AI Paper Atlas

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

ObjEmbed: Towards Universal Multimodal Object Embeddings

Shenghao Fu, Yukun Su, Fengyun Rao, Jing LYU, Xiaohua Xie, Wei-Shi Zheng

ICML 2026 · Vision-language and multimodal › Human and embodied modeling · approach: Language-grounded scene text and structure · 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
use box IoU as the focal-loss label instead of binary labels (single token) (ablation)ObjEmbed · COCOAP: 37.1 -> 42.3Improved
decouple an object into separate class and IoU tokens (ablation)ObjEmbed · COCOAP: 42.3 -> 45.5Improved
place the class token before the IoU token (ablation)ObjEmbed two-token design · COCOAP: 45.1 -> 45.5No clear effect
add object instruction (Object i: <object><iou>) (ablation)ObjEmbed · COCOAP: 42.1 -> 45.5Improved
add task instructions on top of object instruction (ablation)ObjEmbed · COCOAP: 45.5 -> 47.1Improved
add image-level objective to object-level training (ablation)ObjEmbed · local image retrieval (LIR*)average score: 60.2 -> 62.9Improved
add object-level objective to image-level training (ablation)ObjEmbed · global image retrieval (GIR)average score: 81.2 -> 81.6Improved
use separate special tokens for global and local text embeddings (ablation)ObjEmbed · COCO and RefCOCOAP / accuracy: COCO +0.3 AP, RefCOCO +0.8Improved
supervise global image embedding with both short and long captions, with two global image tokens (ablation)ObjEmbed · global image retrieval (GIR)average score: 80.1 -> 81.6Improved
add box offset regression head on IoU embeddings (ablation)ObjEmbed · COCO / RefCOCOAP / accuracy: COCO 52.8 -> 52.5, RefCOCO 87.4 -> 86.2Hurt
increase number of proposals from 50 to 100 / 150 (ablation)ObjEmbed · COCOAP: 52.6 -> 53.0 (100), 51.9 (150)No clear effect
mix ground-truth boxes into proposals (oracle) (ablation)ObjEmbed-4B · COCOAP: 53.0 -> 65.2Improved

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