AI Paper Atlas

Research map / Vision-language and multimodal / CLIP and zero-shot

An Efficient Post-hoc Framework for Reducing Task Discrepancy of Text Encoders for Composed Image Retrieval

Jaeseok Byun, Seokhyeon Jeong, Wonjae Kim, Sanghyuk Chun, Taesup Moon

ICCV 2025 · Vision-language and multimodal › CLIP and zero-shot · approach: Attribute-based zero-shot learning · 14 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 RTD text-encoder post-hoc training to projection-based CIR methodsPic2Word, SEARLE, LinCIR, Context-I2W, FTI4CIR · CIRR/CIRCO/FashionIQaverage gain: average improvement of over 2.78 pointsImproved
use generated text triplets (T_{r+c},T_t) instead of (T_r,T_r) pairs for contrastive loss (ablation)RTD on LinCIR · CIR val (avg)Avg: 38.5 -> 39.2Improved
add refined batch sampling with (T_r,T_r) hard negatives (ablation)RTD on LinCIR · CIR val (avg)Avg: 39.2 -> 39.6Improved
add refined concatenation scheme (ablation)RTD on LinCIR · CIR val (avg)Avg: 39.6 -> 40.7Improved
anchor target caption embedding with frozen text encoder (ablation)RTD on LinCIR · CIR val (avg)Avg: 36.3 -> 39.6Improved
naively fine-tune text encoder with original objective instead of RTD (ablation)Pic2Word · CIR val (avg)Avg: 32.2 -> 10.9Hurt
naively fine-tune text encoder with original objective (ablation)LinCIR · CIR val (avg)Avg: 36.9 -> 35.5Hurt
replace Compodiff LLM triplets with rule-based text triplets (ablation)RTD on LinCIR · CIR val (avg)Avg: +2.9 versus +3.3 over LinCIR 37.4No clear effect
use CASE text triplets (ablation)RTD on LinCIR · CIR val (avg)Avg: +0.3No clear effect
use CoVR text triplets (ablation)RTD on LinCIR · CIR val (avg)Avg: +4.5Improved
replace frozen text encoder with RTD-updated one in composed queryLinCIR ViT-L/14 · CIRCO valcosine similarity to target image: 0.1 -> 0.29Improved
add RTD to projection-based methodsPic2Word, SEARLE, LinCIR · GeneCISR@1: e.g. 12.45 -> 13.18 (LinCIR ViT-L); +0.10 for SEARLE ViT-LMixed

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