Do Composed Image Retrieval Benchmarks Require Multimodal Composition?
Fuente:
arXiv
Gespeichert in:
| Hauptverfasser: | , , , , , , , , , , |
|---|---|
| Format: | Preprint |
| Veröffentlicht: |
2026
|
| Schlagworte: | |
| Online-Zugang: | |
| Tags: |
Tag hinzufügen
Keine Tags, Fügen Sie den ersten Tag hinzu!
|
| _version_ | 1866914571016470528 |
|---|---|
| author | Attimonelli, Matteo De Bellis, Alessandro Gema, Aryo Pradipta Saxena, Rohit Sekoyan, Monica Kwan, Wai-Chung Pomo, Claudio Suglia, Alessandro Jannach, Dietmar Di Noia, Tommaso Minervini, Pasquale |
| author_facet | Attimonelli, Matteo De Bellis, Alessandro Gema, Aryo Pradipta Saxena, Rohit Sekoyan, Monica Kwan, Wai-Chung Pomo, Claudio Suglia, Alessandro Jannach, Dietmar Di Noia, Tommaso Minervini, Pasquale |
| contents | 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. |
| format | Preprint |
| id |
arxiv_https___arxiv_org_abs_2605_14787 |
| institution | arXiv |
| publishDate | 2026 |
| record_format | arxiv |
| spellingShingle | Do Composed Image Retrieval Benchmarks Require Multimodal Composition? Attimonelli, Matteo De Bellis, Alessandro Gema, Aryo Pradipta Saxena, Rohit Sekoyan, Monica Kwan, Wai-Chung Pomo, Claudio Suglia, Alessandro Jannach, Dietmar Di Noia, Tommaso Minervini, Pasquale Computer Vision and Pattern Recognition Computation and Language 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. |
| title | Do Composed Image Retrieval Benchmarks Require Multimodal Composition? |
| topic | Computer Vision and Pattern Recognition Computation and Language |
| url | https://arxiv.org/abs/2605.14787 |