AI Paper Atlas

Research map / Vision-language and multimodal / Multimodal LLMs

DeCoPatch: Revealing Causal Latent Subspaces in Vision-Language Models for GUI Grounding

Yongkang Zhang, Linjia Kang, Zhimin Wang, Duo Wu, Zhi Wang

ECCV 2026 · Vision-language and multimodal › Multimodal LLMs · approach: Bias-aware multi-image robustness · 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
patch top-k selected late-layer FFN neurons at decode timeJedi-7B-1080p · OSWorld-G overallaccuracy (%): 53.19 -> 54.26Improved
patch selected neurons instead of random neurons (ablation)Qwen2.5-VL-3B · OSWorld-G overallaccuracy (%): 26.95 -> 29.79Improved
patch top-k neuronsInternVL2.5-7B · ScreenSpot-v2 avgaccuracy (%): 79.32 -> 84.91Improved
patch top-k neuronsPixtral-12B · ScreenSpot-v2 avgaccuracy (%): 36.95 -> 46.23Improved
patch top-k neuronsUI-Venus-Ground-7B · ScreenSpot-v2 avgaccuracy (%): 93.87 -> 94.50Improved
patch top-k neurons versus overlaying SoM markersJedi-7B-1080p · ScreenSpot-Pro avgaccuracy (%): base 39.29; SoM 40.76; ours 40.09Mixed
patch top-k neuronsQwen3-VL-8B · ScreenSpot-Pro avgaccuracy (%): 46.50 -> 48.96Improved
patch top-k neuronsQwen3-VL-8B · AndroidControl-Curated-Hard SRsuccess rate (%): 31.57 -> 33.96Improved
patch top-k neurons versus SoMQwen2.5-VL-7B · AndroidControl-Curated-Hard SRsuccess rate (%): SoM 31.86 vs ours 31.50; base 29.69No clear effect
replace selected neurons with random neurons (ablation)DeCoPatch · ScreenSpot-v2 avgaccuracy (%): InternVL2.5-7B 84.91 -> 77.52Hurt
suppress (divide) the selected neurons instead of boosting them (ablation)DeCoPatch · OSWorld-G (hyperparameter sweep)accuracy (%)Hurt
vary k, alpha and number of late layers (ablation)DeCoPatch · OSWorld-Gaccuracy (%)Improved

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