AI Paper Atlas

Research map / Vision-language and multimodal / Multimodal LLMs

Suit the Remedy to the Retriever: Interpretable Query Optimization with Retriever Preference Alignment for Vision-Language Retrieval

GuangHao Meng, Jinpeng Wang, Jieming Zhu, Letian Zhang, Yong Jiang, Dan Zhao, Qing Li

AAAI 2026 · Vision-language and multimodal › Multimodal LLMs · approach: Bias-aware multi-image robustness · 0 citations (OpenAlex)

Read the paperCheck your idea against 137,217 papers

AI-extracted summary. Written by AI from the paper's full text. Numbers and claims can be wrong or missing context, so check the paper before you cite or build on it.

Proposes interpretable query optimization that aligns query rewriting with each retriever's preferences for vision-language retrieval (abstract only available).

Problem

Challenging concepts in text or image queries confuse vision-language retrievers, and existing query optimization methods ignore what text descriptions retrievers prefer.

Approach

Optimizes queries interpretably while aligning them to retriever preferences, i.e., rewriting queries into descriptions that better match a given retriever's embedding space. Details beyond the abstract are not available.

Measured effects reported in the paper

What was changedWhereMeasuredEffect
add QRE-SFT on preference data to ReCoT rewriter (ablation)RAQO (CLIP ViT-B/32 retriever) · Flickr30k-CFQI2T R@1: 74.4 -> 75.5Improved
add CCD-SFT (DCCR) on top of QRE-SFT (ablation)RAQO (CLIP ViT-B/32 retriever) · Flickr30k-CFQI2T R@1: 75.5 -> 76.0Improved
add preference rank optimization (DPRO) (ablation)RAQO (CLIP ViT-B/32 retriever) · Flickr30k-CFQI2T R@1: 76.0 -> 77.0Improved
add preference rank optimization (DPRO) (ablation)RAQO (CLIP ViT-B/32 retriever) · Flickr30k-CFQT2I R@1: 54.0 -> 55.1Improved
add ReCoT detect-then-rewrite (ablation)RAQO (CLIP ViT-B/32 retriever) · Flickr30k-CFQI2T R@1: 76.1 -> 77.0Improved
add ReCoT without preference data (ablation)RAQO (CLIP ViT-B/32 retriever) · Flickr30k-CFQI2T R@1: 74.2 -> 74.4No clear effect
full preference post-training vs ReCoT only (ablation)RAQO (CLIP ViT-B/32 retriever) · Flickr30k-CFQI2T R@1: 74.4 -> 77.0Improved
RAQO vs LLMsRewrite (retriever-agnostic rewriting)RAQO (CLIP ViT-B/32 retriever) · Flickr30k-CFQI2T R@1: 74.2 -> 77.0Improved
apply RAQO to BLIP2 retrieverRAQO (CLIP ViT-B/32 retriever) · Flickr30k-CFQI2T R@1: 85.2 -> 86.0Improved
apply RAQO to CLIP, zero-shot to WikiDO OODRAQO (CLIP ViT-B/32 retriever) · WikiDO OODI2T R@1: 73.4 -> 73.9Improved
apply RAQO on Fashion200KRAQO (CLIP ViT-B/32 retriever) · Fashion200KI2T R@1: 9.5 -> 10.9Improved

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.

Limitations

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