研究人员开发了Mobile-Aptus，一个置信度驱动的框架，旨在提高基于多模态大语言模型（MLLM）的移动代理的交互能力。该框架通过使代理能够输出置信度分数以及动作，然后使用语义相似性和直接偏好优化来纠正这些分数，从而解决了过度执行和过度请求的问题。Mobile-Aptus在四个基准测试中展示了最先进的性能，在任务成功率方面取得了显著的提高，并减少了对人工干预的需求。 AI
影响 增强了在移动设备上运行的AI代理的可靠性和效率，减少了不必要的人工干预。
排序理由 该集群包含一篇详细介绍MLLM代理新框架的研究论文。
AI 生成摘要 · Google Gemini · 来自 2 个来源。 我们如何撰写摘要 →
arXiv:2605.28629v1 Announce Type: new Abstract: Recent advancements in multimodal large language models (MLLMs) have shown exceptional potential in enabling mobile-using agents to autonomously execute human instructions. However, fully automated agents often try to execute tasks …
Recent advancements in multimodal large language models (MLLMs) have shown exceptional potential in enabling mobile-using agents to autonomously execute human instructions. However, fully automated agents often try to execute tasks even when they are unable to resolve them, leadi…