{"@context":"https://schema.org","@type":"NewsArticle","generatedAt":"2026-07-28T06:20:51.496Z","headline":"Apple 提出 GH-ESD：面向实例级视觉任务的假设驱动错误切片发现方法","description":"Apple 机器学习研究团队提出 GH-ESD（Grounded Hypothesis-Driven Error Slice Discovery），一种针对目标检测与分割等实例级视觉任务的错误切片发现方法。现有方法主要适用于图像级分类，难以捕捉由上下文关系和空间视觉模式导致的实例级失败。GH-ESD 通过基于假设的驱动方式，系统性地发现模型在语义连贯子集上的系统性失效。","url":"https://www.aioga.com/news/cms3el9cc000orozqcae978wd/","mainEntityOfPage":"https://www.aioga.com/news/cms3el9cc000orozqcae978wd/","datePublished":"2026-07-27T00:00:00.000Z","dateModified":"2026-07-27T00:00:00.000Z","inLanguage":"zh-CN","publisher":{"@type":"NewsMediaOrganization","name":"Aioga","url":"https://www.aioga.com"},"citation":["https://machinelearning.apple.com/research/gh-esd","https://aihot.virxact.com/items/cms3el9cc000orozqcae978wd"],"canonicalUrl":"https://www.aioga.com/news/cms3el9cc000orozqcae978wd/","directAnswer":{"@type":"Answer","text":"Aioga 编辑摘要：Apple 机器学习研究团队提出 GH-ESD（Grounded Hypothesis-Driven Error Slice Discovery），一种针对目标检测与分割等实例级视觉任务的错误切片发现方法。 Aioga 将其归入「论文研究」方向，重点关注它对真实使用和行业竞争的影响。","url":"https://www.aioga.com/news/cms3el9cc000orozqcae978wd/","dateCreated":"2026-07-27T00:00:00.000Z","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/gh-esd","datePublished":"2026-07-27T00:00:00.000Z","provider":{"@type":"Organization","name":"machinelearning.apple.com","url":"https://machinelearning.apple.com/research/gh-esd"}},{"@type":"CreativeWork","name":"AIHot archive record","url":"https://aihot.virxact.com/items/cms3el9cc000orozqcae978wd","datePublished":"2026-07-27T00:00:00.000Z","provider":{"@type":"Organization","name":"AIHot","url":"https://aihot.virxact.com/items/cms3el9cc000orozqcae978wd"}}],"aggregationSource":"Apple Machine Learning Research（RSS）","originalPublisher":{"name":"machinelearning.apple.com","url":"https://machinelearning.apple.com/research/gh-esd"},"article":{"id":"cms3el9cc000orozqcae978wd","slug":"cms3el9cc000orozqcae978wd","url":"https://www.aioga.com/news/cms3el9cc000orozqcae978wd/","title":"Apple 提出 GH-ESD：面向实例级视觉任务的假设驱动错误切片发现方法","title_en":"GH-ESD： Grounded Hypothesis-Driven Error Slice Discovery for Instance-Level Vision Tasks","summary":"Apple 机器学习研究团队提出 GH-ESD（Grounded Hypothesis-Driven Error Slice Discovery），一种针对目标检测与分割等实例级视觉任务的错误切片发现方法。现有方法主要适用于图像级分类，难以捕捉由上下文关系和空间视觉模式导致的实例级失败。GH-ESD 通过基于假设的驱动方式，系统性地发现模型在语义连贯子集上的系统性失效。","source":"Apple Machine Learning Research（RSS）","sourceUrl":"https://machinelearning.apple.com/research/gh-esd","aiHotUrl":"https://aihot.virxact.com/items/cms3el9cc000orozqcae978wd","publishedAt":"2026-07-27T00:00:00.000Z","category":"论文研究","score":40,"selected":false,"articleBody":["GH-ESD: Grounded Hypothesis-Driven Error Slice Discovery for Instance-Level Vision Tasks","Authors Wei Zhang*, Chaoqun Wang*, Zixuan Guan, Ping Sheng Kao, Pengfei Zhao, Peng Wu, Sifeng He","View publication：https://arxiv.org/abs/2512.24592","Instance-Level Task Parameters: A Robust Multi-task Weighting Framework","June 24, 2021 research area Computer Vision：/research/?domain=Computer%20Vision Workshop at CVPR：/research/?event=CVPR%20Workshop","Recent works have shown that deep neural networks benefit from multi-task learning by learning a shared representation across several related tasks. However, performance of such systems depend on relative weighting between various losses involved during training. Prior works on loss weighting schemes assume that instances are equally easy or hard for all tasks. In order to break this assumption, we let the training process dictate the optimal…","Sliced Wasserstein Discrepancy for Unsupervised Domain Adaptation","March 10, 2019 research area Methods and Algorithms：/research/?domain=Methods%20and%20Algorithms conference CVPR：/research/?event=CVPR","In this work, we connect two distinct concepts for unsupervised domain adaptation: feature distribution alignment between domains by utilizing the task-specific decision boundary and the Wasserstein metric. Our proposed sliced Wasserstein discrepancy (SWD) is designed to capture the natural notion of dissimilarity between the outputs of task-specific classifiers. It provides a geometrically meaningful guidance to detect target samples that are…","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":8,"url":"/media/articles/cms3el9cc000orozqcae978wd/64c2324784f2cf67.jpg"}],"mediaStatus":"ok","articleBodyZh":["GH-ESD：用于实例级视觉任务的基于假设驱动的错误切片发现","作者：张威*，王超群*，关子轩，曹平升，赵鹏飞，吴鹏，贺思峰","查看出版物：https://arxiv.org/abs/2512.24592","实例级任务参数：一个稳健的多任务加权框架","2021年6月24日 研究领域 计算机视觉：/research/?domain=Computer%20Vision CVPR研讨会：/research/?event=CVPR%20Workshop","近期研究表明，深度神经网络通过在几个相关任务间学习共享表示，从多任务学习中受益。然而，这类系统的性能取决于训练过程中各损失之间的相对加权。此前关于损失加权方案的研究假设所有实例对所有任务来说难度相等。为了打破该假设，我们让训练过程来决定最优…","用于无监督领域自适应的切片瓦瑟斯坦差异","2019年3月10日 研究领域 方法与算法：/research/?domain=Methods%20and%20Algorithms 会议 CVPR：/research/?event=CVPR","在本工作中，我们将两个无监督领域自适应的不同概念联系起来：通过利用任务特定的决策边界和瓦瑟斯坦度量实现领域间的特征分布对齐。我们提出的切片瓦瑟斯坦差异（SWD）旨在捕捉任务特定分类器输出之间的自然差异概念。它为检测目标样本提供了几何上有意义的指导，使其能够…","我们在机器学习领域的研究每天都在开创新局面。"],"translationStatus":"translated","bodyOrigin":"source-page","editorial":{"summary":"Aioga 编辑摘要：Apple 机器学习研究团队提出 GH-ESD（Grounded Hypothesis-Driven Error Slice Discovery），一种针对目标检测与分割等实例级视觉任务的错误切片发现方法。 Aioga 将其归入「论文研究」方向，重点关注它对真实使用和行业竞争的影响。","background":"背景分析：模型与研究类动态需要结合能力边界、开放方式、成本、可用性和真实任务表现判断，单项指标领先不等于已经形成稳定采用。","viewpoint":"Aioga 判断：这条动态更适合作为行业观察信号，当前信息足以建立线索，但不足以推导长期结论。","implications":"影响分析：对相关团队而言，短期应先核对来源、可用范围和实际成本，再判断是否值得接入或跟进。","nextStep":"后续观察：继续观察官方文档、实际可用性、价格变化、开发者反馈和竞品回应。","evidenceRefs":["title","summary","articleBody"],"confidence":"medium","status":"published","aiGenerated":false,"autoApproved":true,"generatedBy":"rule-safe-fallback","generatedAt":"2026-07-28T06:29:10.482Z","sourceHash":"175e1da11d975a6b","validation":{"passed":true,"mode":"rule-safe-fallback","checks":["schema","length","source-attribution","no-html"]}},"tags":["论文研究","Apple Machine Learning Research（RSS）"],"translations":{"zh-CN":{"title":"Apple 提出 GH-ESD：面向实例级视觉任务的假设驱动错误切片发现方法","summary":"Apple 机器学习研究团队提出 GH-ESD（Grounded Hypothesis-Driven Error Slice Discovery），一种针对目标检测与分割等实例级视觉任务的错误切片发现方法。现有方法主要适用于图像级分类，难以捕捉由上下文关系和空间视觉模式导致的实例级失败。GH-ESD 通过基于假设的驱动方式，系统性地发现模型在语义连贯子集上的系统性失效。","category":"论文研究","source":"machinelearning.apple.com","aggregationSource":"Apple Machine Learning Research（RSS）","pageTitle":"Apple 提出 GH-ESD：面向实例级视觉任务的假设驱动错误切片发现方法 - Aioga AI资讯","description":"Apple 机器学习研究团队提出 GH-ESD（Grounded Hypothesis-Driven Error Slice Discovery），一种针对目标检测与分割等实例级视觉任务的错误切片发现方法。现有方法主要适用于图像级分类，难以捕捉由上下文关系和空间视觉模式导致的实例级失败。GH-ESD 通过基于假设的驱动方式，系统性地发现模型在语义连贯子集上的...","url":"https://www.aioga.com/news/cms3el9cc000orozqcae978wd/"},"en":{"title":"Apple proposes GH-ESD: A hypothesis-driven error slicing discovery method for instance-level visual tasks","summary":"The Apple Machine Learning Research team proposed GH-ESD (Grounded Hypothesis-Driven Error Slice Discovery), a method for discovering error slices in instance-level visual tasks such as object detection and segmentation. Existing methods are mainly suitable for image-level classification and struggle to capture instance-level failures caused by contextual relationships and spatial visual patterns. GH-ESD systematically discovers the model's systematic failures on semantically coherent subsets through a hypothesis-driven approach.","category":"Research","source":"Apple Machine Learning Research（RSS）","aggregationSource":"Apple Machine Learning Research（RSS）","pageTitle":"Apple proposes GH-ESD: A hypothesis-driven error slicing discovery method for instance-level visual tasks - Aioga AI News","description":"The Apple Machine Learning Research team proposed GH-ESD (Grounded Hypothesis-Driven Error Slice Discovery), a method for discovering error slices in instance-level visual tasks su...","url":"https://www.aioga.com/en/news/cms3el9cc000orozqcae978wd/","contentTranslated":true,"sourceHash":"1b3f5b3522b06e3b","translatedAt":"2026-07-27T19:42:55.961Z"},"ja":{"title":"Apple は GH-ESD を提出：インスタンスレベルの視覚タスク向け仮説駆動型エラー切片発見手法","summary":"Appleの機械学習研究チームは、GH-ESD（Grounded Hypothesis-Driven Error Slice Discovery）を提案しました。これは、物体検出やセグメンテーションなどのインスタンスレベルの視覚タスクに対するエラー切り片発見手法です。既存の方法は主に画像レベルの分類に適しており、コンテキスト関係や空間視覚パターンに起因するインスタンスレベルの失敗を捉えることが困難です。GH-ESDは仮説に基づく駆動方式により、モデルが意味的に一貫したサブセットで系統的に失敗する箇所を体系的に発見します。","category":"論文研究","source":"Apple Machine Learning Research（RSS）","aggregationSource":"Apple Machine Learning Research（RSS）","pageTitle":"Apple は GH-ESD を提出：インスタンスレベルの視覚タスク向け仮説駆動型エラー切片発見手法 - Aioga AIニュース","description":"Appleの機械学習研究チームは、GH-ESD（Grounded Hypothesis-Driven Error Slice Discovery）を提案しました。これは、物体検出やセグメンテーションなどのインスタンスレベルの視覚タスクに対するエラー切り片発見手法です。既存の方法は主に画像レベルの分類に適しており、コンテキスト関係や空間視覚パターンに起因するイ...","url":"https://www.aioga.com/ja/news/cms3el9cc000orozqcae978wd/","contentTranslated":true,"sourceHash":"1b3f5b3522b06e3b","translatedAt":"2026-07-27T19:43:06.281Z"},"ko":{"title":"Apple는 GH-ESD를 제안했습니다: 인스턴스 수준 시각 작업을 위한 가설 기반 오류 조각 발견 방법","summary":"Apple 머신러닝 연구팀은 GH-ESD(Grounded Hypothesis-Driven Error Slice Discovery)를 제안했습니다. 이는 객체 검출 및 세분화와 같은 인스턴스 수준의 시각 작업을 위한 오류 슬라이스 발견 방법입니다. 기존의 방법은 주로 이미지 수준의 분류에 적합하며, 컨텍스트 관계와 공간 시각 패턴으로 인해 발생하는 인스턴스 수준의 실패를 포착하기 어렵습니다. GH-ESD는 가설 기반 구동 방식을 통해 모델이 의미상 일관된 하위 집합에서 겪는 체계적인 실패를 체계적으로 발견합니다.","category":"연구","source":"Apple Machine Learning Research（RSS）","aggregationSource":"Apple Machine Learning Research（RSS）","pageTitle":"Apple는 GH-ESD를 제안했습니다: 인스턴스 수준 시각 작업을 위한 가설 기반 오류 조각 발견 방법 - Aioga AI 뉴스","description":"Apple 머신러닝 연구팀은 GH-ESD(Grounded Hypothesis-Driven Error Slice Discovery)를 제안했습니다. 이는 객체 검출 및 세분화와 같은 인스턴스 수준의 시각 작업을 위한 오류 슬라이스 발견 방법입니다. 기존의 방법은 주로 이미지 수준의 분류에 적합하며, 컨텍스트 관계와 공간 시...","url":"https://www.aioga.com/ko/news/cms3el9cc000orozqcae978wd/","contentTranslated":true,"sourceHash":"1b3f5b3522b06e3b","translatedAt":"2026-07-27T19:44:00.448Z"},"es":{"title":"Apple propone GH-ESD: un método de detección de errores basado en hipótesis orientado a tareas visuales a nivel de instancia","summary":"El equipo de investigación en aprendizaje automático de Apple propuso GH-ESD (Grounded Hypothesis-Driven Error Slice Discovery), un método para descubrir fragmentos de error en tareas visuales a nivel de instancia, como la detección y segmentación de objetivos. Los métodos existentes se aplican principalmente a la clasificación a nivel de imagen y difícilmente capturan fallos a nivel de instancia causados por relaciones contextuales y patrones visuales espaciales. GH-ESD, mediante un enfoque impulsado por hipótesis, descubre sistemáticamente los fallos del modelo en subconjuntos semánticamente coherentes.","category":"Investigación","source":"Apple Machine Learning Research（RSS）","aggregationSource":"Apple Machine Learning Research（RSS）","pageTitle":"Apple propone GH-ESD: un método de detección de errores basado en hipótesis orientado a tareas visuales a nivel de instancia - Aioga Noticias de IA","description":"El equipo de investigación en aprendizaje automático de Apple propuso GH-ESD (Grounded Hypothesis-Driven Error Slice Discovery), un método para descubrir fragmentos de error en tar...","url":"https://www.aioga.com/es/news/cms3el9cc000orozqcae978wd/","contentTranslated":true,"sourceHash":"1b3f5b3522b06e3b","translatedAt":"2026-07-27T19:43:53.374Z"},"fr":{"title":"Apple propose GH-ESD : une méthode de découverte d'erreurs basée sur des hypothèses pour les tâches visuelles au niveau des instances","summary":"L'équipe de recherche en apprentissage automatique d'Apple a proposé GH-ESD (Grounded Hypothesis-Driven Error Slice Discovery), une méthode de découverte de tranches d'erreurs pour des tâches visuelles au niveau des instances telles que la détection et la segmentation d'objets. Les méthodes existantes sont principalement adaptées à la classification au niveau de l'image, mais peinent à capturer les échecs au niveau des instances provoqués par les relations contextuelles et les motifs visuels spatiaux. GH-ESD découvre systématiquement, de manière guidée par des hypothèses, les défaillances systématiques du modèle sur des sous-ensembles sémantiquement cohérents.","category":"Recherche","source":"Apple Machine Learning Research（RSS）","aggregationSource":"Apple Machine Learning Research（RSS）","pageTitle":"Apple propose GH-ESD : une méthode de découverte d'erreurs basée sur des hypothèses pour les tâches visuelles au niveau des instances - Aioga Actualités IA","description":"L'équipe de recherche en apprentissage automatique d'Apple a proposé GH-ESD (Grounded Hypothesis-Driven Error Slice Discovery), une méthode de découverte de tranches d'erreurs pour...","url":"https://www.aioga.com/fr/news/cms3el9cc000orozqcae978wd/","contentTranslated":true,"sourceHash":"1b3f5b3522b06e3b","translatedAt":"2026-07-27T19:44:40.148Z"},"de":{"title":"Apple stellt GH-ESD vor: Eine hypothesengesteuerte Fehlerscheibenentdeckungsmethode für instanzbezogene visuelle Aufgaben","summary":"Das Machine-Learning-Forschungsteam von Apple schlug GH-ESD (Grounded Hypothesis-Driven Error Slice Discovery) vor, eine Methode zur Entdeckung von Fehlerschnitten für instanzbasierte visuelle Aufgaben wie Objekterkennung und Segmentierung. Bestehende Methoden eignen sich hauptsächlich für bildbasierte Klassifikationen und erfassen instanzbasierte Fehler, die durch Kontextbeziehungen und räumliche visuelle Muster verursacht werden, nur schwer. GH-ESD entdeckt systematisch modellbezogene systematische Fehler innerhalb semantisch zusammenhängender Teilmengen durch einen hypothesengesteuerten Ansatz.","category":"论文研究","source":"Apple Machine Learning Research（RSS）","aggregationSource":"Apple Machine Learning Research（RSS）","pageTitle":"Apple stellt GH-ESD vor: Eine hypothesengesteuerte Fehlerscheibenentdeckungsmethode für instanzbezogene visuelle Aufgaben - Aioga KI-News","description":"Das Machine-Learning-Forschungsteam von Apple schlug GH-ESD (Grounded Hypothesis-Driven Error Slice Discovery) vor, eine Methode zur Entdeckung von Fehlerschnitten für instanzbasie...","url":"https://www.aioga.com/de/news/cms3el9cc000orozqcae978wd/","contentTranslated":true,"sourceHash":"1b3f5b3522b06e3b","translatedAt":"2026-07-27T19:44:40.662Z"},"pt-BR":{"title":"Apple propõe GH-ESD: um método de descoberta de erros orientado por hipóteses para tarefas visuais em nível de instância","summary":"A equipe de pesquisa em aprendizado de máquina da Apple propôs o GH-ESD (Grounded Hypothesis-Driven Error Slice Discovery), um método de descoberta de fatias de erro para tarefas visuais de nível de instância, como detecção e segmentação de objetos. Os métodos existentes são principalmente aplicáveis à classificação em nível de imagem, sendo difíceis de capturar falhas em nível de instância causadas por relações contextuais e padrões visuais espaciais. O GH-ESD, por meio de uma abordagem orientada por hipóteses, descobre sistematicamente falhas sistemáticas do modelo em subconjuntos semanticamente coerentes.","category":"论文研究","source":"Apple Machine Learning Research（RSS）","aggregationSource":"Apple Machine Learning Research（RSS）","pageTitle":"Apple propõe GH-ESD: um método de descoberta de erros orientado por hipóteses para tarefas visuais em nível de instância - Aioga Notícias de IA","description":"A equipe de pesquisa em aprendizado de máquina da Apple propôs o GH-ESD (Grounded Hypothesis-Driven Error Slice Discovery), um método de descoberta de fatias de erro para tarefas v...","url":"https://www.aioga.com/pt-BR/news/cms3el9cc000orozqcae978wd/","contentTranslated":true,"sourceHash":"1b3f5b3522b06e3b","translatedAt":"2026-07-27T19:45:19.413Z"},"ru":{"title":"Apple представила GH-ESD: метод обнаружения ошибок на основе гипотез для визуальных задач на уровне экземпляров","summary":"Исследовательская команда Apple по машинному обучению предложила GH-ESD (Grounded Hypothesis-Driven Error Slice Discovery), метод обнаружения ошибок для визуальных задач на уровне экземпляров, таких как детекция и сегментация целей. Существующие методы в основном подходят для классификации на уровне изображений и с трудом выявляют ошибки на уровне экземпляров, вызванные контекстными связями и пространственными визуальными шаблонами. GH-ESD с помощью гипотетически-ориентированного подхода систематически выявляет систематические сбои модели на семантически согласованных подмножествах.","category":"论文研究","source":"Apple Machine Learning Research（RSS）","aggregationSource":"Apple Machine Learning Research（RSS）","pageTitle":"Apple представила GH-ESD: метод обнаружения ошибок на основе гипотез для визуальных задач на уровне экземпляров - Aioga Новости ИИ","description":"Исследовательская команда Apple по машинному обучению предложила GH-ESD (Grounded Hypothesis-Driven Error Slice Discovery), метод обнаружения ошибок для визуальных задач на уровне...","url":"https://www.aioga.com/ru/news/cms3el9cc000orozqcae978wd/","contentTranslated":true,"sourceHash":"1b3f5b3522b06e3b","translatedAt":"2026-07-27T19:45:26.762Z"},"ar":{"title":"أبل تقترح GH-ESD: طريقة اكتشاف الشرائح الخاطئة المستندة إلى الافتراضات لمهام الرؤية على مستوى المثال","summary":"قدم فريق أبحاث التعلم الآلي في Apple طريقة GH-ESD (اكتشاف شرائح الأخطاء المدفوعة بالفرضيات المعتمدة على الواقع)، وهي طريقة لاكتشاف شرائح الأخطاء لمهام الرؤية على مستوى الكائن مثل الكشف والتقسيم. الطرق الحالية مناسبة بشكل رئيسي لتصنيف الصور على المستوى الكلي، ويصعب عليها التقاط الفشل على مستوى الكائن الناتج عن العلاقات السياقية والأنماط البصرية المكانية. تقوم GH-ESD من خلال طريقة مدفوعة بالفرضيات باكتشاف الفشل النظامي للنموذج على مجموعات فرعية متماسكة دلاليًا بشكل منهجي.","category":"论文研究","source":"Apple Machine Learning Research（RSS）","aggregationSource":"Apple Machine Learning Research（RSS）","pageTitle":"أبل تقترح GH-ESD: طريقة اكتشاف الشرائح الخاطئة المستندة إلى الافتراضات لمهام الرؤية على مستوى المثال - Aioga أخبار الذكاء الاصطناعي","description":"قدم فريق أبحاث التعلم الآلي في Apple طريقة GH-ESD (اكتشاف شرائح الأخطاء المدفوعة بالفرضيات المعتمدة على الواقع)، وهي طريقة لاكتشاف شرائح الأخطاء لمهام الرؤية على مستوى الكائن مثل ا...","url":"https://www.aioga.com/ar/news/cms3el9cc000orozqcae978wd/","contentTranslated":true,"sourceHash":"1b3f5b3522b06e3b","translatedAt":"2026-07-27T19:46:06.743Z"},"hi":{"title":"Apple ने GH-ESD प्रस्तुत किया: उदाहरण-स्तरीय दृश्य कार्यों के लिए अनुमान-चालित त्रुटि स्लाइस पहचान विधि","summary":"Apple मशीन लर्निंग रिसर्च टीम ने GH-ESD (Grounded Hypothesis-Driven Error Slice Discovery) पेश किया, जो लक्ष्य पहचान और विभाजन जैसी उदाहरण-स्तरीय दृश्य कार्यों के लिए त्रुटि स्लाइस खोजने की एक विधि है। मौजूदा तरीकों का मुख्य रूप से छवि-स्तरीय वर्गीकरण पर ही उपयोग होता है और ये संदर्भ संबंध और स्थानिक दृश्य पैटर्नों से उत्पन्न उदाहरण-स्तरीय विफलताओं को पकड़ने में कठिन होते हैं। GH-ESD परिकल्पना-आधारित संचालित तरीके के माध्यम से मॉडल की सिमेंटिक रूप से सुसंगत उपसेट पर व्यवस्थित विफलताओं का व्यवस्थित रूप से पता लगाता है।","category":"论文研究","source":"Apple Machine Learning Research（RSS）","aggregationSource":"Apple Machine Learning Research（RSS）","pageTitle":"Apple ने GH-ESD प्रस्तुत किया: उदाहरण-स्तरीय दृश्य कार्यों के लिए अनुमान-चालित त्रुटि स्लाइस पहचान विधि - Aioga AI समाचार","description":"Apple मशीन लर्निंग रिसर्च टीम ने GH-ESD (Grounded Hypothesis-Driven Error Slice Discovery) पेश किया, जो लक्ष्य पहचान और विभाजन जैसी उदाहरण-स्तरीय दृश्य कार्यों के लिए त्रुटि स्लाइस...","url":"https://www.aioga.com/hi/news/cms3el9cc000orozqcae978wd/","contentTranslated":true,"sourceHash":"1b3f5b3522b06e3b","translatedAt":"2026-07-27T19:46:12.382Z"},"it":{"title":"Apple propone GH-ESD: un metodo di scoperta degli errori guidato da ipotesi per compiti visivi a livello di istanza","summary":"Il team di ricerca sul machine learning di Apple ha proposto GH-ESD (Grounded Hypothesis-Driven Error Slice Discovery), un metodo per la scoperta di slice di errore mirato a compiti visivi a livello di istanza come il rilevamento e la segmentazione degli oggetti. I metodi esistenti sono principalmente adatti alla classificazione a livello di immagine e faticano a catturare i fallimenti a livello di istanza causati dalle relazioni contestuali e dai modelli visivi spaziali. GH-ESD, attraverso un approccio guidato basato su ipotesi, individua sistematicamente i fallimenti sistematici del modello su sottoinsiemi semanticamente coerenti.","category":"论文研究","source":"Apple Machine Learning Research（RSS）","aggregationSource":"Apple Machine Learning Research（RSS）","pageTitle":"Apple propone GH-ESD: un metodo di scoperta degli errori guidato da ipotesi per compiti visivi a livello di istanza - Aioga Notizie IA","description":"Il team di ricerca sul machine learning di Apple ha proposto GH-ESD (Grounded Hypothesis-Driven Error Slice Discovery), un metodo per la scoperta di slice di errore mirato a compit...","url":"https://www.aioga.com/it/news/cms3el9cc000orozqcae978wd/","contentTranslated":true,"sourceHash":"1b3f5b3522b06e3b","translatedAt":"2026-07-27T19:46:54.626Z"},"nl":{"title":"Apple stelt GH-ESD voor: een hypothesegestuurde fout-slice ontdekkingmethode voor instance-level visuele taken","summary":"Het machine learning-onderzoeksteam van Apple heeft GH-ESD (Grounded Hypothesis-Driven Error Slice Discovery) voorgesteld, een methode voor het ontdekken van foutensneden voor instance-level visuele taken zoals objectdetectie en segmentatie. Bestaande methoden zijn voornamelijk geschikt voor image-level classificatie en hebben moeite met het vastleggen van instance-level fouten die worden veroorzaakt door contextuele relaties en ruimtelijke visuele patronen. GH-ESD ontdekt systematisch systematische fouten van het model op semantisch samenhangende subsets via een hypothese-gedreven benadering.","category":"论文研究","source":"Apple Machine Learning Research（RSS）","aggregationSource":"Apple Machine Learning Research（RSS）","pageTitle":"Apple stelt GH-ESD voor: een hypothesegestuurde fout-slice ontdekkingmethode voor instance-level visuele taken - Aioga AI-nieuws","description":"Het machine learning-onderzoeksteam van Apple heeft GH-ESD (Grounded Hypothesis-Driven Error Slice Discovery) voorgesteld, een methode voor het ontdekken van foutensneden voor inst...","url":"https://www.aioga.com/nl/news/cms3el9cc000orozqcae978wd/","contentTranslated":true,"sourceHash":"1b3f5b3522b06e3b","translatedAt":"2026-07-27T19:46:58.478Z"},"tr":{"title":"Apple GH-ESD'yi önerdi: Örnek düzeyindeki görsel görevler için varsayıma dayalı hata dilimi keşif yöntemi","summary":"Apple makine öğrenimi araştırma ekibi, GH-ESD (Grounded Hypothesis-Driven Error Slice Discovery) adlı, hedef tespit ve segmentasyon gibi örnek bazlı görsel görevler için bir hata dilimi keşif yöntemi önerdi. Mevcut yöntemler esas olarak görüntü düzeyinde sınıflandırmaya uygundur ve bağlam ilişkileri ile mekansal görsel desenlerden kaynaklanan örnek bazlı hataları yakalamakta zorlanır. GH-ESD, varsayıma dayalı bir yaklaşımla, modelin anlamsal olarak tutarlı alt kümelerdeki sistematik başarısızlıklarını sistematik bir şekilde keşfeder.","category":"论文研究","source":"Apple Machine Learning Research（RSS）","aggregationSource":"Apple Machine Learning Research（RSS）","pageTitle":"Apple GH-ESD'yi önerdi: Örnek düzeyindeki görsel görevler için varsayıma dayalı hata dilimi keşif yöntemi - Aioga AI Haberleri","description":"Apple makine öğrenimi araştırma ekibi, GH-ESD (Grounded Hypothesis-Driven Error Slice Discovery) adlı, hedef tespit ve segmentasyon gibi örnek bazlı görsel görevler için bir hata d...","url":"https://www.aioga.com/tr/news/cms3el9cc000orozqcae978wd/","contentTranslated":true,"sourceHash":"1b3f5b3522b06e3b","translatedAt":"2026-07-27T19:47:43.828Z"},"vi":{"title":"Apple đề xuất GH-ESD: Phương pháp phát hiện lỗi dựa trên giả thuyết cho các nhiệm vụ thị giác cấp thực thể","summary":"Nhóm nghiên cứu học máy của Apple đã đề xuất GH-ESD (Grounded Hypothesis-Driven Error Slice Discovery), một phương pháp phát hiện lát cắt lỗi dành cho các nhiệm vụ thị giác cấp độ thực thể như phát hiện và phân đoạn mục tiêu. Các phương pháp hiện có chủ yếu áp dụng cho phân loại cấp độ hình ảnh, khó có thể nắm bắt các thất bại cấp độ thực thể gây ra bởi mối quan hệ ngữ cảnh và mô hình thị giác không gian. GH-ESD thông qua cách tiếp cận dựa trên giả thuyết, có hệ thống phát hiện các thất bại hệ thống của mô hình trên các tập con ngữ nghĩa liền mạch.","category":"论文研究","source":"Apple Machine Learning Research（RSS）","aggregationSource":"Apple Machine Learning Research（RSS）","pageTitle":"Apple đề xuất GH-ESD: Phương pháp phát hiện lỗi dựa trên giả thuyết cho các nhiệm vụ thị giác cấp thực thể - Tin tức AI Aioga","description":"Nhóm nghiên cứu học máy của Apple đã đề xuất GH-ESD (Grounded Hypothesis-Driven Error Slice Discovery), một phương pháp phát hiện lát cắt lỗi dành cho các nhiệm vụ thị giác cấp độ...","url":"https://www.aioga.com/vi/news/cms3el9cc000orozqcae978wd/","contentTranslated":true,"sourceHash":"1b3f5b3522b06e3b","translatedAt":"2026-07-27T19:47:44.838Z"},"id":{"title":"Apple Mengajukan GH-ESD: Metode Penemuan Irisan Kesalahan Berbasis Hipotesis untuk Tugas Visual Tingkat Instance","summary":"Tim penelitian pembelajaran mesin Apple mengusulkan GH-ESD (Grounded Hypothesis-Driven Error Slice Discovery), sebuah metode penemuan potongan kesalahan untuk tugas visual tingkat instance seperti deteksi dan segmentasi objek. Metode yang ada saat ini sebagian besar cocok untuk klasifikasi tingkat gambar, sehingga sulit menangkap kegagalan tingkat instance yang disebabkan oleh hubungan konteks dan pola visual spasial. GH-ESD, melalui pendekatan berbasis hipotesis, secara sistematis menemukan kegagalan sistematis model pada subset yang memiliki koherensi semantik.","category":"论文研究","source":"Apple Machine Learning Research（RSS）","aggregationSource":"Apple Machine Learning Research（RSS）","pageTitle":"Apple Mengajukan GH-ESD: Metode Penemuan Irisan Kesalahan Berbasis Hipotesis untuk Tugas Visual Tingkat Instance - Berita AI Aioga","description":"Tim penelitian pembelajaran mesin Apple mengusulkan GH-ESD (Grounded Hypothesis-Driven Error Slice Discovery), sebuah metode penemuan potongan kesalahan untuk tugas visual tingkat...","url":"https://www.aioga.com/id/news/cms3el9cc000orozqcae978wd/","contentTranslated":true,"sourceHash":"1b3f5b3522b06e3b","translatedAt":"2026-07-27T19:48:25.359Z"},"th":{"title":"Apple เสนอ GH-ESD: วิธีการค้นหาข้อผิดพลาดโดยสมมติฐานสำหรับงานวิสัยทัศน์ระดับตัวอย่าง","summary":"ทีมวิจัยการเรียนรู้ของเครื่องของ Apple ได้นำเสนอ GH-ESD (Grounded Hypothesis-Driven Error Slice Discovery) ซึ่งเป็นวิธีการค้นหาชิ้นส่วนข้อผิดพลาดสำหรับงานด้านภาพระดับตัวอย่าง เช่น การตรวจจับวัตถุและการแบ่งส่วน วิธีการที่มีอยู่ส่วนใหญ่ใช้ได้กับการจำแนกประเภทระดับภาพ ทำให้ยากต่อการจับความล้มเหลวระดับตัวอย่างที่เกิดจากความสัมพันธ์บริบทและรูปแบบภาพเชิงพื้นที่ GH-ESD ใช้วิธีการขับเคลื่อนโดยสมมติฐานเพื่อค้นพบความล้มเหลวอย่างเป็นระบบของโมเดลบนกลุ่มย่อยที่มีความหมายเชิงบริบท","category":"论文研究","source":"Apple Machine Learning Research（RSS）","aggregationSource":"Apple Machine Learning Research（RSS）","pageTitle":"Apple เสนอ GH-ESD: วิธีการค้นหาข้อผิดพลาดโดยสมมติฐานสำหรับงานวิสัยทัศน์ระดับตัวอย่าง - ข่าว AI Aioga","description":"ทีมวิจัยการเรียนรู้ของเครื่องของ Apple ได้นำเสนอ GH-ESD (Grounded Hypothesis-Driven Error Slice Discovery) ซึ่งเป็นวิธีการค้นหาชิ้นส่วนข้อผิดพลาดสำหรับงานด้านภาพระดับตัวอย่าง เช่น...","url":"https://www.aioga.com/th/news/cms3el9cc000orozqcae978wd/","contentTranslated":true,"sourceHash":"1b3f5b3522b06e3b","translatedAt":"2026-07-27T19:48:28.237Z"},"pl":{"title":"Apple przedstawiło GH-ESD: Metodę wykrywania błędnych fragmentów napędzaną hipotezami dla zadań wizualnych na poziomie instancji","summary":"Zespół badawczy Apple zajmujący się uczeniem maszynowym zaproponował GH-ESD (Grounded Hypothesis-Driven Error Slice Discovery), metodę odkrywania błędnych wycinków dla zadań wizualnych na poziomie obiektów, takich jak wykrywanie i segmentacja obiektów. Dotychczasowe metody głównie nadają się do klasyfikacji na poziomie obrazów i trudno im uchwycić niepowodzenia na poziomie obiektów spowodowane zależnościami kontekstowymi i przestrzennymi wzorcami wizualnymi. GH-ESD w oparciu o podejście napędzane hipotezami systematycznie odkrywa systematyczne błędy modelu w semantycznie spójnych podzbiorach.","category":"论文研究","source":"Apple Machine Learning Research（RSS）","aggregationSource":"Apple Machine Learning Research（RSS）","pageTitle":"Apple przedstawiło GH-ESD: Metodę wykrywania błędnych fragmentów napędzaną hipotezami dla zadań wizualnych na poziomie instancji - Aioga Wiadomości AI","description":"Zespół badawczy Apple zajmujący się uczeniem maszynowym zaproponował GH-ESD (Grounded Hypothesis-Driven Error Slice Discovery), metodę odkrywania błędnych wycinków dla zadań wizual...","url":"https://www.aioga.com/pl/news/cms3el9cc000orozqcae978wd/","contentTranslated":true,"sourceHash":"1b3f5b3522b06e3b","translatedAt":"2026-07-27T19:49:07.924Z"}}}}