AI Paper Atlas

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

Pix2Key: Controllable Open-Vocabulary Retrieval with Semantic Decomposition and Self-Supervised Visual Dictionary Learning

Guoyizhe Wei, Yang Jiao, Nan Xi, Zhishen Huang, Jingjing Meng, Rama Chellappa, Yan Gao

ICML 2026 · Vision-language and multimodal › Human and embodied modeling · approach: Composed and text-based image retrieval · 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 negated constraints to the query dictionary (ablation)Pix2Key+V-Dict-AE · DFMM-ComposeRecall@50: 37.27 -> 39.41Improved
add open-set anchors together with negation (ablation)Pix2Key+V-Dict-AE · DFMM-ComposeRecall@50: 39.41 -> 40.48Improved
add open-set anchors without negation (ablation)Pix2Key · DFMM-ComposeAC@50: 53.31 -> 51.72Hurt
add MMR diversity reranking (ablation)Pix2Key+V-Dict-AE · DFMM-ComposeILD@50: 49.22 -> 53.96Improved
add MMR diversity reranking (ablation)Pix2Key+V-Dict-AE · DFMM-ComposeRecall@50: 40.48 -> 40.96Improved
add MMR reranking to CIReVLCIReVL · CIRRRecall@50: 86.95 -> 86.45Hurt
add V-Dict-AE self-supervised visual dictionary pretrainingPix2Key · CIRRRecall@5: 54.26 -> 59.44Improved
add V-Dict-AE pretrainingPix2Key · FashionIQ (average)Recall@10: 29.18 -> 29.96Improved
add LoRA (ablation)Pix2Key+V-Dict-AE · CIRRRecall@5: 57.20 -> 59.44Improved
use refined prompt (ablation)Pix2Key+V-Dict-AE · CIRRRecall@5: 58.27 -> 59.44Improved
increase input resolution from 64 to 224 (ablation)Pix2Key+V-Dict-AE · CIRRRecall@5: 58.01 -> 59.44Improved
pooler depth 9 instead of 5 (ablation)Pix2Key+V-Dict-AE · CIRRRecall@5: 59.44 -> 56.90Hurt

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