Do Composed Image Retrieval Benchmarks Require Multimodal Composition?

TL;DR

This paper reveals that many composed image retrieval benchmarks can be solved using single modalities, indicating that current models may not truly require multimodal composition for high performance.

Contribution

The study uncovers the prevalence of unimodal shortcuts in CIR benchmarks and emphasizes the need for more rigorous evaluation of multimodal composition.

Findings

A large fraction of queries are solvable with a single modality (32.2% to 83.6%).

Re-evaluation on validated shortcut-free queries shows reliance on multimodal signals increases.

Current benchmarks conflate shortcut-solvable, noisy, and genuine compositional queries.

Abstract

Composed Image Retrieval (CIR) is a multimodal retrieval task where a query consists of a reference image and a textual modification, and the goal is to retrieve a target image satisfying both. In principle, strong performance on CIR benchmarks is assumed to require multimodal composition, i.e., combining complementary information from reference image and textual modification. In this work, we show that this assumption does not always hold. Across four widely used CIR benchmarks and eleven Generalist Multimodal Embedding models, a large fraction of queries can be solved using a single modality (from 32.2% to 83.6%), revealing pervasive unimodal shortcuts. Thus, high CIR performance can arise from unimodal signals rather than true multimodal composition. To better understand this issue, we perform a two-stage audit. First, we identify shortcut-solvable queries through cross-model analysis. Second, we conduct human validation on 4,741 shortcut-free queries, of which only 1,689 are well-formed, with common issues including ambiguous edits and mismatched targets. Re-evaluating models on this validated subset reveals qualitatively different behaviour: queries can no longer be solved with a single modality, and successful retrieval requires combining both inputs. While accuracy decreases, reliance on multimodal information increases. Overall, current CIR benchmarks conflate shortcut-solvable, noisy, and genuinely compositional queries, leading to an overestimation of model capability in multimodal composition.