AI Paper Atlas

Research map / Video / Video-language understanding

Imagine Before Concentration: Diffusion-Guided Registers Enhance Partially Relevant Video Retrieval

Jun Li, Xuhang Lou, Jinpeng Wang, Yuting Wang, Yaowei Wang, Shu-Tao Xia, Bin Chen

CVPR 2026 · Video › Video-language understanding · approach: Text-video retrieval · 2 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 diffusion-generated global registers (ablation)DreamPRVR · ActivityNet CaptionsSumR: 153.4 -> 156.1Improved
add diffusion-generated global registers (ablation)DreamPRVR · TVRSumR: 187.0 -> 193.1Improved
use diffusion registers instead of adaptive pooling (ablation)DreamPRVR · Charades-STASumR: 78.1 -> 80.0Improved
add diffusion refinement (DRE) instead of one-step mapping (ablation)DreamPRVR · ActivityNet CaptionsSumR: 150.6 -> 156.1Improved
initialize registers from PVS video-centric distribution rather than random noise (ablation)DreamPRVR · Charades-STASumR: 77.6 -> 80.0Improved
initialize registers from PVS (ablation)DreamPRVR · ActivityNet CaptionsSumR: 154.9 -> 156.1No clear effect
add Lpvs KL loss (ablation)DreamPRVR · ActivityNet CaptionsSumR: 155.1 -> 156.1Improved
add Ldre textual supervision (ablation)DreamPRVR · Charades-STASumR: 78.5 -> 80.0Improved
add textual semantic structure learning (QSP + diversity) (ablation)DreamPRVR · Charades-STASumR: 76.9 -> 80.0Improved
add all proposed losses over Lsim only (ablation)DreamPRVR · TVRSumR: 187.0 -> 193.1Improved
increase register count above 8 (ablation)DreamPRVR · ActivityNet CaptionsSumRHurt
increase diffusion timesteps beyond 10 (ablation)DreamPRVR · ActivityNet CaptionsSumRHurt

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