AI Paper Atlas

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

Synergistic Space-Vision Processing for Predicate Inference

Zhenhua Lei, Zefang Han, yu qiu

ICML 2026 · Vision-language and multimodal › Human and embodied modeling · approach: Scene graph generation

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 vision stream and cross-stream fusion to the space stream alone (ablation)DS-Net · VG PredClsmR@50/100: 32.5/34.7 -> 40.2/41.1Improved
add space stream and fusion to the vision stream alone (ablation)DS-Net · VG PredClsmR@50/100: 34.9/36.1 -> 40.2/41.1Improved
use proposed space stream instead of common space stream (ablation)DS-Net · VG PredClsmR@50/100: 38.1/39.3 -> 40.2/41.1Improved
use proposed vision stream instead of common vision stream (ablation)DS-Net · VG PredClsmR@50/100: 36.2/38.4 -> 40.2/41.1Improved
replace concatenation with cross-stream fusion (ablation)DS-Net · VG PredClsmR@50/100: 37.5/39.1 -> 40.2/41.1Improved
add space-to-vision fusion to vision-to-space fusion only (ablation)DS-Net · VG PredClsmR@50/100: 38.4/40.3 -> 40.2/41.1Improved
add vision-to-space fusion to space-to-vision fusion only (ablation)DS-Net · VG PredClsmR@50/100: 39.1/40.9 -> 40.2/41.1Improved
increase latent parts from 4 to 8 (ablation)DS-Net · VG PredClsmR@50/100: 37.4/38.9 -> 40.2/41.1Improved
increase latent parts from 8 to 14 (ablation)DS-Net · VG PredClsmR@50/100: 40.2/41.1 -> 39.9/41.0No clear effect
plug DS-Net into PENETPENET · VG SGGenmR@100: 14.5 -> 19.6Improved
plug DS-Net into RepSGGRepSGG · VG SGGenmR@100: 18.9 -> 21.7Improved
plug DS-Net into RA-SGGRA-SGG · VG SGGenmR@100: 17.1 -> 20.7Improved

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