智能体通过智能选择恢复历史截图，提升长任务成功率。
CausalCache: Conditional High-Fidelity Restoration for Long-Horizon GUI Agents
- 按预测价值动态选择恢复旧截图，而非只保留最新画面。
- 在30步任务中达46.7%成功率，比固定最近4帧高4.3个百分点。
- 跨应用任务提升显著，适合长流程、多窗口操作场景。
长时序GUI智能体可将完整操作历史压缩为文本，但仅少数历史截图能保留在当前上下文中。本文提出预算化保真度恢复问题：每个事件均被摘要保留，固定预算 $B$ 决定哪些事件可恢复归档截图。近期策略 $\text{Recent-}B$ 将所有视觉资源用于最新事件。而 CausalCache 评估完整历史，仅当旧事件预测效用高于近期事件时才替换。一个历史门控的键值适配器仅修改恢复的历史图像标记，在无激活历史图像时完全绕过，确保当前屏幕处理不受影响。该适配器与选择器在桌面轨迹上进行匹配预算干预训练，并在移动端零样本评估。在 OSWorld-Verified 上，激活历史截图使成功率相比仅摘要记忆提升约13个百分点。在官方15步限制下，CausalCache 与 Recent-$4$ 无统计差异；在30步诊断任务中，成功率达46.7%（对比42.4%，+4.3点）。在117个MobileWorld任务上零样本测试，成功率从30.2%提升至36.8%。增益集中在预定义的跨应用记忆候选集（30.6% vs. 19.4%，+11.2点），单应用控制无明显差异（43.6% vs. 42.4%）。结果表明，选择性恢复过去事件的视觉信息比固定优先近期更有效。
原文摘要 · Abstract (English)
Long-horizon GUI agents can retain complete action histories as compact text, but only a few historical screenshots fit in active context. We formulate this as budgeted fidelity restoration: every event remains summarized, while a fixed budget $B$ determines which events regain their archived screenshots. Recent-$B$ assigns all visual slots to the latest events. CausalCache instead scores the complete history and swaps in an older event only when its predicted utility exceeds that of a recent event. A history-gated key/value adapter modifies only restored history-image tokens and is exactly bypassed when no history image is active, preserving current-screen processing. The adapter and selector are trained with matched-budget interventions on desktop trajectories and evaluated zero-shot on mobile. On OSWorld-Verified, activating historical screenshots improves success by about $13$ percentage points over summary-only memory. Under the official $15$-step limit, CausalCache and Recent-$4$ are statistically indistinguishable; in a $30$-step diagnostic, CausalCache achieves $46.7\%$ success versus $42.4\%$ ($+4.3$ points). Zero-shot on $117$ MobileWorld tasks, CausalCache improves over Recent-$4$ from $30.2\%$ to $36.8\%$. The gain is concentrated on a pre-defined cross-app memory-candidate split ($30.6\%$ vs. $19.4\%$, $+11.2$ points), while single-app controls show no detectable difference ($43.6\%$ vs. $42.4\%$). These results show that selecting which past events regain pixels is more effective than spending a fixed visual budget entirely on recency.
Thank you to arXiv for use of its open access interoperability. PaperDance 不是 arXiv 官方产品；中文卡片由大模型生成，请以原文为准。