arXiv cs.CL· Amirhossein Abaskohi, Yuhang He, Peter West, Giuseppe Carenini, Pranit Chawla, Vibhav Vineet·· 15 小时前AI 评分38
ReVision：通过时序视觉冗余削减扩展计算机使用智能体
ReVision: Scaling Computer-Use Agents via Temporal Visual Redundancy Reduction
AI 导读
ReVision 通过可学习的 patch 选择器比较连续截图的 patch 表示并去除冗余视觉 patch，在保留空间结构的前提下训练多模态语言模型。在 OSWorld、WebTailBench 和 AgentNetBench 三个基准上，用 Qwen2.5-VL-7B 处理含 5 张历史截图的轨迹时，token 用量平均减少 46%，成功率较不裁剪基线提升 3%。
正文 · 原文
Abstract:Computer-use agents (CUAs) rely on visual observations of graphical user interfaces, where each screenshot is encoded into a large number of visual tokens. As interaction trajectories grow, the token cost increases rapidly, limiting the amount of history that can be incorporated under fixed context and compute budgets. This has resulted in no or very limited improvement in the performance when using history unlike other domains. We address this inefficiency by introducing ReVision, which is used to train multimodal language models on trajectories where redundant visual patches are removed using a learned patch selector that compares patch representations across consecutive screenshots while preserving spatial structure required by the model. Across three benchmarks, OSWorld, WebTailBench, and AgentNetBench, when processing trajectories with 5 history screenshots using Qwen2.5-VL-7B, ReVision reduces token usage by 46% on average while improving success rate by 3% over the no drop baseline. This establishes a clear efficiency gain, enabling agents to process longer trajectories with fewer tokens. With this improved efficiency, we revisit the role of history in CUAs and find that performance continues to improve as more past observations are incorporated when redundancy is removed.
| Subjects: | Computation and Language (cs.CL) |
| Cite as: | arXiv:2605.11212 [cs.CL] |
| (or arXiv:2605.11212v4 [cs.CL] for this version) | |
| https://doi.org/10.48550/arXiv.2605.11212 arXiv-issued DOI via DataCite |
Submission history
From: Amirhossein Abaskohi [view email]
[v1]
Mon, 11 May 2026 20:27:54 UTC (6,757 KB)
[v2]
Wed, 13 May 2026 16:34:05 UTC (6,757 KB)
[v3]
Fri, 5 Jun 2026 22:18:46 UTC (6,753 KB)
[v4]
Sun, 25 Oct 2026 06:06:01 UTC (6,754 KB)
来源：arXiv cs.CL · arxiv.org