研究人员开发了 GUI-CIDER，一种新颖的中间训练方法，旨在增强使用多模态大型语言模型构建的 GUI 代理的世界知识。该方法通过因果内化和密度感知示例重选明确内化 GUI 操作知识，解决了传统训练后方法的局限性。GUI-CIDER 合成数据，通过优先考虑因果结构和减少冗余来精炼数据，然后使用这些精炼的数据进行中间训练。实验表明，使用此方法训练的代理在 GUI 理解和任务成功率方面有了显著提高。 AI
影响 该方法可能带来更强大、更可靠的 GUI 代理，从而改善用户与软件的交互。
排序理由 该集群包含一篇详细介绍 AI 代理新训练方法的论文。
在 Hugging Face Daily Papers 阅读 →
AI 生成摘要 · Google Gemini · 来自 5 个来源。 我们如何撰写摘要 →
arXiv:2605.28534v1 Announce Type: new Abstract: Despite the rapid progress of multimodal large language models in building Graphical User Interface (GUI) agents, their real-world task completion is fundamentally bottlenecked by a lack of world knowledge about GUI operations. Exis…
Despite the rapid progress of multimodal large language models in building Graphical User Interface (GUI) agents, their real-world task completion is fundamentally bottlenecked by a lack of world knowledge about GUI operations. Existing solutions typically rely on expensive multi…
Despite the rapid progress of multimodal large language models in building Graphical User Interface (GUI) agents, their real-world task completion is fundamentally bottlenecked by a lack of world knowledge about GUI operations. Existing solutions typically rely on expensive multi…
GUI-CIDER is a mid-training method that explicitly incorporates GUI world knowledge through causal internalization and density-aware exemplar reselection to improve GUI agent performance.
arXiv:2605.30884v1 Announce Type: new Abstract: Existing agentic reinforcement learning methods for GUI grounding have limitations at two levels. At the data level, current approaches typically treat all training samples equally, although their training value to the baseline mode…