Towards Vision-Free CIR: Attribute-Augmented Scoring and LLM-Based Reranking for Zero-Shot Composed Image Retrieval

2026-07-14

AI Summary

This work addresses the performance limitations of vision-free methods in zero-shot compositional image retrieval (CIR), which stem from information loss in textual descriptions. To mitigate the absence of visual details, the authors propose a novel vision-free CIR framework featuring an attribute-augmented hybrid scoring mechanism that explicitly integrates attribute matching with large language model (LLM)-driven semantic consistency reranking. Evaluated on the CIRR dataset, the method achieves an R@1 score of 44.04%, substantially outperforming existing zero-shot approaches by 8.79%. Further experiments on FashionIQ reveal a nuanced trade-off between semantic reasoning and fine-grained visual alignment, underscoring the effectiveness and novelty of the proposed approach in complex multimodal retrieval scenarios.

Abstract

Recent work has shown that "Vision-Free'' approaches (representing images as text) can be effective for standard image retrieval tasks. However, it remains unclear whether this paradigm can effectively handle a more complex, multimodal task, Composed Image Retrieval (CIR), due to the inherent information loss in textual descriptions. In this paper, we introduce a Vision-Free CIR framework that addresses this challenge through two key techniques: (1) Attribute-Augmented Hybrid Scoring, which compensates for lost visual details via explicit attribute matching, and (2) LLM-Based Reranking, which verifies semantic consistency of top candidates. Experiments on the open-domain CIRR dataset show that our approach outperforms existing Zero-shot CIR methods (44.04% R@1, +8.79%). On FashionIQ, our results highlight the trade-off between semantic reasoning and fine-grained visual matching. Ablation studies reveal that both attribute-augmented scoring and LLM-Based Reranking consistently improve performance.

Problem

• Vision-Free
• Composed Image Retrieval
• Zero-Shot
• Information Loss
• Multimodal

Innovation

• Vision-Free CIR
• Attribute-Augmented Scoring
• LLM-Based Reranking
• Zero-Shot Composed Image Retrieval
• Text-Based Image Representation