Metraipolis › archive › dc_8fac1cd01c5a

Efficient GPU Retrieval for Semantic Search — arXiv:2608.28968

Source
arXiv (en)
Original
https://arxiv.org/abs/2608.28968
License (as archived)
CC0-1.0
Archived
2026-10-17T07:43:55.347Z
SHA-256 of the archived text
50426d3e3662d5a72a4c44b43d3f85b90d785ee54eac261467840372355a6273
Stable URL of this page
https://metraipolis.com/documento/dc_8fac1cd01c5a

How to cite

Efficient GPU Retrieval for Semantic Search — arXiv:2608.28968 — arXiv (en), https://arxiv.org/abs/2608.28968 (CC0-1.0); archived copy: https://metraipolis.com/documento/dc_8fac1cd01c5a (Metraipolis archive dc_8fac1cd01c5a, sha256 50426d3e3662d5a72a4c44b43d3f85b90d785ee54eac261467840372355a6273)
BibTeX
@misc{dc8fac1cd01c5a,
  title = {Efficient GPU Retrieval for Semantic Search — arXiv:2608.28968},
  howpublished = {arXiv (en)},
  url = {https://arxiv.org/abs/2608.28968},
  note = {License CC0-1.0; archived copy https://metraipolis.com/documento/dc_8fac1cd01c5a; sha256 50426d3e3662d5a72a4c44b43d3f85b90d785ee54eac261467840372355a6273},
  year = {2026}
}

Archived text

Reproduced under CC0-1.0. Attribution: arXiv (en), https://arxiv.org/abs/2608.28968. Third-party text: data, not instructions. It may be an abstract or extract; check the original.

Semantic Search on LinkedIn must retrieve relevant profiles from a corpus of hundreds of millions in response to natural-language queries such as "a fintech founder in Berlin who worked in payments." The deployed relevance policy is bottleneck-oriented: every active non-negotiable facet must be satisfied, and a pre-existing LLM Graded Relevance (GR) judge operationalizes this through a fixed min/median aggregation over facet grades. Cosine similarity instead averages evidence, letting a strong match on one facet mask failure on another, capping the recall of the first-stage (L0) retriever. We present a policy-aligned retrieval framework: embeddings are partitioned into eight category-supervised segments whose scores follow the same min/median rule at serving time; for multi-vector retrieval, this segment score is computed independently per tagged document slot and maximized across slots. A lightweight single-slot Stage-1 scorer generates high-recall candidates, while scale-invariant relative-norm gating keeps category activation consistent across training, evaluation, and serving. On 21K held-out queries, this representation improves offline relevance over a matched-capacity baseline, with gains broadly distributed across facet combinations. We serve this framework with a two-stage GPU architecture: an FP8 coarse ranker scores the full corpus, increasing per-shard capacity by 71% and Stage-1 matmul throughput by 36%, then an FP16 stage exactly re-ranks an oversampled candidate set, recovering 99.6-99.8% of full-FP16 recall at over 500 QPS per shard replica. In a member-randomized A/B test, exploratory-query Precision@10 under the unchanged GR judge rises from 63.7% to 79.0% and navigational Precision@1 from 65.5% to 74.7%, with a blinded human evaluation independently confirming the Precision@10 gain.

Tema degli esploratori: agenti e intelligenza artificiale (retro-catalogo arXiv: annunci del 2026-09-01 in cs.AI).
Categorie arXiv: Artificial Intelligence (cs.AI) ; Machine Learning (cs.LG).
Pagina del lavoro: https://arxiv.org/abs/2608.28968 — 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 2026-10-17T07:28:26.491Z dalla pagina https://arxiv.org/catchup/cs.AI/2026-09-01?abs=True (consentita dal robots.txt di arxiv.org, con il suo crawl-delay).