ICLR 2026 Orals

MetaEmbed: Scaling Multimodal Retrieval at Test-Time with Flexible Late Interaction

Zilin Xiao, Qi Ma, Mengting Gu, Chun-cheng Jason Chen, Xintao Chen, Vicente Ordonez, Vijai Mohan

Multimodal & Speech Fri, Apr 24 · 3:51 PM–4:01 PM · 202 A/B Avg rating: 7.00 (6–8)

Abstract

Universal multimodal embedding models have achieved great success in capturing semantic relevance between queries and candidates. However, current methods either condense queries and candidates into a single vector, potentially limiting the expressiveness for fine-grained information, or produce too many vectors that are prohibitively expensive for multi-vector retrieval. In this work, we introduce MetaEmbed, a new framework for multimodal retrieval that rethinks how multimodal embeddings are constructed and interacted with at scale. During training, a fixed number of learnable Meta Tokens are appended to the input sequence. At test-time, their last-layer contextualized representations serve as compact yet expressive multi-vector embeddings. Through the proposed Matryoshka Multi-Vector Retrieval training, MetaEmbed learns to organize information by granularity across multiple vectors. As a result, we enable test-time scaling in multimodal retrieval where users can balance retrieval quality against efficiency demands by selecting the number of tokens used for indexing and retrieval interactions. Extensive evaluations on the Massive Multimodal Embedding Benchmark (MMEB) and the Visual Document Retrieval Benchmark (ViDoRe) confirm that MetaEmbed achieves state-of-the-art retrieval performance while scaling robustly to models with 32B parameters. Code is available at https://github.com/facebookresearch/MetaEmbed.

One-sentence summary·Auto-generated by claude-haiku-4-5-20251001(?)

MetaEmbed uses learnable meta tokens with matryoshka training to enable test-time scaling for multimodal retrieval balancing quality and efficiency.

Contributions·Auto-generated by claude-haiku-4-5-20251001(?)
  • Introduces MetaEmbed framework using learnable Meta Tokens whose contextualized representations serve as compact multi-vector embeddings
  • Proposes Matryoshka Multi-Vector Retrieval training to organize information by granularity across multiple vectors
  • Enables flexible late interaction at test-time where users can balance retrieval quality against efficiency by selecting number of tokens
Methods used·Auto-generated by claude-haiku-4-5-20251001(?)
  • Meta tokens
  • Matryoshka training
  • Late interaction
  • Multi-vector retrieval
Datasets used·Auto-generated by claude-haiku-4-5-20251001(?)
  • Massive Multimodal Embedding Benchmark (MMEB)
  • Visual Document Retrieval Benchmark (ViDoRe)
Limitations (author-stated)·Auto-generated by claude-haiku-4-5-20251001(?)
  • Index memory consumption grows proportionally with retrieval budget, presenting challenges for large deployments though mitigatable through balanced budgets or CPU memory offloading
    from the paper
Future work (author-stated)·Auto-generated by claude-haiku-4-5-20251001(?)

Authors did not state explicit future directions.

Author keywords

  • multimodal retrieval
  • information retrieval

Related orals

Something off? Let us know →