AI Paper Atlas

Research map / Vision-language and multimodal / Cross-modal learning

Efficient and High-Fidelity Omni Modality Retrieval

Chuong Huynh, Manh Luong, Abhinav Shrivastava

CVPR 2026 · Vision-language and multimodal › Cross-modal learning · approach: Probabilistic embeddings for cross-modal 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
use ASWP pooling instead of single [EOS] vector (ablation)OmniRet · ablation avg recallaverage recall: 43.4 -> 50.2Improved
use ASWP instead of 16 multi-vector late interaction (ablation)OmniRet · ablation avg recallaverage recall: 49.8 -> 50.2No clear effect
use STM hard selection instead of average pooling in ASWP (ablation)OmniRet · ablation avg recallaverage recall: 20.7 -> 50.2Improved
use STM instead of max pooling or weighted sum (ablation)OmniRet · ablation avg recallaverage recall: 49.2 / 48.5 -> 50.2Improved
add the shared media resampler (ablation)OmniRet · ablation avg recallaverage recall: 46.7 -> 50.2Improved
add modality-specific media latents to the shared resampler (ablation)OmniRet · ablation avg recallaverage recall: 49.8 -> 50.2No clear effect
add diversity loss on resampled tokens (ablation)OmniRet · ablation avg recallaverage recall: 47.1 -> 50.2Improved
add triplet loss (ablation)OmniRet · ablation avg recallaverage recall: 49.7 -> 50.2No clear effect
increase projections L from 1536 to 4096 (ablation)OmniRet · ablation avg recallaverage recall: 47.5 -> 50.2Improved
use OmniRet rather than VLM2VecV2 on image classificationOmniRet · MMEBv2 Image CLSRecall@1: 62.9 -> 51.7Hurt
use OmniRet rather than VLM2VecV2 on video tasksOmniRet · MMEBv2 VideoRecall@1: CLS 39.3 -> 48.6; RET 28.8 -> 36.5Improved
use single-stage OmniRet vs caption+Gemma pipelineOmniRet · ACM composed audioRecall@5: 44.6 (pipeline) vs 23.0Hurt

Earlier methods this paper compares with

Named baselines from earlier accepted papers in this paper's results tables: better than 1 of 2 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