{"@context":"https://schema.org","@type":"NewsArticle","generatedAt":"2026-07-27T21:40:51.174Z","headline":"MineExplorer：首个《我的世界》分钟级长程任务评测基准发布","description":"美团 LongCat 团队发布 MineExplorer，这是首个在《我的世界》开放世界中实现分钟级长程任务的评测基准，包含 813 个人工验证实例。评测 18 款顶级多模态大模型发现，最强模型 Claude-Opus-4.6 整体任务成功率仅 41%，从 1 跳任务的 77% 骤降至 4 跳任务的 12%，近 60% 的失败源于导航失败。MineExplorer 已全面开源。","url":"https://www.aioga.com/news/cmryewabt00mnrolg1nxcjf18/","mainEntityOfPage":"https://www.aioga.com/news/cmryewabt00mnrolg1nxcjf18/","datePublished":"2026-07-23T11:58:00.000Z","dateModified":"2026-07-23T11:58:00.000Z","inLanguage":"zh-CN","publisher":{"@type":"NewsMediaOrganization","name":"Aioga","url":"https://www.aioga.com"},"citation":["https://mp.weixin.qq.com/s/P3yzceXkVxth7Q63nRfBLg","https://aihot.virxact.com/items/cmryewabt00mnrolg1nxcjf18"],"canonicalUrl":"https://www.aioga.com/news/cmryewabt00mnrolg1nxcjf18/","directAnswer":{"@type":"Answer","text":"美团 LongCat 团队发布 MineExplorer，面向《我的世界》开放世界中的分钟级长程任务，提供包含 813 个人工验证实例的评测基准。对 18 款顶级多模态大模型的测试显示，最强模型整体成功率为 41%。","url":"https://www.aioga.com/news/cmryewabt00mnrolg1nxcjf18/","dateCreated":"2026-07-23T11:58:00.000Z","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":"公众号：龙猫LongCat（美团） source article","url":"https://mp.weixin.qq.com/s/P3yzceXkVxth7Q63nRfBLg","datePublished":"2026-07-23T11:58:00.000Z","provider":{"@type":"Organization","name":"公众号：龙猫LongCat（美团）","url":"https://mp.weixin.qq.com/s/P3yzceXkVxth7Q63nRfBLg"}},{"@type":"CreativeWork","name":"AIHot archive record","url":"https://aihot.virxact.com/items/cmryewabt00mnrolg1nxcjf18","datePublished":"2026-07-23T11:58:00.000Z","provider":{"@type":"Organization","name":"AIHot","url":"https://aihot.virxact.com/items/cmryewabt00mnrolg1nxcjf18"}}],"aggregationSource":"公众号：龙猫LongCat（美团）","originalPublisher":{"name":"公众号：龙猫LongCat（美团）","url":"https://mp.weixin.qq.com/s/P3yzceXkVxth7Q63nRfBLg"},"article":{"id":"cmryewabt00mnrolg1nxcjf18","slug":"cmryewabt00mnrolg1nxcjf18","url":"https://www.aioga.com/news/cmryewabt00mnrolg1nxcjf18/","title":"MineExplorer：首个《我的世界》分钟级长程任务评测基准发布","title_en":"让AI离开温室，走向动态世界：MineExplorer揭示顶级多模态大模型被忽视的能力断层","summary":"美团 LongCat 团队发布 MineExplorer，这是首个在《我的世界》开放世界中实现分钟级长程任务的评测基准，包含 813 个人工验证实例。评测 18 款顶级多模态大模型发现，最强模型 Claude-Opus-4.6 整体任务成功率仅 41%，从 1 跳任务的 77% 骤降至 4 跳任务的 12%，近 60% 的失败源于导航失败。MineExplorer 已全面开源。","source":"公众号：龙猫LongCat（美团）","sourceUrl":"https://mp.weixin.qq.com/s/P3yzceXkVxth7Q63nRfBLg","aiHotUrl":"https://aihot.virxact.com/items/cmryewabt00mnrolg1nxcjf18","publishedAt":"2026-07-23T11:58:00.000Z","category":"论文研究","score":68,"selected":true,"articleBody":["美团 LongCat 团队发布 MineExplorer，这是首个在《我的世界》开放世界中实现分钟级长程任务的评测基准，包含 813 个人工验证实例。","评测 18 款顶级多模态大模型发现，最强模型 Claude-Opus-4.6 整体任务成功率仅 41%，从 1 跳任务的 77% 骤降至 4 跳任务的 12%，近 60% 的失败源于导航失败。","MineExplorer 已全面开源。"],"articleImages":[],"mediaStatus":"none","articleBodyZh":[],"translationStatus":"","bodyOrigin":"summary-fallback","editorial":{"summary":"美团 LongCat 团队发布 MineExplorer，面向《我的世界》开放世界中的分钟级长程任务，提供包含 813 个人工验证实例的评测基准。对 18 款顶级多模态大模型的测试显示，最强模型整体成功率为 41%。","background":"材料显示，Claude-Opus-4.6 在 1 跳任务中的成功率为 77%，到 4 跳任务降至 12%；近 60% 的失败源于导航。MineExplorer 已全面开源，评测聚焦开放世界中的连续任务执行能力。","viewpoint":"Aioga 判断，MineExplorer 的价值在于把多模态模型从短步骤响应带入更长链路、开放环境下检验。成功率随任务跳数明显下降，可能反映规划、执行衔接与空间导航仍是关键限制。","implications":"这项评测可能促使模型能力比较从单步完成扩展到长程任务稳定性，也为导航失败等具体问题提供可量化观察维度。由于材料未说明开源范围和使用方式，相关影响仍需结合实际复现判断。","nextStep":"值得关注后续基于 MineExplorer 的复测结果，包括不同模型在任务跳数增加后的表现，以及导航失败是否持续成为主要失败来源。对开源基准的实际采用情况，也应等待更多公开材料确认。","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":"2026-07-26T17:49:52.985Z","sourceHash":"d50d134225c66a53","review":{"approved":true,"groundedness":96,"clarity":93,"duplicationRisk":18,"blockingIssues":[],"notes":["“可能反映规划、执行衔接与空间导航仍是关键限制”属于基于评测结果的合理分析，已使用“可能”限定，没有冒充来源事实。","“从短步骤响应带入更长链路、开放环境下检验”属于观点性概括，并已通过“Aioga 判断”明确标注。","候选内容未提及来源所称“首个”评测基准，属于信息取舍，不构成事实错误。"]},"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","公众号：龙猫LongCat（美团）"],"translations":{"zh-CN":{"title":"MineExplorer：首个《我的世界》分钟级长程任务评测基准发布","summary":"美团 LongCat 团队发布 MineExplorer，这是首个在《我的世界》开放世界中实现分钟级长程任务的评测基准，包含 813 个人工验证实例。评测 18 款顶级多模态大模型发现，最强模型 Claude-Opus-4.6 整体任务成功率仅 41%，从 1 跳任务的 77% 骤降至 4 跳任务的 12%，近 60% 的失败源于导航失败。MineExplorer 已全面开源。","category":"论文研究","source":"公众号：龙猫LongCat（美团）","aggregationSource":"公众号：龙猫LongCat（美团）","pageTitle":"MineExplorer：首个《我的世界》分钟级长程任务评测基准发布 - Aioga AI资讯","description":"美团 LongCat 团队发布 MineExplorer，这是首个在《我的世界》开放世界中实现分钟级长程任务的评测基准，包含 813 个人工验证实例。评测 18 款顶级多模态大模型发现，最强模型 Claude-Opus-4.6 整体任务成功率仅 41%，从 1 跳任务的 77% 骤降至 4 跳任务的 12%，近 60% 的失败源于导航失败。MineExplo...","url":"https://www.aioga.com/news/cmryewabt00mnrolg1nxcjf18/"},"en":{"title":"MineExplorer: The first minute-level long-term mission evaluation benchmark for 'Minecraft' released","summary":"The Meituan LongCat team released MineExplorer, the first evaluation benchmark to achieve minute-level long-range tasks in the open world of Minecraft, containing 813 manually verified instances. Evaluating 18 top multimodal large models found that the strongest model, Claude-Opus-4.6, had an overall task success rate of only 41%, dropping sharply from 77% for 1-hop tasks to 12% for 4-hop tasks, with nearly 60% of failures due to navigation errors. MineExplorer has been fully open-sourced.","category":"Research","source":"公众号：龙猫LongCat（美团）","aggregationSource":"公众号：龙猫LongCat（美团）","pageTitle":"MineExplorer: The first minute-level long-term mission evaluation benchmark for 'Minecraft' released - Aioga AI News","description":"The Meituan LongCat team released MineExplorer, the first evaluation benchmark to achieve minute-level long-range tasks in the open world of Minecraft, containing 813 manually veri...","url":"https://www.aioga.com/en/news/cmryewabt00mnrolg1nxcjf18/","contentTranslated":true,"sourceHash":"a7096d6bbdcff7dd","translatedAt":"2026-07-26T02:43:47.876Z"},"ja":{"title":"MineExplorer：初の『Minecraft』分単位長距離ミッション評価ベンチマークを公開","summary":"美団のLongCatチームはMineExplorerを発表しました。これは『マインクラフト』のオープンワールドで分単位の長距離タスクを実現する初の評価ベンチマークで、813件の人工検証済みインスタンスを含みます。18種類のトップマルチモーダル大規模モデルを評価した結果、最強モデルClaude-Opus-4.6でも全体タスク成功率はわずか41%で、1ステップタスクの77%から4ステップタスクでは12%に急落し、約60%の失敗はナビゲーションの失敗に起因しています。MineExplorerは完全にオープンソース化されています。","category":"論文研究","source":"公众号：龙猫LongCat（美团）","aggregationSource":"公众号：龙猫LongCat（美团）","pageTitle":"MineExplorer：初の『Minecraft』分単位長距離ミッション評価ベンチマークを公開 - Aioga AIニュース","description":"美団のLongCatチームはMineExplorerを発表しました。これは『マインクラフト』のオープンワールドで分単位の長距離タスクを実現する初の評価ベンチマークで、813件の人工検証済みインスタンスを含みます。18種類のトップマルチモーダル大規模モデルを評価した結果、最強モデルClaude-Opus-4.6でも全体タスク成功率はわずか41%で、1ステップタ...","url":"https://www.aioga.com/ja/news/cmryewabt00mnrolg1nxcjf18/","contentTranslated":true,"sourceHash":"a7096d6bbdcff7dd","translatedAt":"2026-07-26T02:44:02.045Z"},"ko":{"title":"MineExplorer：첫 번째 《마인크래프트》 분 단위 장기 임무 평가 기준 발표","summary":"메이투안 LongCat 팀이 MineExplorer를 출시했습니다. 이는 《마인크래프트》 오픈 월드에서 분 단위 장거리 임무를 구현한 최초의 평가 기준으로, 813개의 인공 검증 사례를 포함하고 있습니다. 18종의 최고급 멀티모달 대형 모델을 평가한 결과, 가장 강력한 모델 Claude-Opus-4.6의 전체 임무 성공률은 단 41%였으며, 1점 점프 임무의 77%에서 4점 점프 임무의 12%로 급격히 감소했고, 실패의 약 60%가 내비게이션 실패에서 비롯되었습니다. MineExplorer는 이미 전면적으로 오픈 소스화되었습니다.","category":"연구","source":"公众号：龙猫LongCat（美团）","aggregationSource":"公众号：龙猫LongCat（美团）","pageTitle":"MineExplorer：첫 번째 《마인크래프트》 분 단위 장기 임무 평가 기준 발표 - Aioga AI 뉴스","description":"메이투안 LongCat 팀이 MineExplorer를 출시했습니다. 이는 《마인크래프트》 오픈 월드에서 분 단위 장거리 임무를 구현한 최초의 평가 기준으로, 813개의 인공 검증 사례를 포함하고 있습니다. 18종의 최고급 멀티모달 대형 모델을 평가한 결과, 가장 강력한 모델 Claude-Opus-4.6의 전체 임무 성공률...","url":"https://www.aioga.com/ko/news/cmryewabt00mnrolg1nxcjf18/","contentTranslated":true,"sourceHash":"a7096d6bbdcff7dd","translatedAt":"2026-07-26T02:44:39.595Z"},"es":{"title":"MineExplorer: Se publica el primer referente de evaluación de misiones de larga duración por minutos de 'Minecraft'","summary":"El equipo LongCat de Meituan lanzó MineExplorer, el primer benchmark que implementa tareas de largo alcance a nivel de minutos en el mundo abierto de \"Minecraft\", que incluye 813 casos de verificación manual. La evaluación de 18 modelos multimodales de primer nivel mostró que el modelo más fuerte, Claude-Opus-4.6, tuvo una tasa de éxito general en las tareas de solo 41%, descendiendo drásticamente del 77% en tareas de un salto al 12% en tareas de cuatro saltos, y casi el 60% de los fracasos se debe a errores de navegación. MineExplorer ya se encuentra completamente abierto al público.","category":"Investigación","source":"公众号：龙猫LongCat（美团）","aggregationSource":"公众号：龙猫LongCat（美团）","pageTitle":"MineExplorer: Se publica el primer referente de evaluación de misiones de larga duración por minutos de 'Minecraft' - Aioga Noticias de IA","description":"El equipo LongCat de Meituan lanzó MineExplorer, el primer benchmark que implementa tareas de largo alcance a nivel de minutos en el mundo abierto de \"Minecraft\", que incluye 813 c...","url":"https://www.aioga.com/es/news/cmryewabt00mnrolg1nxcjf18/","contentTranslated":true,"sourceHash":"a7096d6bbdcff7dd","translatedAt":"2026-07-26T02:44:35.624Z"},"fr":{"title":"MineExplorer : Première publication de référence d'évaluation des missions longues à l'échelle minute dans Minecraft","summary":"L'équipe LongCat de Meituan a publié MineExplorer, le premier benchmark de tâches longues à l'échelle de minutes dans le monde ouvert de Minecraft, comprenant 813 instances vérifiées manuellement. L'évaluation de 18 grands modèles multimodaux de premier plan a révélé que le modèle le plus performant, Claude-Opus-4.6, avait un taux de réussite global de seulement 41 %, chutant de 77 % pour les tâches en 1 saut à 12 % pour les tâches en 4 sauts, près de 60 % des échecs étant dus à des erreurs de navigation. MineExplorer est désormais entièrement open source.","category":"Recherche","source":"公众号：龙猫LongCat（美团）","aggregationSource":"公众号：龙猫LongCat（美团）","pageTitle":"MineExplorer : Première publication de référence d'évaluation des missions longues à l'échelle minute dans Minecraft - Aioga Actualités IA","description":"L'équipe LongCat de Meituan a publié MineExplorer, le premier benchmark de tâches longues à l'échelle de minutes dans le monde ouvert de Minecraft, comprenant 813 instances vérifié...","url":"https://www.aioga.com/fr/news/cmryewabt00mnrolg1nxcjf18/","contentTranslated":true,"sourceHash":"a7096d6bbdcff7dd","translatedAt":"2026-07-26T02:45:21.231Z"},"de":{"title":"MineExplorer: Erste Minutenlange Langstrecken-Aufgabenbewertungs-Benchmark für ‚Minecraft‘ veröffentlicht","summary":"Das Meituan LongCat-Team hat MineExplorer veröffentlicht, das erste Bewertungsbenchmark, das minutengenaue Langstreckenaufgaben in der offenen Welt von „Minecraft“ realisiert und 813 manuell verifizierte Beispiele enthält. Beim Testen von 18 führenden multimodalen Großmodellen zeigte sich, dass das stärkste Modell Claude-Opus-4.6 eine Gesamtaufgabenerfolgsrate von nur 41 % erreicht, wobei die Erfolgsrate von Aufgaben mit 1 Sprung bei 77 % auf Aufgaben mit 4 Sprüngen auf 12 % sinkt, und fast 60 % der Misserfolge auf Navigationsfehler zurückzuführen sind. MineExplorer ist vollständig Open Source verfügbar.","category":"论文研究","source":"公众号：龙猫LongCat（美团）","aggregationSource":"公众号：龙猫LongCat（美团）","pageTitle":"MineExplorer: Erste Minutenlange Langstrecken-Aufgabenbewertungs-Benchmark für ‚Minecraft‘ veröffentlicht - Aioga KI-News","description":"Das Meituan LongCat-Team hat MineExplorer veröffentlicht, das erste Bewertungsbenchmark, das minutengenaue Langstreckenaufgaben in der offenen Welt von „Minecraft“ realisiert und 8...","url":"https://www.aioga.com/de/news/cmryewabt00mnrolg1nxcjf18/","contentTranslated":true,"sourceHash":"a7096d6bbdcff7dd","translatedAt":"2026-07-26T02:45:17.936Z"},"pt-BR":{"title":"MineExplorer: Primeiro benchmark de avaliação de missão de longa duração em minutos de Minecraft publicado","summary":"A equipe LongCat da Meituan lançou o MineExplorer, que é o primeiro benchmark de avaliação a implementar tarefas de longo alcance em minutos no mundo aberto de Minecraft, contendo 813 casos verificados manualmente. Avaliando 18 dos principais modelos multimodais, descobriu-se que o modelo mais forte, Claude-Opus-4.6, teve uma taxa de sucesso geral de apenas 41%, caindo de 77% em tarefas de 1 salto para 12% em tarefas de 4 saltos, com quase 60% das falhas decorrentes de falhas de navegação. O MineExplorer já está totalmente disponível como código aberto.","category":"论文研究","source":"公众号：龙猫LongCat（美团）","aggregationSource":"公众号：龙猫LongCat（美团）","pageTitle":"MineExplorer: Primeiro benchmark de avaliação de missão de longa duração em minutos de Minecraft publicado - Aioga Notícias de IA","description":"A equipe LongCat da Meituan lançou o MineExplorer, que é o primeiro benchmark de avaliação a implementar tarefas de longo alcance em minutos no mundo aberto de Minecraft, contendo...","url":"https://www.aioga.com/pt-BR/news/cmryewabt00mnrolg1nxcjf18/","contentTranslated":true,"sourceHash":"a7096d6bbdcff7dd","translatedAt":"2026-07-26T02:45:55.186Z"},"ru":{"title":"MineExplorer: опубликован первый минутный долгосрочный тестовый эталон для «Minecraft»","summary":"Команда LongCat компании Meituan выпустила MineExplorer — это первый эталон оценки для выполнения долгосрочных задач за минуты в открытом мире Minecraft, включающий 813 проверенных вручную примеров. Оценка 18 ведущих многомодальных крупных моделей показала, что самая мощная модель Claude-Opus-4.6 имеет общий коэффициент успешного выполнения задач всего 41%, при этом для задач с одним прыжком успех составляет 77%, а для задач с четырьмя прыжками падает до 12%; почти 60% неудач связаны с ошибками навигации. MineExplorer полностью открыт.","category":"论文研究","source":"公众号：龙猫LongCat（美团）","aggregationSource":"公众号：龙猫LongCat（美团）","pageTitle":"MineExplorer: опубликован первый минутный долгосрочный тестовый эталон для «Minecraft» - Aioga Новости ИИ","description":"Команда LongCat компании Meituan выпустила MineExplorer — это первый эталон оценки для выполнения долгосрочных задач за минуты в открытом мире Minecraft, включающий 813 проверенных...","url":"https://www.aioga.com/ru/news/cmryewabt00mnrolg1nxcjf18/","contentTranslated":true,"sourceHash":"a7096d6bbdcff7dd","translatedAt":"2026-07-26T02:46:00.564Z"},"ar":{"title":"MineExplorer: إطلاق أول معيار تقييم لمهام طويلة المدى بالدقائق في \"ماين كرافت\"","summary":"أصدرت فريق مي توان LongCat أداة MineExplorer، وهي أول معيار تقييم للمهام طويلة المدى بالدقائق في العالم المفتوح للعبة \"ماين كرافت\"، ويحتوي على 813 مثالًا تم التحقق منه يدويًا. عند تقييم 18 نموذجًا متعدد الوسائط من الطراز الأعلى، وُجد أن أقوى نموذج Claude-Opus-4.6 لديه معدل نجاح عام للمهام يبلغ 41% فقط، حيث انخفض من 77% للمهام المؤلفة من خطوة واحدة إلى 12% للمهام المؤلفة من أربع خطوات، ويعود نحو 60% من حالات الفشل إلى فشل في التنقل. وقد تم فتح المصدر لجميع أدوات MineExplorer.","category":"论文研究","source":"公众号：龙猫LongCat（美团）","aggregationSource":"公众号：龙猫LongCat（美团）","pageTitle":"MineExplorer: إطلاق أول معيار تقييم لمهام طويلة المدى بالدقائق في \"ماين كرافت\" - Aioga أخبار الذكاء الاصطناعي","description":"أصدرت فريق مي توان LongCat أداة MineExplorer، وهي أول معيار تقييم للمهام طويلة المدى بالدقائق في العالم المفتوح للعبة \"ماين كرافت\"، ويحتوي على 813 مثالًا تم التحقق منه يدويًا. عند...","url":"https://www.aioga.com/ar/news/cmryewabt00mnrolg1nxcjf18/","contentTranslated":true,"sourceHash":"a7096d6bbdcff7dd","translatedAt":"2026-07-26T02:46:37.986Z"},"hi":{"title":"MineExplorer: पहला \"Minecraft\" मिनट-स्तरीय लंबी अवधि के मिशन मूल्यांकन बेंचमार्क जारी","summary":"Meituan LongCat टीम ने MineExplorer लॉन्च किया, जो दुनिया में मिनट-स्तरीय लंबी दूरी के कार्यों को मापने वाला पहला माइक्रोसॉफ्ट वर्ल्ड बेंचमार्क है, जिसमें 813 व्यक्तिगत सत्यापित उदाहरण शामिल हैं। 18 शीर्षक बहु-मॉडल बड़े मॉडलों का मूल्यांकन करने पर पता चला कि सबसे शक्तिशाली मॉडल Claude-Opus-4.6 की कुल कार्य सफलता दर केवल 41% थी, 1-कूद कार्य से 77% से घटकर 4-कूद कार्य में 12% हो गई, और लगभग 60% विफलताओं का कारण नेविगेशन में असफलता थी। MineExplorer पूरी तरह से ओपन-सोर्स है।","category":"论文研究","source":"公众号：龙猫LongCat（美团）","aggregationSource":"公众号：龙猫LongCat（美团）","pageTitle":"MineExplorer: पहला \"Minecraft\" मिनट-स्तरीय लंबी अवधि के मिशन मूल्यांकन बेंचमार्क जारी - Aioga AI समाचार","description":"Meituan LongCat टीम ने MineExplorer लॉन्च किया, जो दुनिया में मिनट-स्तरीय लंबी दूरी के कार्यों को मापने वाला पहला माइक्रोसॉफ्ट वर्ल्ड बेंचमार्क है, जिसमें 813 व्यक्तिगत सत्यापित उद...","url":"https://www.aioga.com/hi/news/cmryewabt00mnrolg1nxcjf18/","contentTranslated":true,"sourceHash":"a7096d6bbdcff7dd","translatedAt":"2026-07-26T02:46:42.395Z"},"it":{"title":"MineExplorer: il primo benchmark di valutazione delle missioni a lungo termine a livello di minuti di Minecraft pubblicato","summary":"Il team LongCat di Meituan ha rilasciato MineExplorer, il primo benchmark di valutazione che realizza attività a lungo termine a livello di minuti nel mondo aperto di Minecraft, comprendente 813 casi verificati manualmente. La valutazione di 18 principali modelli multimodali ha rilevato che il modello più potente, Claude-Opus-4.6, ha un tasso di successo complessivo delle attività di solo il 41%, scendendo dal 77% nelle attività con 1 salto al 12% nelle attività con 4 salti, con quasi il 60% dei fallimenti dovuto a errori di navigazione. MineExplorer è stato completamente reso open source.","category":"论文研究","source":"公众号：龙猫LongCat（美团）","aggregationSource":"公众号：龙猫LongCat（美团）","pageTitle":"MineExplorer: il primo benchmark di valutazione delle missioni a lungo termine a livello di minuti di Minecraft pubblicato - Aioga Notizie IA","description":"Il team LongCat di Meituan ha rilasciato MineExplorer, il primo benchmark di valutazione che realizza attività a lungo termine a livello di minuti nel mondo aperto di Minecraft, co...","url":"https://www.aioga.com/it/news/cmryewabt00mnrolg1nxcjf18/","contentTranslated":true,"sourceHash":"a7096d6bbdcff7dd","translatedAt":"2026-07-26T02:47:17.896Z"},"nl":{"title":"MineExplorer: Eerste publicatie van een minutenschaal langetermijnmissiebeoordelingsbenchmark voor 'Minecraft'","summary":"Het Meituan LongCat-team heeft MineExplorer uitgebracht, de eerste evaluatiebenchmark die minutenschaal lange-afstandstaken realiseert in de open wereld van 'Minecraft', met 813 handmatig geverifieerde voorbeelden. Bij het evalueren van 18 toonaangevende multimodale grote modellen bleek dat het beste model, Claude-Opus-4.6, een algehele taakvoltooiingsgraad van slechts 41% had, die daalde van 77% bij éénstapstaken tot 12% bij vierstapstaken, waarbij bijna 60% van de mislukkingen te wijten was aan navigatiefouten. MineExplorer is volledig open source gemaakt.","category":"论文研究","source":"公众号：龙猫LongCat（美团）","aggregationSource":"公众号：龙猫LongCat（美团）","pageTitle":"MineExplorer: Eerste publicatie van een minutenschaal langetermijnmissiebeoordelingsbenchmark voor 'Minecraft' - Aioga AI-nieuws","description":"Het Meituan LongCat-team heeft MineExplorer uitgebracht, de eerste evaluatiebenchmark die minutenschaal lange-afstandstaken realiseert in de open wereld van 'Minecraft', met 813 ha...","url":"https://www.aioga.com/nl/news/cmryewabt00mnrolg1nxcjf18/","contentTranslated":true,"sourceHash":"a7096d6bbdcff7dd","translatedAt":"2026-07-26T02:47:16.128Z"},"tr":{"title":"MineExplorer: İlk 'Minecraft' Dakika Düzeyinde Uzun Süreli Görev Değerlendirme Standardı Yayınlandı","summary":"Meituan LongCat ekibi MineExplorer’ı yayınladı, bu, Minecraft açık dünyasında dakikalık uzun menzilli görevleri gerçekleştiren ilk değerlendirme benchmark’ıdır ve 813 elle doğrulanmış örnek içerir. 18 farklı üst düzey çok modlu büyük modeli değerlendirdiklerinde, en güçlü model olan Claude-Opus-4.6'nın genel görev başarı oranı yalnızca %41, tek adımlı görevden %77’den dört adımlı göreve %12’ye keskin bir düşüş gözlemlendi ve başarısızlıkların yaklaşık %60’ı navigasyon hatalarından kaynaklandı. MineExplorer tamamen açık kaynak olarak sunulmuştur.","category":"论文研究","source":"公众号：龙猫LongCat（美团）","aggregationSource":"公众号：龙猫LongCat（美团）","pageTitle":"MineExplorer: İlk 'Minecraft' Dakika Düzeyinde Uzun Süreli Görev Değerlendirme Standardı Yayınlandı - Aioga AI Haberleri","description":"Meituan LongCat ekibi MineExplorer’ı yayınladı, bu, Minecraft açık dünyasında dakikalık uzun menzilli görevleri gerçekleştiren ilk değerlendirme benchmark’ıdır ve 813 elle doğrulan...","url":"https://www.aioga.com/tr/news/cmryewabt00mnrolg1nxcjf18/","contentTranslated":true,"sourceHash":"a7096d6bbdcff7dd","translatedAt":"2026-07-26T02:47:55.403Z"},"vi":{"title":"MineExplorer: Phát hành tiêu chuẩn đánh giá nhiệm vụ dài hạn theo phút đầu tiên của 《Minecraft》","summary":"Đội ngũ LongCat của Meituan đã phát hành MineExplorer, đây là tiêu chuẩn đánh giá đầu tiên thực hiện nhiệm vụ dài hạn theo phút trong thế giới mở của Minecraft, bao gồm 813 trường hợp xác thực thủ công. Đánh giá 18 mô hình lớn đa phương diện hàng đầu cho thấy, mô hình mạnh nhất Claude-Opus-4.6 có tỷ lệ thành công tổng thể chỉ 41%, từ 77% cho nhiệm vụ 1 bước giảm xuống 12% cho nhiệm vụ 4 bước, gần 60% thất bại do lỗi điều hướng. MineExplorer đã được mở mã hoàn toàn.","category":"论文研究","source":"公众号：龙猫LongCat（美团）","aggregationSource":"公众号：龙猫LongCat（美团）","pageTitle":"MineExplorer: Phát hành tiêu chuẩn đánh giá nhiệm vụ dài hạn theo phút đầu tiên của 《Minecraft》 - Tin tức AI Aioga","description":"Đội ngũ LongCat của Meituan đã phát hành MineExplorer, đây là tiêu chuẩn đánh giá đầu tiên thực hiện nhiệm vụ dài hạn theo phút trong thế giới mở của Minecraft, bao gồm 813 trường...","url":"https://www.aioga.com/vi/news/cmryewabt00mnrolg1nxcjf18/","contentTranslated":true,"sourceHash":"a7096d6bbdcff7dd","translatedAt":"2026-07-26T02:47:53.080Z"},"id":{"title":"MineExplorer: Tolok Ukur Penilaian Misi Jangka Panjang Level Menit Pertama di 'Minecraft' Dirilis","summary":"Tim Meituan LongCat merilis MineExplorer, ini adalah tolok ukur evaluasi pertama untuk tugas jarak jauh menit-demi-menit di dunia terbuka 《Minecraft》, mencakup 813 contoh yang diverifikasi secara manual. Evaluasi terhadap 18 model besar multimodal top menemukan bahwa model terkuat Claude-Opus-4.6 hanya memiliki tingkat keberhasilan tugas keseluruhan 41%, menurun tajam dari 77% pada tugas 1-lompat menjadi 12% pada tugas 4-lompat, hampir 60% kegagalan disebabkan oleh kegagalan navigasi. MineExplorer telah sepenuhnya bersifat open source.","category":"论文研究","source":"公众号：龙猫LongCat（美团）","aggregationSource":"公众号：龙猫LongCat（美团）","pageTitle":"MineExplorer: Tolok Ukur Penilaian Misi Jangka Panjang Level Menit Pertama di 'Minecraft' Dirilis - Berita AI Aioga","description":"Tim Meituan LongCat merilis MineExplorer, ini adalah tolok ukur evaluasi pertama untuk tugas jarak jauh menit-demi-menit di dunia terbuka 《Minecraft》, mencakup 813 contoh yang dive...","url":"https://www.aioga.com/id/news/cmryewabt00mnrolg1nxcjf18/","contentTranslated":true,"sourceHash":"a7096d6bbdcff7dd","translatedAt":"2026-07-26T02:48:29.014Z"},"th":{"title":"MineExplorer: การเปิดตัวเกณฑ์การประเมินภารกิจระยะยาวระดับนาทีชุดแรกของ 'Minecraft'","summary":"ทีม Meituan LongCat ได้เปิดตัว MineExplorer ซึ่งเป็นเกณฑ์การประเมินครั้งแรกที่สามารถทำภารกิจระยะยาวระดับนาทีในโลกเปิดของ 'Minecraft' รวมถึงตัวอย่างที่ได้รับการยืนยันด้วยตนเอง 813 รายการ การประเมิน 18 โมเดลขนาดใหญ่มัลติโมดัลระดับแนวหน้าพบว่า โมเดลที่แข็งแกร่งที่สุด Claude-Opus-4.6 มีอัตราความสำเร็จของภารกิจโดยรวมเพียง 41% ลดลงจากภารกิจกระโดดครั้งเดียวที่ 77% เป็นภารกิจกระโดด 4 ครั้งที่ 12% เกือบ 60% ของความล้มเหลวมาจากการนำทางไม่สำเร็จ MineExplorer ได้เปิดซอร์สอย่างครบถ้วนแล้ว","category":"论文研究","source":"公众号：龙猫LongCat（美团）","aggregationSource":"公众号：龙猫LongCat（美团）","pageTitle":"MineExplorer: การเปิดตัวเกณฑ์การประเมินภารกิจระยะยาวระดับนาทีชุดแรกของ 'Minecraft' - ข่าว AI Aioga","description":"ทีม Meituan LongCat ได้เปิดตัว MineExplorer ซึ่งเป็นเกณฑ์การประเมินครั้งแรกที่สามารถทำภารกิจระยะยาวระดับนาทีในโลกเปิดของ 'Minecraft' รวมถึงตัวอย่างที่ได้รับการยืนยันด้วยตนเอง 813 ร...","url":"https://www.aioga.com/th/news/cmryewabt00mnrolg1nxcjf18/","contentTranslated":true,"sourceHash":"a7096d6bbdcff7dd","translatedAt":"2026-07-26T02:48:40.641Z"},"pl":{"title":"MineExplorer: Opublikowano pierwszy benchmark oceniający długoterminowe zadania minutowe w „Minecraft”","summary":"Zespół Meituan LongCat opublikował MineExplorer, który jest pierwszym w świecie otwartym gry „Minecraft” benchmarkiem zadań długoterminowych realizowanych w minutach, zawierającym 813 ręcznie zweryfikowanych przypadków. Ocena 18 najlepszych wielomodalnych dużych modeli wykazała, że najsilniejszy model Claude-Opus-4.6 osiągnął ogólny wskaźnik sukcesu zadań zaledwie 41%, spadając gwałtownie z 77% w zadaniach skoków 1 do 12% w zadaniach skoków 4, a prawie 60% niepowodzeń wynikało z błędów nawigacji. MineExplorer został w pełni udostępniony jako open source.","category":"论文研究","source":"公众号：龙猫LongCat（美团）","aggregationSource":"公众号：龙猫LongCat（美团）","pageTitle":"MineExplorer: Opublikowano pierwszy benchmark oceniający długoterminowe zadania minutowe w „Minecraft” - Aioga Wiadomości AI","description":"Zespół Meituan LongCat opublikował MineExplorer, który jest pierwszym w świecie otwartym gry „Minecraft” benchmarkiem zadań długoterminowych realizowanych w minutach, zawierającym...","url":"https://www.aioga.com/pl/news/cmryewabt00mnrolg1nxcjf18/","contentTranslated":true,"sourceHash":"a7096d6bbdcff7dd","translatedAt":"2026-07-26T02:49:21.422Z"}}}}