微软研究人员发现，当前 AI 模型和代理存在一个重大局限性：它们无法有效管理长期任务。这些系统在需要长时间持续运行或保持记忆的任务中表现不佳。这种缺陷影响了它们执行复杂、多阶段操作的潜力，并指出了未来 AI 发展的一个方向。 AI
影响 凸显了当前 AI 能力的局限性，表明当前模型和代理尚无法执行复杂的长期操作。
排序理由 微软研究人员的研究论文，指出了 AI 模型的局限性。
AI 生成摘要 · Google Gemini · 来自 3 个来源。 我们如何撰写摘要 →
微软研究人员发现 AI 模型难以处理长期任务
报道来源 [3]
-
Microsoft researchers find AI models and agents can't handle long-running tasks
An intern who failed this much would be shown the door
-
Microsoft Research warns AI agents still struggle with long autonomous workflows. Researchers found frontier models corrupted documents, lost critical context,
Microsoft Research warns AI agents still struggle with long autonomous workflows. Researchers found frontier models corrupted documents, lost critical context, and degraded content quality over multi-step tasks - with agentic tooling often worsening outcomes. Human oversight stil…
-
Microsoft researchers find AI models and agents can't handle long-running tasks https://www. theregister.com/ai-ml/2026/05/ 11/microsoft-researchers-find-ai-mod
Microsoft researchers find AI models and agents can't handle long-running tasks https://www. theregister.com/ai-ml/2026/05/ 11/microsoft-researchers-find-ai-models-and-agents-cant-handle-long-running-tasks/5238263 # ai