{"@context":"https://schema.org","@type":"NewsArticle","generatedAt":"2026-07-23T06:40:50.084Z","headline":"Apple 提出无环境合成数据生成方法，用于训练 API 调用型 LLM 智能体","description":"Apple 研究人员提出一种无需可执行环境即可生成高质量训练数据的方法，用于训练 API 调用型大语言模型（LLM）智能体。该方法仅需 API 规格说明，利用 LLM 作为数字世界模型，通过教师智能体与 LLM 模拟器交互生成轨迹，并由 LLM 裁判过滤。在 AppWorld 和 OfficeBench 基准上，微调模型使用该合成数据取得了显著的性能提升。","url":"https://www.aioga.com/news/cmrurrjvx04zdbi9t9c30lp8p/","mainEntityOfPage":"https://www.aioga.com/news/cmrurrjvx04zdbi9t9c30lp8p/","datePublished":"2026-07-21T00:00:00.000Z","dateModified":"2026-07-21T00:00:00.000Z","inLanguage":"zh-CN","publisher":{"@type":"NewsMediaOrganization","name":"Aioga","url":"https://www.aioga.com"},"citation":["https://machinelearning.apple.com/research/environment-free","https://aihot.virxact.com/items/cmrurrjvx04zdbi9t9c30lp8p"],"canonicalUrl":"https://www.aioga.com/news/cmrurrjvx04zdbi9t9c30lp8p/","directAnswer":{"@type":"Answer","text":"Aioga 编辑摘要：Apple 研究人员提出一种无需可执行环境即可生成高质量训练数据的方法，用于训练 API 调用型大语言模型（LLM）智能体。 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/cmrurrjvx04zdbi9t9c30lp8p/","dateCreated":"2026-07-21T00:00:00.000Z","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/environment-free","datePublished":"2026-07-21T00:00:00.000Z","provider":{"@type":"Organization","name":"machinelearning.apple.com","url":"https://machinelearning.apple.com/research/environment-free"}},{"@type":"CreativeWork","name":"AIHot archive record","url":"https://aihot.virxact.com/items/cmrurrjvx04zdbi9t9c30lp8p","datePublished":"2026-07-21T00:00:00.000Z","provider":{"@type":"Organization","name":"AIHot","url":"https://aihot.virxact.com/items/cmrurrjvx04zdbi9t9c30lp8p"}}],"aggregationSource":"Apple Machine Learning Research（RSS）","originalPublisher":{"name":"machinelearning.apple.com","url":"https://machinelearning.apple.com/research/environment-free"},"article":{"id":"cmrurrjvx04zdbi9t9c30lp8p","slug":"cmrurrjvx04zdbi9t9c30lp8p","url":"https://www.aioga.com/news/cmrurrjvx04zdbi9t9c30lp8p/","title":"Apple 提出无环境合成数据生成方法，用于训练 API 调用型 LLM 智能体","title_en":"Environment-free Synthetic Data Generation for API-Calling Agents","summary":"Apple 研究人员提出一种无需可执行环境即可生成高质量训练数据的方法，用于训练 API 调用型大语言模型（LLM）智能体。该方法仅需 API 规格说明，利用 LLM 作为数字世界模型，通过教师智能体与 LLM 模拟器交互生成轨迹，并由 LLM 裁判过滤。在 AppWorld 和 OfficeBench 基准上，微调模型使用该合成数据取得了显著的性能提升。","source":"Apple Machine Learning Research（RSS）","sourceUrl":"https://machinelearning.apple.com/research/environment-free","aiHotUrl":"https://aihot.virxact.com/items/cmrurrjvx04zdbi9t9c30lp8p","publishedAt":"2026-07-21T00:00:00.000Z","category":"论文研究","score":52,"selected":false,"articleBody":["Environment-free Synthetic Data Generation for API-Calling Agents","Authors Seanie Lee, Sanjoy Chowdhury, Chao Jiang, Cheng-Yu Hsieh, Ting-Yao Hu, Alexander T Toshev, Oncel Tuzel, Raviteja Vemulapalli","View publication：https://arxiv.org/abs/2607.16900","Reinforcement Learning for Long-Horizon Interactive LLM Agents","February 5, 2025 research area Methods and Algorithms：/research/?domain=Methods%20and%20Algorithms","Interactive digital agents (IDAs) leverage APIs of stateful digital environments to perform tasks in response to user requests. While IDAs powered by instruction-tuned large language models (LLMs) can react to feedback from interface invocations in multi-step exchanges, they have not been trained in their respective digital environments. Prior methods accomplish less than half of tasks in sophisticated benchmarks such as AppWorld. We present a…","Hierarchical and Dynamic Prompt Compression for Efficient Zero-shot API Usage","April 15, 2024 research area Speech and Natural Language Processing：/research/?domain=Speech%20and%20Natural%20Language%20Processing conference EACL：/research/?event=EACL","Long prompts present a significant challenge for practical LLM-based systems that need to operate with low latency and limited resources. We investigate prompt compression for zero-shot dialogue systems that learn to use unseen APIs directly in-context from their documentation, which may take up hundreds of prompt tokens per API. We start from a recently introduced approach (Mu et al., 2023) that learns to compress the prompt into a few “gist…","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/cmrurrjvx04zdbi9t9c30lp8p/64c2324784f2cf67.jpg"}],"mediaStatus":"ok","articleBodyZh":["面向 API 调用代理的无环境合成数据生成","作者 Seanie Lee、Sanjoy Chowdhury、Chao Jiang、Cheng-Yu Hsieh、Ting-Yao Hu、Alexander T Toshev、Oncel Tuzel、Raviteja Vemulapalli","查看出版物：https://arxiv.org/abs/2607.16900","面向长远交互的强化学习大型语言模型代理","2025 年 2 月 5 日 研究领域 方法与算法：/research/?domain=Methods%20and%20Algorithms","交互式数字代理（IDAs）利用有状态数字环境的 API 来响应用户请求执行任务。虽然由指令微调的大型语言模型（LLMs）支持的 IDAs 可以在多步交互中对接口调用的反馈做出反应，但它们尚未在各自的数字环境中进行训练。以往方法在诸如 AppWorld 等复杂基准测试中完成的任务不足一半。我们提出了一个…","高效零样本 API 使用的分层动态提示压缩","2024 年 4 月 15 日 研究领域 语音与自然语言处理：/research/?domain=Speech%20and%20Natural%20Language%20Processing 会议 EACL：/research/?event=EACL","长提示对需要低延迟和有限资源操作的实际基于 LLM 的系统构成了重大挑战。我们研究零样本对话系统的提示压缩，该系统学习从文档中直接以内联方式使用未见过的 API，每个 API 可能占据数百个提示 token。我们从最近提出的一种方法（Mu 等，2023）开始，该方法学习将提示压缩为少量“要点…","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":"Aioga 编辑摘要：Apple 研究人员提出一种无需可执行环境即可生成高质量训练数据的方法，用于训练 API 调用型大语言模型（LLM）智能体。 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":"2026-07-23T06:49:19.093Z","sourceHash":"0ffe8fe5e47b4284","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 提出无环境合成数据生成方法，用于训练 API 调用型 LLM 智能体","summary":"Apple 研究人员提出一种无需可执行环境即可生成高质量训练数据的方法，用于训练 API 调用型大语言模型（LLM）智能体。该方法仅需 API 规格说明，利用 LLM 作为数字世界模型，通过教师智能体与 LLM 模拟器交互生成轨迹，并由 LLM 裁判过滤。在 AppWorld 和 OfficeBench 基准上，微调模型使用该合成数据取得了显著的性能提升。","category":"论文研究","source":"machinelearning.apple.com","aggregationSource":"Apple Machine Learning Research（RSS）","pageTitle":"Apple 提出无环境合成数据生成方法，用于训练 API 调用型 LLM 智能体 - Aioga AI资讯","description":"Apple 研究人员提出一种无需可执行环境即可生成高质量训练数据的方法，用于训练 API 调用型大语言模型（LLM）智能体。该方法仅需 API 规格说明，利用 LLM 作为数字世界模型，通过教师智能体与 LLM 模拟器交互生成轨迹，并由 LLM 裁判过滤。在 AppWorld 和 OfficeBench 基准上，微调模型使用该合成数据取得了显著的性能提升。","url":"https://www.aioga.com/news/cmrurrjvx04zdbi9t9c30lp8p/"},"en":{"title":"Apple proposes a method for generating environmental-free synthetic data, used for training API-calling LLM agents","summary":"Apple researchers proposed a method to generate high-quality training data without the need for an executable environment, aimed at training API-calling large language model (LLM) agents. This method only requires API specifications, uses LLMs as a model of the digital world, generates trajectories through interactions between a teacher agent and the LLM simulator, and filters them with an LLM referee. On the AppWorld and OfficeBench benchmarks, models fine-tuned using this synthetic data achieved significant performance improvements.","category":"Research","source":"Apple Machine Learning Research（RSS）","aggregationSource":"Apple Machine Learning Research（RSS）","pageTitle":"Apple proposes a method for generating environmental-free synthetic data, used for training API-calling LLM agents - Aioga AI News","description":"Apple researchers proposed a method to generate high-quality training data without the need for an executable environment, aimed at training API-calling large language model (LLM)...","url":"https://www.aioga.com/en/news/cmrurrjvx04zdbi9t9c30lp8p/","contentTranslated":true,"sourceHash":"498277917fc098f5","translatedAt":"2026-07-23T06:42:37.785Z"},"ja":{"title":"Appleは環境のない合成データ生成方法を提案し、API呼び出し型LLMエージェントの訓練に使用する","summary":"Appleの研究者は、実行可能な環境を必要とせず、高品質なトレーニングデータを生成する方法を提案しました。この方法は、API 呼び出し型の大規模言語モデル（LLM）エージェントのトレーニングに用いられます。この手法は API 仕様書のみを使用し、LLM をデジタル世界のモデルとして活用し、教師エージェントと LLM シミュレーターの相互作用により軌跡を生成し、LLM レフェリーによってフィルタリングされます。AppWorld と OfficeBench のベンチマーク上で、微調整したモデルはこの合成データを使用することで顕著な性能向上を達成しました。","category":"論文研究","source":"Apple Machine Learning Research（RSS）","aggregationSource":"Apple Machine Learning Research（RSS）","pageTitle":"Appleは環境のない合成データ生成方法を提案し、API呼び出し型LLMエージェントの訓練に使用する - Aioga AIニュース","description":"Appleの研究者は、実行可能な環境を必要とせず、高品質なトレーニングデータを生成する方法を提案しました。この方法は、API 呼び出し型の大規模言語モデル（LLM）エージェントのトレーニングに用いられます。この手法は API 仕様書のみを使用し、LLM をデジタル世界のモデルとして活用し、教師エージェントと LLM シミュレーターの相互作用により軌跡を生成し...","url":"https://www.aioga.com/ja/news/cmrurrjvx04zdbi9t9c30lp8p/","contentTranslated":true,"sourceHash":"498277917fc098f5","translatedAt":"2026-07-23T06:42:46.689Z"},"ko":{"title":"Apple은 API 호출형 LLM 에이전트를 훈련하기 위해 환경 없는 합성 데이터 생성 방법을 제안했다","summary":"Apple 연구원들은 실행 환경이 필요 없이 높은 품질의 학습 데이터를 생성할 수 있는 방법을 제안했으며, 이는 API 호출형 대형 언어 모델(LLM) 에이전트를 학습시키는 데 사용됩니다. 이 방법은 API 명세만을 필요로 하며, LLM을 디지털 세계 모델로 활용하여 교사 에이전트와 LLM 시뮬레이터의 상호작용을 통해 궤적을 생성하고, LLM 심판이 이를 필터링합니다. AppWorld와 OfficeBench 벤치마크에서, 이 합성 데이터를 사용하여 미세 조정된 모델은 눈에 띄는 성능 향상을 달성했습니다.","category":"연구","source":"Apple Machine Learning Research（RSS）","aggregationSource":"Apple Machine Learning Research（RSS）","pageTitle":"Apple은 API 호출형 LLM 에이전트를 훈련하기 위해 환경 없는 합성 데이터 생성 방법을 제안했다 - Aioga AI 뉴스","description":"Apple 연구원들은 실행 환경이 필요 없이 높은 품질의 학습 데이터를 생성할 수 있는 방법을 제안했으며, 이는 API 호출형 대형 언어 모델(LLM) 에이전트를 학습시키는 데 사용됩니다. 이 방법은 API 명세만을 필요로 하며, LLM을 디지털 세계 모델로 활용하여 교사 에이전트와 LLM 시뮬레이터의 상호작용을 통해 궤...","url":"https://www.aioga.com/ko/news/cmrurrjvx04zdbi9t9c30lp8p/","contentTranslated":true,"sourceHash":"498277917fc098f5","translatedAt":"2026-07-23T06:43:32.542Z"},"es":{"title":"Apple propone un método de generación de datos sintéticos sin entorno, utilizado para entrenar agentes inteligentes LLM de tipo llamada a API","summary":"Los investigadores de Apple propusieron un método para generar datos de entrenamiento de alta calidad sin necesidad de un entorno ejecutable, destinado a entrenar agentes de modelos de lenguaje grande (LLM) que hacen llamadas a API. Este método solo requiere especificaciones de la API, utiliza el LLM como modelo del mundo digital, genera trayectorias mediante la interacción entre un agente maestro y un simulador LLM, y es filtrado por un árbitro LLM. En los benchmarks AppWorld y OfficeBench, los modelos ajustados con estos datos sintéticos lograron mejoras de rendimiento significativas.","category":"Investigación","source":"Apple Machine Learning Research（RSS）","aggregationSource":"Apple Machine Learning Research（RSS）","pageTitle":"Apple propone un método de generación de datos sintéticos sin entorno, utilizado para entrenar agentes inteligentes LLM de tipo llamada a API - Aioga Noticias de IA","description":"Los investigadores de Apple propusieron un método para generar datos de entrenamiento de alta calidad sin necesidad de un entorno ejecutable, destinado a entrenar agentes de modelo...","url":"https://www.aioga.com/es/news/cmrurrjvx04zdbi9t9c30lp8p/","contentTranslated":true,"sourceHash":"498277917fc098f5","translatedAt":"2026-07-23T06:43:35.115Z"},"fr":{"title":"Apple propose une méthode de génération de données synthétiques sans environnement, utilisée pour entraîner des agents intelligents LLM appelant des API","summary":"Les chercheurs d'Apple ont proposé une méthode pour générer des données d'entraînement de haute qualité sans environnement exécutable, destinée à former des agents de grands modèles de langage (LLM) appelés via API. Cette méthode nécessite uniquement une spécification de l'API et utilise le LLM comme modèle du monde numérique, en générant des trajectoires par l'interaction entre un agent enseignant et un simulateur LLM, puis filtrées par un arbitre LLM. Sur les benchmarks AppWorld et OfficeBench, le modèle affiné utilisant ces données synthétiques a obtenu une amélioration de performance significative.","category":"Recherche","source":"Apple Machine Learning Research（RSS）","aggregationSource":"Apple Machine Learning Research（RSS）","pageTitle":"Apple propose une méthode de génération de données synthétiques sans environnement, utilisée pour entraîner des agents intelligents LLM appelant des API - Aioga Actualités IA","description":"Les chercheurs d'Apple ont proposé une méthode pour générer des données d'entraînement de haute qualité sans environnement exécutable, destinée à former des agents de grands modèle...","url":"https://www.aioga.com/fr/news/cmrurrjvx04zdbi9t9c30lp8p/","contentTranslated":true,"sourceHash":"498277917fc098f5","translatedAt":"2026-07-23T06:44:26.853Z"},"de":{"title":"Apple schlägt eine Methode zur Generierung umweltfreier synthetischer Daten vor, die zur Schulung von API-Aufruf-LLM-Agenten verwendet wird","summary":"Apple-Forscher haben eine Methode vorgeschlagen, um qualitativ hochwertige Trainingsdaten zu erzeugen, ohne eine ausführbare Umgebung zu benötigen, die für das Training von API-Aufruf-basierten großen Sprachmodellagenten (LLM) verwendet werden kann. Diese Methode benötigt lediglich API-Spezifikationen und nutzt LLM als Modell der digitalen Welt. Durch die Interaktion eines Lehreragenten mit dem LLM-Simulator werden Trajektorien generiert, die anschließend von einem LLM-Schiedsrichter gefiltert werden. Auf den Benchmarks AppWorld und OfficeBench erzielte das feinabgestimmte Modell mit diesen synthetischen Daten eine signifikante Leistungssteigerung.","category":"论文研究","source":"Apple Machine Learning Research（RSS）","aggregationSource":"Apple Machine Learning Research（RSS）","pageTitle":"Apple schlägt eine Methode zur Generierung umweltfreier synthetischer Daten vor, die zur Schulung von API-Aufruf-LLM-Agenten verwendet wird - Aioga KI-News","description":"Apple-Forscher haben eine Methode vorgeschlagen, um qualitativ hochwertige Trainingsdaten zu erzeugen, ohne eine ausführbare Umgebung zu benötigen, die für das Training von API-Auf...","url":"https://www.aioga.com/de/news/cmrurrjvx04zdbi9t9c30lp8p/","contentTranslated":true,"sourceHash":"498277917fc098f5","translatedAt":"2026-07-23T06:44:21.407Z"},"pt-BR":{"title":"A Apple propôs um método de geração de dados sintéticos sem ambiente, usado para treinar agentes inteligentes LLM do tipo chamada de API","summary":"Pesquisadores da Apple propuseram um método para gerar dados de treinamento de alta qualidade sem a necessidade de um ambiente executável, destinado ao treinamento de agentes de modelo de linguagem grande (LLM) que realizam chamadas de API. Este método requer apenas a especificação da API, utilizando o LLM como um modelo do mundo digital, gerando trajetórias por meio da interação entre o agente professor e o simulador LLM, e filtradas pelo LLM árbitro. Nos benchmarks AppWorld e OfficeBench, o modelo ajustado com esses dados sintéticos obteve melhorias de desempenho significativas.","category":"论文研究","source":"Apple Machine Learning Research（RSS）","aggregationSource":"Apple Machine Learning Research（RSS）","pageTitle":"A Apple propôs um método de geração de dados sintéticos sem ambiente, usado para treinar agentes inteligentes LLM do tipo chamada de API - Aioga Notícias de IA","description":"Pesquisadores da Apple propuseram um método para gerar dados de treinamento de alta qualidade sem a necessidade de um ambiente executável, destinado ao treinamento de agentes de mo...","url":"https://www.aioga.com/pt-BR/news/cmrurrjvx04zdbi9t9c30lp8p/","contentTranslated":true,"sourceHash":"498277917fc098f5","translatedAt":"2026-07-23T06:45:10.672Z"},"ru":{"title":"Apple предложила метод генерации синтетических данных без учета окружающей среды, предназначенный для обучения интеллектуальных агентов LLM с вызовами API","summary":"Исследователи Apple предложили метод генерации высококачественных обучающих данных без необходимости исполняемой среды, предназначенный для обучения агентов больших языковых моделей (LLM) с вызовом API. Метод требует только спецификации API, использует LLM в качестве модели цифрового мира, генерирует траектории через взаимодействие симулированного учителя с эмулятором LLM и фильтруется судьей LLM. На бенчмарках AppWorld и OfficeBench модели, дообученные с использованием этих синтетических данных, показали значительное улучшение производительности.","category":"论文研究","source":"Apple Machine Learning Research（RSS）","aggregationSource":"Apple Machine Learning Research（RSS）","pageTitle":"Apple предложила метод генерации синтетических данных без учета окружающей среды, предназначенный для обучения интеллектуальных агентов LLM с вызовами API - Aioga Новости ИИ","description":"Исследователи Apple предложили метод генерации высококачественных обучающих данных без необходимости исполняемой среды, предназначенный для обучения агентов больших языковых моделе...","url":"https://www.aioga.com/ru/news/cmrurrjvx04zdbi9t9c30lp8p/","contentTranslated":true,"sourceHash":"498277917fc098f5","translatedAt":"2026-07-23T06:45:12.436Z"},"ar":{"title":"قدمت شركة آبل طريقة لتوليد بيانات تركيبية بدون بيئة، لاستخدامها في تدريب وكلاء LLM الذين يستدعون واجهات برمجة التطبيقات","summary":"اقترح باحثو شركة آبل طريقة لتوليد بيانات تدريب عالية الجودة دون الحاجة إلى بيئة تنفيذية، وذلك لتدريب وكلاء نماذج اللغة الكبيرة (LLM) التي تستدعي واجهات برمجة التطبيقات (API). تتطلب هذه الطريقة فقط مواصفات API، وتستخدم نموذج اللغة الكبير كنموذج للعالم الرقمي، من خلال تفاعل وكيل المعلم مع محاكي LLM لتوليد المسارات، ويتم تصفيتها بواسطة حكَم LLM. على معايير AppWorld وOfficeBench، حققت النماذج المعدلة باستخدام هذه البيانات المُولَّدة تحسنًا ملحوظًا في الأداء.","category":"论文研究","source":"Apple Machine Learning Research（RSS）","aggregationSource":"Apple Machine Learning Research（RSS）","pageTitle":"قدمت شركة آبل طريقة لتوليد بيانات تركيبية بدون بيئة، لاستخدامها في تدريب وكلاء LLM الذين يستدعون واجهات برمجة التطبيقات - Aioga أخبار الذكاء الاصطناعي","description":"اقترح باحثو شركة آبل طريقة لتوليد بيانات تدريب عالية الجودة دون الحاجة إلى بيئة تنفيذية، وذلك لتدريب وكلاء نماذج اللغة الكبيرة (LLM) التي تستدعي واجهات برمجة التطبيقات (API). تتطلب...","url":"https://www.aioga.com/ar/news/cmrurrjvx04zdbi9t9c30lp8p/","contentTranslated":true,"sourceHash":"498277917fc098f5","translatedAt":"2026-07-23T06:46:01.267Z"},"hi":{"title":"एप्पल ने पर्यावरण-रहित सिंथेटिक डेटा निर्माण विधि प्रस्तावित की, जिसका उपयोग API कॉल-आधारित LLM एजेंट को प्रशिक्षित करने के लिए किया जाता है","summary":"Apple के शोधकर्ताओं ने एक ऐसी विधि प्रस्तावित की है जिसके लिए निष्पादन वातावरण की आवश्यकता नहीं है और जो उच्च गुणवत्ता वाले प्रशिक्षण डेटा उत्पन्न कर सकती है, जिसका उपयोग API कॉल-आधारित बड़े भाषा मॉडल (LLM) एजेंट को प्रशिक्षित करने के लिए किया जा सकता है। इस विधि के लिए केवल API विनिर्देश की आवश्यकता होती है, और यह LLM का उपयोग डिजिटल दुनिया के मॉडल के रूप में करती है, शिक्षक एजेंट और LLM सिमुलेटर के बीच इंटरैक्शन के माध्यम से ट्रैजेक्टरी उत्पन्न करती है, जिसे LLM न्यायाधीश द्वारा फ़िल्टर किया जाता है। AppWorld और OfficeBench मानकों पर, सूक्ष्म-संशोधित मॉडल ने इस संश्लेषित डेटा का उपयोग करके उल्लेखनीय प्रदर्शन सुधार हासिल किया।","category":"论文研究","source":"Apple Machine Learning Research（RSS）","aggregationSource":"Apple Machine Learning Research（RSS）","pageTitle":"एप्पल ने पर्यावरण-रहित सिंथेटिक डेटा निर्माण विधि प्रस्तावित की, जिसका उपयोग API कॉल-आधारित LLM एजेंट को प्रशिक्षित करने के लिए किया जाता है - Aioga AI समाचार","description":"Apple के शोधकर्ताओं ने एक ऐसी विधि प्रस्तावित की है जिसके लिए निष्पादन वातावरण की आवश्यकता नहीं है और जो उच्च गुणवत्ता वाले प्रशिक्षण डेटा उत्पन्न कर सकती है, जिसका उपयोग API कॉल-आ...","url":"https://www.aioga.com/hi/news/cmrurrjvx04zdbi9t9c30lp8p/","contentTranslated":true,"sourceHash":"498277917fc098f5","translatedAt":"2026-07-23T06:46:03.660Z"},"it":{"title":"Apple propone un metodo di generazione di dati sintetici senza ambiente, utilizzato per addestrare agenti intelligenti LLM che effettuano chiamate API","summary":"I ricercatori di Apple hanno proposto un metodo per generare dati di addestramento di alta qualità senza la necessità di un ambiente eseguibile, destinato all'addestramento di agenti di grandi modelli linguistici (LLM) per chiamate API. Questo metodo richiede solamente le specifiche dell'API e utilizza LLM come modello del mondo digitale, generando traiettorie tramite l'interazione tra un agente insegnante e un simulatore LLM, con la filtrazione affidata a un arbitro LLM. Sui benchmark AppWorld e OfficeBench, il modello fine-tuned che ha utilizzato questi dati sintetici ha ottenuto un miglioramento significativo delle prestazioni.","category":"论文研究","source":"Apple Machine Learning Research（RSS）","aggregationSource":"Apple Machine Learning Research（RSS）","pageTitle":"Apple propone un metodo di generazione di dati sintetici senza ambiente, utilizzato per addestrare agenti intelligenti LLM che effettuano chiamate API - Aioga Notizie IA","description":"I ricercatori di Apple hanno proposto un metodo per generare dati di addestramento di alta qualità senza la necessità di un ambiente eseguibile, destinato all'addestramento di agen...","url":"https://www.aioga.com/it/news/cmrurrjvx04zdbi9t9c30lp8p/","contentTranslated":true,"sourceHash":"498277917fc098f5","translatedAt":"2026-07-23T06:46:49.124Z"},"nl":{"title":"Apple stelt een methode voor het genereren van omgevingsvrije synthetische gegevens voor het trainen van API-aanroepende LLM-agenten","summary":"Apple-onderzoekers hebben een methode voorgesteld om hoogwaardige trainingsgegevens te genereren zonder een uitvoeringsomgeving, bedoeld voor het trainen van grote taalmodellen (LLM) die API-aanroepen doen. Deze methode heeft alleen de API-specificaties nodig en gebruikt LLM als model van de digitale wereld, waarbij trajecten worden gegenereerd door interactie tussen een docentagent en de LLM-simulator, en vervolgens door de LLM-scheidsrechter worden gefilterd. Op de benchmarks AppWorld en OfficeBench behaalde het fijn afgestelde model met deze synthetische gegevens een aanzienlijke prestatieverbetering.","category":"论文研究","source":"Apple Machine Learning Research（RSS）","aggregationSource":"Apple Machine Learning Research（RSS）","pageTitle":"Apple stelt een methode voor het genereren van omgevingsvrije synthetische gegevens voor het trainen van API-aanroepende LLM-agenten - Aioga AI-nieuws","description":"Apple-onderzoekers hebben een methode voorgesteld om hoogwaardige trainingsgegevens te genereren zonder een uitvoeringsomgeving, bedoeld voor het trainen van grote taalmodellen (LL...","url":"https://www.aioga.com/nl/news/cmrurrjvx04zdbi9t9c30lp8p/","contentTranslated":true,"sourceHash":"498277917fc098f5","translatedAt":"2026-07-23T06:46:49.864Z"},"tr":{"title":"Apple, API çağrısı yapan LLM ajanlarını eğitmek için ortamdan bağımsız sentez veri üretme yöntemi önerdi","summary":"Apple araştırmacıları, API çağrısı yapabilen büyük dil modeli (LLM) ajanlarını eğitmek için yüksek kaliteli eğitim verileri üretmeye yönelik, yürütülebilir bir ortama ihtiyaç duymayan bir yöntem önerdi. Bu yöntem yalnızca API spesifikasyonunu gerektirir, LLM'yi dijital dünya modeli olarak kullanır, öğretici ajan ile LLM simülatörü arasındaki etkileşimle izler üretir ve LLM hakem tarafından filtrelenir. AppWorld ve OfficeBench ölçütlerinde, bu sentetik verileri kullanan ince ayar modeli önemli performans artışı sağlamıştır.","category":"论文研究","source":"Apple Machine Learning Research（RSS）","aggregationSource":"Apple Machine Learning Research（RSS）","pageTitle":"Apple, API çağrısı yapan LLM ajanlarını eğitmek için ortamdan bağımsız sentez veri üretme yöntemi önerdi - Aioga AI Haberleri","description":"Apple araştırmacıları, API çağrısı yapabilen büyük dil modeli (LLM) ajanlarını eğitmek için yüksek kaliteli eğitim verileri üretmeye yönelik, yürütülebilir bir ortama ihtiyaç duyma...","url":"https://www.aioga.com/tr/news/cmrurrjvx04zdbi9t9c30lp8p/","contentTranslated":true,"sourceHash":"498277917fc098f5","translatedAt":"2026-07-23T06:47:37.269Z"},"vi":{"title":"Apple đề xuất phương pháp tạo dữ liệu tổng hợp không môi trường, dùng để huấn luyện đại lý thông minh LLM gọi API","summary":"Các nhà nghiên cứu của Apple đề xuất một phương pháp tạo dữ liệu huấn luyện chất lượng cao mà không cần môi trường thực thi, dùng để huấn luyện các đại mô hình ngôn ngữ (LLM) dạng gọi API. Phương pháp này chỉ cần thông số kỹ thuật API, sử dụng LLM làm mô hình thế giới số, thông qua tương tác giữa tác nhân giáo viên và bộ mô phỏng LLM để tạo ra các quỹ đạo, và sau đó được LLM trọng tài lọc. Trên các chuẩn AppWorld và OfficeBench, mô hình tinh chỉnh sử dụng dữ liệu tổng hợp này đã đạt được cải thiện hiệu suất đáng kể.","category":"论文研究","source":"Apple Machine Learning Research（RSS）","aggregationSource":"Apple Machine Learning Research（RSS）","pageTitle":"Apple đề xuất phương pháp tạo dữ liệu tổng hợp không môi trường, dùng để huấn luyện đại lý thông minh LLM gọi API - Tin tức AI Aioga","description":"Các nhà nghiên cứu của Apple đề xuất một phương pháp tạo dữ liệu huấn luyện chất lượng cao mà không cần môi trường thực thi, dùng để huấn luyện các đại mô hình ngôn ngữ (LLM) dạng...","url":"https://www.aioga.com/vi/news/cmrurrjvx04zdbi9t9c30lp8p/","contentTranslated":true,"sourceHash":"498277917fc098f5","translatedAt":"2026-07-23T06:47:33.095Z"},"id":{"title":"Apple mengusulkan metode pembuatan data sintetis tanpa lingkungan, digunakan untuk melatih agen cerdas LLM tipe panggilan API","summary":"Peneliti Apple mengusulkan metode untuk menghasilkan data pelatihan berkualitas tinggi tanpa memerlukan lingkungan eksekusi, yang digunakan untuk melatih agen model bahasa besar (LLM) jenis pemanggilan API. Metode ini hanya memerlukan spesifikasi API, menggunakan LLM sebagai model dunia digital, melalui interaksi agen guru dengan simulator LLM untuk menghasilkan jejak, dan disaring oleh hakim LLM. Pada tolok ukur AppWorld dan OfficeBench, model yang disesuaikan menggunakan data sintetis ini mencapai peningkatan kinerja yang signifikan.","category":"论文研究","source":"Apple Machine Learning Research（RSS）","aggregationSource":"Apple Machine Learning Research（RSS）","pageTitle":"Apple mengusulkan metode pembuatan data sintetis tanpa lingkungan, digunakan untuk melatih agen cerdas LLM tipe panggilan API - Berita AI Aioga","description":"Peneliti Apple mengusulkan metode untuk menghasilkan data pelatihan berkualitas tinggi tanpa memerlukan lingkungan eksekusi, yang digunakan untuk melatih agen model bahasa besar (L...","url":"https://www.aioga.com/id/news/cmrurrjvx04zdbi9t9c30lp8p/","contentTranslated":true,"sourceHash":"498277917fc098f5","translatedAt":"2026-07-23T06:48:25.752Z"},"th":{"title":"Apple นำเสนอวิธีการสร้างข้อมูลสังเคราะห์ที่ไม่ใช้สิ่งแวดล้อม สำหรับฝึกผู้ปัญญาประดิษฐ์แบบเรียกใช้ API","summary":"นักวิจัยของ Apple ได้นำเสนอวิธีการสร้างข้อมูลการฝึกอบรมคุณภาพสูงโดยไม่ต้องใช้สภาพแวดล้อมที่สามารถรันได้ สำหรับฝึกฝนเอเจนต์ LLM ที่เรียกใช้ API วิธีการนี้เพียงแค่ต้องการสเปคของ API ใช้ LLM เป็นโมเดลของโลกดิจิทัล ผ่านการปฏิสัมพันธ์ระหว่างเอเจนต์ครูและจำลอง LLM เพื่อสร้างเส้นทาง จากนั้น LLM จะทำหน้าที่ตัดสินและกรอง ในเกณฑ์มาตรฐาน AppWorld และ OfficeBench โมเดลที่ผ่านการปรับจูนโดยใช้ข้อมูลสังเคราะห์นี้สามารถทำให้ประสิทธิภาพดีขึ้นอย่างมีนัยสำคัญ","category":"论文研究","source":"Apple Machine Learning Research（RSS）","aggregationSource":"Apple Machine Learning Research（RSS）","pageTitle":"Apple นำเสนอวิธีการสร้างข้อมูลสังเคราะห์ที่ไม่ใช้สิ่งแวดล้อม สำหรับฝึกผู้ปัญญาประดิษฐ์แบบเรียกใช้ API - ข่าว AI Aioga","description":"นักวิจัยของ Apple ได้นำเสนอวิธีการสร้างข้อมูลการฝึกอบรมคุณภาพสูงโดยไม่ต้องใช้สภาพแวดล้อมที่สามารถรันได้ สำหรับฝึกฝนเอเจนต์ LLM ที่เรียกใช้ API วิธีการนี้เพียงแค่ต้องการสเปคของ API...","url":"https://www.aioga.com/th/news/cmrurrjvx04zdbi9t9c30lp8p/","contentTranslated":true,"sourceHash":"498277917fc098f5","translatedAt":"2026-07-23T06:48:26.285Z"},"pl":{"title":"Apple przedstawiło metodę generowania syntetycznych danych bez środowiska, przeznaczoną do szkolenia agentów inteligentnych typu wywołującego API opartych na LLM","summary":"Badacze Apple zaproponowali metodę generowania wysokiej jakości danych treningowych bez potrzeby środowiska wykonawczego, przeznaczoną do trenowania agentów dużych modeli językowych (LLM) wywołujących API. Metoda ta wymaga jedynie specyfikacji API, wykorzystuje LLM jako model cyfrowego świata, generuje trajektorie poprzez interakcję agenta nauczyciela z symulatorem LLM, a następnie filtruje je sędzia LLM. Na benchmarkach AppWorld i OfficeBench, modele dostrajane przy użyciu tych danych syntetycznych osiągnęły znaczącą poprawę wydajności.","category":"论文研究","source":"Apple Machine Learning Research（RSS）","aggregationSource":"Apple Machine Learning Research（RSS）","pageTitle":"Apple przedstawiło metodę generowania syntetycznych danych bez środowiska, przeznaczoną do szkolenia agentów inteligentnych typu wywołującego API opartych na LLM - Aioga Wiadomości AI","description":"Badacze Apple zaproponowali metodę generowania wysokiej jakości danych treningowych bez potrzeby środowiska wykonawczego, przeznaczoną do trenowania agentów dużych modeli językowyc...","url":"https://www.aioga.com/pl/news/cmrurrjvx04zdbi9t9c30lp8p/","contentTranslated":true,"sourceHash":"498277917fc098f5","translatedAt":"2026-07-23T06:49:15.314Z"}}}}