AI Paper Atlas

Research map / Vision-language and multimodal / Human and embodied modeling

ScreenParse: Moving Beyond Sparse Grounding with Complete Screen Parsing Supervision

A. Said Gurbuz, Sunghwan Hong, Ahmed Nassar, Marc Pollefeys, Peter W. J. Staar

ICML 2026 · Vision-language and multimodal › Human and embodied modeling · approach: Layout-aware document understanding · 0 citations (OpenAlex)

Read the paperCheck your idea against 137,217 papers

AI-extracted. The measured effects below were read by AI from the paper's full text (no summary yet). Numbers and claims can be wrong or missing context, so check the paper before you cite or build on it.

Measured effects reported in the paper

What was changedWhereMeasuredEffect
add structure-aware weighted loss (upweight tag and location tokens) (ablation)ScreenVLM · ScreenParse testPage IoU: 0.592 -> 0.606Improved
add structure-aware weighted loss (ablation)ScreenVLM · ScreenSpot PCRecall@50: 0.129 -> 0.222Improved
add structure-aware weighted loss (ablation)ScreenVLM · GroundCUAPage IoU: 0.226 -> 0.251Improved
fine-tune on ScreenParseQwen3-VL-2B-Instruct · ScreenParse testPage IoU: 0.228 -> 0.585Improved
fine-tune on ScreenParseInternVL3-2B · ScreenParse testPage IoU: 0.111 -> 0.509Improved
fine-tune on ScreenParseInternVL3-2B · GroundCUAPage IoU: 0.025 -> 0.203Improved
fine-tune on ScreenParseQwen3-VL-2B-Instruct · ScreenSpot MobileRecall@50: 0.143 -> 0.108Hurt
fine-tune on ScreenParseQwen3-VL-2B-Instruct · ScreenSpot WebPixCov: 0.292 -> 0.720Improved
fine-tune OmniParser V2 on ScreenParseOmniParser V2 · ScreenParse testPage IoU: 0.270 -> 0.503Improved
fine-tune OmniParser V2 on ScreenParseOmniParser V2 · GroundCUAPage IoU: 0.361 -> 0.398Improved
swap OmniParser v2 detector for ScreenParse-trained YOLOSet-of-Mark pipeline · Mind2Web (task)step success: 26.8 -> 27.0No clear effect
swap OmniParser v2 detector for ScreenParse-trained YOLOSet-of-Mark pipeline · OSWorld-G fine-grained manipulationaccuracy: 25.50 -> 29.53Improved

Earlier methods this paper compares with

Named baselines from earlier accepted papers in this paper's results tables: better than 1 of 1 on most shared results.

Matched by method name in results tables read by AI. Authors' own tables tend to favour the newer method, and a comparison can be missed when a method is renamed.

Closest papers by meaning

Writing a paper in this area? scime draft is a free LaTeX editor with AI included, real-time co-authors and a check of your references and venue rules before you submit. It also imports OpenAI Prism projects.

Open scime draft

Get the next scime Dispatch and a monthly field update

Free. We send a confirmation link first; every email has a one-click unsubscribe. Privacy