AI Paper Atlas

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

RetFormer: Multimodal Retrieval for Enhancing Image Recognition

Tianrui Yu, Xiubo Liang, Hongzhi Wang

CVPR 2026 · Vision-language and multimodal › Human and embodied modeling · approach: Composed and text-based image retrieval

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
retrieve both text and image instead of image only (ablation)RetFormer · ImageNet-LTOverall top-1 accuracy: 76.0 -> 81.9Improved
retrieve both text and image instead of text only (ablation)RetFormer · ImageNet-LTOverall top-1 accuracy: 78.1 -> 81.9Improved
add retrieval module to CLIP classifier fine-tuningCLIP ViT-B/16 · ImageNet-LTFew-shot top-1 accuracy: 52.9 (full FT) -> 76.8Improved
fine-tune the image encoder instead of freezing it (ablation)RetFormer · ImageNet-LTOverall top-1 accuracy: 78.6 -> 81.9Improved
increase cross-fusion layers from L=1 to L=4 (ablation)RetFormer · ImageNet-LTOverall top-1 accuracy: 81.2 -> 81.9Improved
increase cross-fusion layers from L=4 to L=8 (ablation)RetFormer · ImageNet-LTOverall top-1 accuracy: 81.9 -> 82.0No clear effect
increase retrieved neighbors k up to 32 (ablation)RetFormer · ImageNet-LTaccuracyImproved
increase k further (ablation)RetFormer · ImageNet-LTMany-shot accuracyHurt
enlarge the knowledge base (downstream, DataComp, All) (ablation)RetFormer · ImageNet-LTaccuracyImproved
use approximate HNSW index (recall 0.97) instead of exact k-NN (ablation)RetFormer · ImageNet-LTOverall top-1 accuracy: 81.9 -> 81.8 (23.34 ms -> 7.69 ms)No clear effect
lower HNSW recall to 0.65 (ablation)RetFormer · ImageNet-LTOverall top-1 accuracy: 81.9 -> 78.7Hurt
use smaller k (8, 16) on Places-LT (ablation)RetFormer · Places-LTMany-shot accuracy: Many +0.3 to 0.8%, Few dippedMixed

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