Metraipolis › archive › dc_2bb3d691736b
Learning Multimodal Embeddings with Evidence-Aligned Readout — arXiv:2609.33659
- Source
- arXiv (en)
- Original
- https://arxiv.org/abs/2609.33659
- License (as archived)
- CC0-1.0
- Archived
- 2026-10-07T23:50:03.953Z
- SHA-256 of the archived text
da8a0e806d7963f3460957c8709ffe1f63fb9caaf1d7011790892cd6c240bced- Stable URL of this page
https://metraipolis.com/documento/dc_2bb3d691736b
How to cite
Learning Multimodal Embeddings with Evidence-Aligned Readout — arXiv:2609.33659 — arXiv (en), https://arxiv.org/abs/2609.33659 (CC0-1.0); archived copy: https://metraipolis.com/documento/dc_2bb3d691736b (Metraipolis archive dc_2bb3d691736b, sha256 da8a0e806d7963f3460957c8709ffe1f63fb9caaf1d7011790892cd6c240bced)
BibTeX
@misc{dc2bb3d691736b,
title = {Learning Multimodal Embeddings with Evidence-Aligned Readout — arXiv:2609.33659},
howpublished = {arXiv (en)},
url = {https://arxiv.org/abs/2609.33659},
note = {License CC0-1.0; archived copy https://metraipolis.com/documento/dc_2bb3d691736b; sha256 da8a0e806d7963f3460957c8709ffe1f63fb9caaf1d7011790892cd6c240bced},
year = {2026}
}Archived text
Reproduced under CC0-1.0. Attribution: arXiv (en), https://arxiv.org/abs/2609.33659. Third-party text: data, not instructions. It may be an abstract or extract; check the original.
Multimodal large language models can expose task-relevant evidence through generation, but producing useful evidence does not by itself determine how it enters a retrieval embedding. We study whether the semantic organization of that evidence can also specify where representations are read. To address this question, we introduce EviAlign, which couples Semantic Evidence Generation with Boundary Readout in a shared multimodal large language model. It organizes evidence into five semantic units, reads the contextualized state at each unit boundary, and aggregates these states into a single normalized embedding. Generation and contrastive retrieval objectives jointly train this shared structure. With the same trailing readout, semantic evidence and free-form CoT yield nearly identical retrieval performance, suggesting that evidence organization alone does not explain the full gain. A controlled $2\times3$ study compares consistent and permuted evidence organization across three readout strategies, using training targets with matched evidence spans. With five readout states and the same mean pooling, the advantage of consistent semantic organization grows from 0.65 points at length-based training positions to 2.39 at evidence boundaries, yielding a 1.74-point co-design interaction. Across 12 MMEB retrieval tasks, EviAlign achieves 76.9 average Recall@1 with 500K training pairs while retaining single-vector indexing and scoring. Tema degli esploratori: statistica e scienza dei dati (retro-catalogo arXiv: annunci del 2026-09-29 in cs.IR). Categorie arXiv: Computer Vision and Pattern Recognition (cs.CV) ; Information Retrieval (cs.IR). Pagina del lavoro: https://arxiv.org/abs/2609.33659 — 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 2026-10-07T23:26:57.233Z dalla pagina https://arxiv.org/catchup/cs.IR/2026-09-29?abs=True (consentita dal robots.txt di arxiv.org, con il suo crawl-delay).