشفير الأحادي (one-hot)، يمكن تحقيق توليد مستمر للبيانات المتقطعة، وأكدت خرائط التدفق التصنيفي (CFMs) إمكانية التسريع في أخذ العينات، مع جودة عينات تنافسية. 🔗 اقرأ النص الأصلي عبر AIHOT · https://aihot.virxact.com/items/cmsj4jrne24d5ronkvb40rucl","category":"行业动态","source":"Apple Machine Learning Research（RSS）","aggregationSource":"Apple Machine Learning Research（RSS）","pageTitle":"مقياس خرائط التدفق التصنيفي - Aioga أخبار الذكاء الاصطناعي","description":"من المتوقع أن يصبح نموذج الانتشار المتواصل ومطابقة التدفق بديلاً قوياً للطريقة التوليدية الذاتية في نمذجة اللغة، ويمكن أن يفتح مزايا النماذج المستمرة مثل التسريع في العينة والتحيز....","url":"https://www.aioga.com/ar/news/cmsj4jrne24d5ronkvb40rucl/","contentTranslated":true,"sourceHash":"fdeda6806310cc7d","translatedAt":"2026-08-10T11:17:27.527Z"},"hi":{"title":"श्रेणीबद्ध प्रवाह मानचित्रों का पैमाना","summary":"सतत फैलाव और प्रवाह मिलान मॉडल भाषा मॉडलिंग में ऑटोरेग्रेसीव पद्धतियों के लिए एक शक्तिशाली विकल्प बनने की उम्मीद रखते हैं, जो तेज़ नमूना लेने और झुकाव जैसी सतत मोडलिटीज़ के लाभों को खोल सकते हैं। हाल के शोध में गुसीय वितरण और वन-हॉट एन्कोडेड डेटा वितरण के बीच सरल प्रवाह मिलान प्रक्रिया के माध्यम से डिस्क्रीट डेटा का सतत निर्माण किया गया, और वर्गीकरण फ्लो मैपिंग (CFMs) की मदद से तेज़ नमूना लेने की संभावना की पुष्टि की गई, और नमूने की गुणवत्ता प्रतिस्पर्धी है। 🔗 मूल लेख पढ़ें via AIHOT · https://aihot.virxact.com/items/cmsj4jrne24d5ronkvb40rucl","category":"行业动态","source":"Apple Machine Learning Research（RSS）","aggregationSource":"Apple Machine Learning Research（RSS）","pageTitle":"श्रेणीबद्ध प्रवाह मानचित्रों का पैमाना - Aioga AI समाचार","description":"सतत फैलाव और प्रवाह मिलान मॉडल भाषा मॉडलिंग में ऑटोरेग्रेसीव पद्धतियों के लिए एक शक्तिशाली विकल्प बनने की उम्मीद रखते हैं, जो तेज़ नमूना लेने और झुकाव जैसी सतत मोडलिटीज़ के लाभों क...","url":"https://www.aioga.com/hi/news/cmsj4jrne24d5ronkvb40rucl/","contentTranslated":true,"sourceHash":"fdeda6806310cc7d","translatedAt":"2026-08-10T11:17:28.308Z"},"it":{"title":"Scala delle mappe di flusso categorico","summary":"I modelli di diffusione continua e di corrispondenza del flusso hanno il potenziale per diventare una valida alternativa ai metodi autoregressivi nella modellazione linguistica, sbloccando vantaggi legati alla modalità continua come il campionamento accelerato e la distorsione. Studi recenti hanno realizzato la generazione continua di dati discreti attraverso un semplice processo di corrispondenza del flusso tra la distribuzione dei dati codificati in one-hot e la distribuzione gaussiana, e hanno verificato la fattibilità del campionamento accelerato tramite le mappe di flusso di classificazione (CFMs), mostrando una qualità dei campioni competitiva. 🔗 Leggi l'articolo completo via AIHOT · https://aihot.virxact.com/items/cmsj4jrne24d5ronkvb40rucl","category":"行业动态","source":"Apple Machine Learning Research（RSS）","aggregationSource":"Apple Machine Learning Research（RSS）","pageTitle":"Scala delle mappe di flusso categorico - Aioga Notizie IA","description":"I modelli di diffusione continua e di corrispondenza del flusso hanno il potenziale per diventare una valida alternativa ai metodi autoregressivi nella modellazione linguistica, sb...","url":"https://www.aioga.com/it/news/cmsj4jrne24d5ronkvb40rucl/","contentTranslated":true,"sourceHash":"fdeda6806310cc7d","translatedAt":"2026-08-10T11:17:43.741Z"},"nl":{"title":"Schaal van categorische stroomkaarten","summary":"Het continue diffusie- en stromingsafstemmingsmodel belooft een krachtig alternatief te worden voor autoregressieve methoden in taalmodellering en kan voordelen bieden zoals versnelde sampling en scheve continue modaliteiten. Recente studies hebben door een eenvoudig stromingsafstemmingsproces tussen Gauss-verdelingen en one-hot gecodeerde data de continue generatie van discrete data gerealiseerd, en met behulp van classificatieflowmappings (CFM's) werd de haalbaarheid van versnelde sampling bevestigd, met concurrerende samplekwaliteit. 🔗 Lees het originele artikel via AIHOT · https://aihot.virxact.com/items/cmsj4jrne24d5ronkvb40rucl","category":"行业动态","source":"Apple Machine Learning Research（RSS）","aggregationSource":"Apple Machine Learning Research（RSS）","pageTitle":"Schaal van categorische stroomkaarten - Aioga AI-nieuws","description":"Het continue diffusie- en stromingsafstemmingsmodel belooft een krachtig alternatief te worden voor autoregressieve methoden in taalmodellering en kan voordelen bieden zoals versne...","url":"https://www.aioga.com/nl/news/cmsj4jrne24d5ronkvb40rucl/","contentTranslated":true,"sourceHash":"fdeda6806310cc7d","translatedAt":"2026-08-10T11:17:43.968Z"},"tr":{"title":"Kategorik Akış Haritalarının Ölçeği","summary":"Sürekli yayılma ve akış eşleştirme modelleri, dil modellemede oto-regresif yöntemlerin güçlü bir alternatifi olma potansiyeline sahiptir ve sürekli mod avantajları olan hızlandırılmış örnekleme ve eğiklik gibi özellikleri açığa çıkarabilir. Son dönemdeki araştırmalar, Gauss dağılımı ile one-hot kodlama veri dağılımı arasındaki basit akış eşleştirme süreci aracılığıyla, ayrık verinin sürekli üretimini gerçekleştirmiş ve sınıflandırma akış eşlemeleri (CFM'ler) sayesinde hızlandırılmış örneklemenin uygulanabilirliğini doğrulamıştır; örnek kalitesi rekabetçi seviyededir. 🔗 Orijinalini oku via AIHOT · https://aihot.virxact.com/items/cmsj4jrne24d5ronkvb40rucl","category":"行业动态","source":"Apple Machine Learning Research（RSS）","aggregationSource":"Apple Machine Learning Research（RSS）","pageTitle":"Kategorik Akış Haritalarının Ölçeği - Aioga AI Haberleri","description":"Sürekli yayılma ve akış eşleştirme modelleri, dil modellemede oto-regresif yöntemlerin güçlü bir alternatifi olma potansiyeline sahiptir ve sürekli mod avantajları olan hızlandırıl...","url":"https://www.aioga.com/tr/news/cmsj4jrne24d5ronkvb40rucl/","contentTranslated":true,"sourceHash":"fdeda6806310cc7d","translatedAt":"2026-08-10T11:17:59.258Z"},"vi":{"title":"Tỷ lệ bản đồ dòng chảy phân loại","summary":"Mô hình khuếch tán liên tục và khớp luồng hứa hẹn trở thành một phương pháp thay thế mạnh mẽ cho phương pháp tự hồi quy trong mô hình hóa ngôn ngữ, có thể mở khóa các ưu điểm của chế độ liên tục như tăng tốc lấy mẫu và lệch. Nghiên cứu gần đây đã thực hiện việc tạo dữ liệu rời rạc liên tục thông qua quá trình khớp luồng đơn giản giữa phân phối Gaussian và dữ liệu mã hóa one-hot, đồng thời xác minh khả năng tăng tốc lấy mẫu với sự hỗ trợ của ánh xạ luồng phân loại (CFMs), chất lượng mẫu cạnh tranh. 🔗 Đọc bài gốc via AIHOT · https://aihot.virxact.com/items/cmsj4jrne24d5ronkvb40rucl","category":"行业动态","source":"Apple Machine Learning Research（RSS）","aggregationSource":"Apple Machine Learning Research（RSS）","pageTitle":"Tỷ lệ bản đồ dòng chảy phân loại - Tin tức AI Aioga","description":"Mô hình khuếch tán liên tục và khớp luồng hứa hẹn trở thành một phương pháp thay thế mạnh mẽ cho phương pháp tự hồi quy trong mô hình hóa ngôn ngữ, có thể mở khóa các ưu điểm của c...","url":"https://www.aioga.com/vi/news/cmsj4jrne24d5ronkvb40rucl/","contentTranslated":true,"sourceHash":"fdeda6806310cc7d","translatedAt":"2026-08-10T11:18:00.838Z"},"id":{"title":"Skala peta aliran kategorikal","summary":"Model difusi kontinu dan pencocokan aliran diharapkan menjadi alternatif yang kuat terhadap metode autoregresif dalam pemodelan bahasa, yang dapat membuka keuntungan modalitas kontinu seperti sampling yang dipercepat dan bias. Penelitian terbaru menggunakan proses pencocokan aliran sederhana antara distribusi Gaussian dan distribusi data one-hot untuk menghasilkan data diskrit secara kontinu, dan memvalidasi kelayakan sampling yang dipercepat dengan bantuan pemetaan aliran klasifikasi (CFMs), dengan kualitas sampel yang kompetitif. 🔗 Baca teks asli via AIHOT · https://aihot.virxact.com/items/cmsj4jrne24d5ronkvb40rucl","category":"行业动态","source":"Apple Machine Learning Research（RSS）","aggregationSource":"Apple Machine Learning Research（RSS）","pageTitle":"Skala peta aliran kategorikal - Berita AI Aioga","description":"Model difusi kontinu dan pencocokan aliran diharapkan menjadi alternatif yang kuat terhadap metode autoregresif dalam pemodelan bahasa, yang dapat membuka keuntungan modalitas kont...","url":"https://www.aioga.com/id/news/cmsj4jrne24d5ronkvb40rucl/","contentTranslated":true,"sourceHash":"fdeda6806310cc7d","translatedAt":"2026-08-10T11:18:15.235Z"},"th":{"title":"มาตราส่วนของแผนที่การไหลเชิงหมวดหมู่","summary":"โมเดลการแพร่กระจายต่อเนื่องและการจับคู่ฟลูว์มีแนวโน้มที่จะกลายเป็นทางเลือกที่แข็งแกร่งสำหรับวิธีการออโตรีเกรสซีฟในการสร้างภาษา สามารถเปิดประตูให้กับข้อดีของโหมดต่อเนื่อง เช่น การเร่งความเร็วการสุ่มตัวอย่างและการเอียง งานวิจัยล่าสุดผ่านกระบวนการจับคู่ฟลูว์ที่เรียบง่ายระหว่างการแจกแจงแบบเกาส์และการแจกแจงข้อมูลแบบ one-hot เพื่อสร้างข้อมูลเชิงดิสครีตในรูปแบบต่อเนื่อง และยืนยันความเป็นไปได้ของการเร่งความเร็วการสุ่มตัวอย่างด้วยการทำ Mapping ของฟลูว์เชิงจัดประเภท (CFMs) คุณภาพตัวอย่างสามารถแข่งขันได้ 🔗 อ่านบทความต้นฉบับ via AIHOT · https://aihot.virxact.com/items/cmsj4jrne24d5ronkvb40rucl","category":"行业动态","source":"Apple Machine Learning Research（RSS）","aggregationSource":"Apple Machine Learning Research（RSS）","pageTitle":"มาตราส่วนของแผนที่การไหลเชิงหมวดหมู่ - ข่าว AI Aioga","description":"โมเดลการแพร่กระจายต่อเนื่องและการจับคู่ฟลูว์มีแนวโน้มที่จะกลายเป็นทางเลือกที่แข็งแกร่งสำหรับวิธีการออโตรีเกรสซีฟในการสร้างภาษา สามารถเปิดประตูให้กับข้อดีของโหมดต่อเนื่อง เช่น การเร...","url":"https://www.aioga.com/th/news/cmsj4jrne24d5ronkvb40rucl/","contentTranslated":true,"sourceHash":"fdeda6806310cc7d","translatedAt":"2026-08-10T11:18:15.725Z"},"pl":{"title":"Skala kategorycznych map przepływu","summary":"Model ciągłego dyfuzowania i dopasowania przepływu ma szansę stać się silną alternatywą dla metod autoregresyjnych w modelowaniu języka, umożliwiając wykorzystanie zalet ciągłych modalności, takich jak przyspieszone próbkowanie i nachylenie. Ostatnie badania wykazały, że poprzez prosty proces dopasowania przepływu między rozkładem Gaussa a rozkładem danych zakodowanych w one-hot, możliwe jest ciągłe generowanie danych dyskretnych, a przy pomocy mapowania przepływu klasyfikacyjnego (CFM) potwierdzono wykonalność przyspieszonego próbkowania, przy konkurencyjnej jakości próbek. 🔗 Przeczytaj oryginał via AIHOT · https://aihot.virxact.com/items/cmsj4jrne24d5ronkvb40rucl","category":"行业动态","source":"Apple Machine Learning Research（RSS）","aggregationSource":"Apple Machine Learning Research（RSS）","pageTitle":"Skala kategorycznych map przepływu - Aioga Wiadomości AI","description":"Model ciągłego dyfuzowania i dopasowania przepływu ma szansę stać się silną alternatywą dla metod autoregresyjnych w modelowaniu języka, umożliwiając wykorzystanie zalet ciągłych m...","url":"https://www.aioga.com/pl/news/cmsj4jrne24d5ronkvb40rucl/","contentTranslated":true,"sourceHash":"fdeda6806310cc7d","translatedAt":"2026-08-10T11:18:31.131Z"}},"evidenceTier":"verified-news","reviewStatus":"editorial-selected","indexable":true,"editorialCover":""}}