ReVision removes visually redundant patches across consecutive screenshots, reducing token usage and enabling models to handle longer histories, leading to improved performance in computer-use agents.</p>\n","updatedAt":"2026-06-11T18:56:17.962Z","author":{"_id":"60e32baedc56466240084155","avatarUrl":"https://cdn-avatars.huggingface.co/v1/production/uploads/60e32baedc56466240084155/Q-cJh3Q3-vvMbe749Gt5B.jpeg","fullname":"Amirhossein Abaskohi","name":"AmirhosseinAbaskohi","type":"user","isPro":false,"isHf":false,"isHfAdmin":false,"isMod":false,"followerCount":3,"isUserFollowing":false}},"numEdits":0,"identifiedLanguage":{"language":"en","probability":0.8902086615562439},"editors":["AmirhosseinAbaskohi"],"editorAvatarUrls":["https://cdn-avatars.huggingface.co/v1/production/uploads/60e32baedc56466240084155/Q-cJh3Q3-vvMbe749Gt5B.jpeg"],"reactions":[],"isReport":false}}],"primaryEmailConfirmed":false,"paper":{"id":"2605.11212","authors":[{"_id":"6a27c1f46dde1c5ef75bd22b","name":"Amirhossein Abaskohi","hidden":false},{"_id":"6a27c1f46dde1c5ef75bd22c","name":"Yuhang He","hidden":false},{"_id":"6a27c1f46dde1c5ef75bd22d","name":"Peter West","hidden":false},{"_id":"6a27c1f46dde1c5ef75bd22e","name":"Giuseppe Carenini","hidden":false},{"_id":"6a27c1f46dde1c5ef75bd22f","name":"Pranit Chawla","hidden":false},{"_id":"6a27c1f46dde1c5ef75bd230","name":"Vibhav Vineet","hidden":false}],"publishedAt":"2026-06-05T00:00:00.000Z","submittedOnDailyAt":"2026-06-11T00:00:00.000Z","title":"ReVision: Scaling Computer-Use Agents via Temporal Visual Redundancy Reduction","submittedOnDailyBy":{"_id":"60e32baedc56466240084155","avatarUrl":"https://cdn-avatars.huggingface.co/v1/production/uploads/60e32baedc56466240084155/Q-cJh3Q3-vvMbe749Gt5B.jpeg","isPro":false,"fullname":"Amirhossein Abaskohi","user":"AmirhosseinAbaskohi","type":"user","name":"AmirhosseinAbaskohi"},"summary":"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.","upvotes":3,"discussionId":"6a27c1f46dde1c5ef75bd231","ai_summary":"ReVision improves computer-use agent efficiency by removing redundant visual patches from consecutive screenshots while preserving spatial structure, reducing token usage by 46% and improving success rates.","ai_keywords":["computer-use agents","visual tokens","multimodal language models","patch selector","visual patches","consecutive screenshots","spatial structure","token usage","success rate"],"ai_summary_model":"Qwen/Qwen2.5-Coder-32B-Instruct","organization":{"_id":"68151d0f51add3813f3f7d1b","name":"MicrosoftResearch","fullname":"Microsoft Research","avatar":"https://cdn-avatars.huggingface.co/v1/production/uploads/6529a4f2f1205983224fa513/PeuVr7jSuJflmDBBGxoDX.png"}},"canReadDatabase":false,"canManagePapers":false,"canSubmit":false,"hasHfLevelAccess":false,"upvoted":false,"upvoters":[{"_id":"60e32baedc56466240084155","avatarUrl":"https://cdn-avatars.huggingface.co/v1/production/uploads/60e32baedc56466240084155/Q-cJh3Q3-vvMbe749Gt5B.jpeg","isPro":false,"fullname":"Amirhossein Abaskohi","user":"AmirhosseinAbaskohi","type":"user"},{"_id":"63c8527becdb7c9fdd9cacc6","avatarUrl":"/avatars/c8a3f5e1e5159ae5ead41bd9fc2b9b34.svg","isPro":false,"fullname":"Vibhav Vineet","user":"vibhav-vineet","type":"user"},{"_id":"6228ede94323cef93a956b24","avatarUrl":"/avatars/2f01099f102889f2a621a68dcd61b6b6.svg","isPro":false,"fullname":"AmirHossein DabiriAghdam","user":"AmirHossein1378","type":"user"}],"acceptLanguages":["en"],"dailyPaperRank":0,"organization":{"_id":"68151d0f51add3813f3f7d1b","name":"MicrosoftResearch","fullname":"Microsoft Research","avatar":"https://cdn-avatars.huggingface.co/v1/production/uploads/6529a4f2f1205983224fa513/PeuVr7jSuJflmDBBGxoDX.png"},"markdownContentUrl":"https://huggingface.co/buckets/huggingchat/papers-content/resolve/2605/2605.11212.md"}">

Get this paper in your agent:

Don't have the latest CLI?

Cite arxiv.org/abs/2605.11212 in a model README.md to link it from this page.

Cite arxiv.org/abs/2605.11212 in a dataset README.md to link it from this page.

Cite arxiv.org/abs/2605.11212 in a Space README.md to link it from this page.

Add this paper to a collection to link it from this page.