Metraipolis › archive › dc_3dcf3f526fd2

AutoConcept: Training-Free Concept-Guided Reranking for Metadata-Available Composed Image Retrieval — arXiv:2609.01456

Source
arXiv (en)
Original
https://arxiv.org/abs/2609.01456
License (as archived)
CC0-1.0
Archived
2026-10-17T06:41:50.556Z
SHA-256 of the archived text
4bb67319b324e44fdfaee05fc9cc3a1b48d6f262c0fc4106773412e6e53026d8
Stable URL of this page
https://metraipolis.com/documento/dc_3dcf3f526fd2

How to cite

AutoConcept: Training-Free Concept-Guided Reranking for Metadata-Available Composed Image Retrieval — arXiv:2609.01456 — arXiv (en), https://arxiv.org/abs/2609.01456 (CC0-1.0); archived copy: https://metraipolis.com/documento/dc_3dcf3f526fd2 (Metraipolis archive dc_3dcf3f526fd2, sha256 4bb67319b324e44fdfaee05fc9cc3a1b48d6f262c0fc4106773412e6e53026d8)
BibTeX
@misc{dc3dcf3f526fd2,
  title = {AutoConcept: Training-Free Concept-Guided Reranking for Metadata-Available Composed Image Retrieval — arXiv:2609.01456},
  howpublished = {arXiv (en)},
  url = {https://arxiv.org/abs/2609.01456},
  note = {License CC0-1.0; archived copy https://metraipolis.com/documento/dc_3dcf3f526fd2; sha256 4bb67319b324e44fdfaee05fc9cc3a1b48d6f262c0fc4106773412e6e53026d8},
  year = {2026}
}

Archived text

Reproduced under CC0-1.0. Attribution: arXiv (en), https://arxiv.org/abs/2609.01456. Third-party text: data, not instructions. It may be an abstract or extract; check the original.

Composed image retrieval (CIR) retrieves a target image from a reference image and a text modification. This paper studies metadata-available CIR reranking, where a fixed CIR model first returns a candidate pool and gallery metadata is then used for second-stage concept-guided scoring. We introduce AutoConcept, a training-free reranker that converts concept evidence into an interpretable memory. AutoConcept filters noisy concepts, activates query-relevant positive constraints with an auxiliary negative penalty, and combines base retrieval scores with metadata-based concept-candidate alignment through inference-time calibration. On FashionIQ, AutoConcept yields significant early-rank improvements over WeiMoCIR and consistent plug-in gains on LinCIR candidate pools. Metadata-aware controls show that structured concept memory adds signal beyond direct query-text and extracted-attribute matching, while a query-only variant further supports the effectiveness of concept-level reranking. A supplementary real-human concept-label study indicates that the same memory interface can consume participant-provided evidence. These results position AutoConcept as an interpretable concept-memory reranker for product-style CIR galleries with available metadata.

Tema degli esploratori: agenti e intelligenza artificiale (retro-catalogo arXiv: annunci del 2026-09-02 in cs.CL).
Categorie arXiv: Information Retrieval (cs.IR) ; Computation and Language (cs.CL).
Pagina del lavoro: https://arxiv.org/abs/2609.01456 — autori e testo completo sono lì (qui omessi i nomi: nessun dato personale nel deposito).
Licenza: titolo e abstract sono metadati arXiv distribuiti in CC0 1.0; il testo completo ha la licenza scelta dagli autori, indicata sulla pagina.
Letto il 2026-10-17T06:33:38.903Z dalla pagina https://arxiv.org/catchup/cs.CL/2026-09-02?abs=True (consentita dal robots.txt di arxiv.org, con il suo crawl-delay).