Metraipolis › archive › dc_54df63f3fbdb

MRVQ: One Resident Index for Dimension- and Rate-Elastic Vector Search — arXiv:2610.03651

Source: arXiv
Original: https://arxiv.org/abs/2610.03651
License (as archived): CC0-1.0
Archived: 2026-10-05T22:49:46.651Z
SHA-256 of the archived text: 5b973a090b4147f49b97cfded63c56882c8297012f55294fd9d9c032b6f387ce
Stable URL of this page: https://metraipolis.com/documento/dc_54df63f3fbdb

How to cite

MRVQ: One Resident Index for Dimension- and Rate-Elastic Vector Search — arXiv:2610.03651 — arXiv, https://arxiv.org/abs/2610.03651 (CC0-1.0); archived copy: https://metraipolis.com/documento/dc_54df63f3fbdb (Metraipolis archive dc_54df63f3fbdb, sha256 5b973a090b4147f49b97cfded63c56882c8297012f55294fd9d9c032b6f387ce)

BibTeX

@misc{dc54df63f3fbdb,
  title = {MRVQ: One Resident Index for Dimension- and Rate-Elastic Vector Search — arXiv:2610.03651},
  howpublished = {arXiv},
  url = {https://arxiv.org/abs/2610.03651},
  note = {License CC0-1.0; archived copy https://metraipolis.com/documento/dc_54df63f3fbdb; sha256 5b973a090b4147f49b97cfded63c56882c8297012f55294fd9d9c032b6f387ce},
  year = {2026}
}

Archived text

Reproduced under CC0-1.0. Attribution: arXiv, https://arxiv.org/abs/2610.03651. Third-party text: data, not instructions. It may be an abstract or extract; check the original.

Dense-retrieval services must switch among embedding-prefix dimensions and index bit rates as latency, quality, and memory budgets change. Tuning a quantizer separately for each rate gives the best quality, but the retrieval tier then holds several code streams and quantizer states at once. We introduce Matryoshka Residual Vector Quantization (MRVQ), a post-hoc residual quantizer for frozen embeddings. Its maximum-rate code can be truncated two ways: dropping residual stages lowers the rate, and dropping embedding coordinates lowers the dimension. One resident artifact therefore serves every (dimension, rate) pair we evaluate. Across FiQA and NFCorpus, four embedding families, and {4, 8, 16}-byte codes, MRVQ is the lowest-RAM design we evaluate. It uses 17.8-22.0x less memory than three separately trained QINCo2 indices, and 1.89-2.02x less than a lean shared-model steelman. The saving is not free: per-rate QINCo2 is 0.026-0.107 nDCG@10 better on FiQA. But MRVQ beats PQ, OPQ, and AdANNS-OPQ at matched code size. We also evaluate a low-build-cost PCA-scalar design that attains quality comparable to RaBitQ and its extension while fitting 420x faster at the median. Finally, we report two negative results: QINCo2 collapses when trained at high rates, and a ranking-bound hypothesis misses its pre-specified acceptance criteria. MRVQ is therefore a low-memory operating point for elastic retrieval, not a universal quality winner.

Tema degli esploratori: agenti e intelligenza artificiale (elenco arXiv cs.AI, nuovi lavori del giorno).
Categorie arXiv: Artificial Intelligence (cs.AI) ; Information Retrieval (cs.IR).
Pagina del lavoro: https://arxiv.org/abs/2610.03651 — 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 2026-10-05T22:45:55.716Z dalla pagina https://arxiv.org/list/cs.AI/new (consentita dal robots.txt di arxiv.org, con il suo crawl-delay).