Metraipolis › archive › dc_133eb6bf0abf

Learning from What You Retrieve: Online RL Fine-Tuning for Semantic Retrieval — arXiv:2608.30753

Source
arXiv (en)
Original
https://arxiv.org/abs/2608.30753
License (as archived)
CC0-1.0
Archived
2026-10-17T07:52:58.070Z
SHA-256 of the archived text
e49e21193db24fe4d16629fc68c259b732ab192d5a34e839358ca263beffc301
Stable URL of this page
https://metraipolis.com/documento/dc_133eb6bf0abf

How to cite

Learning from What You Retrieve: Online RL Fine-Tuning for Semantic Retrieval — arXiv:2608.30753 — arXiv (en), https://arxiv.org/abs/2608.30753 (CC0-1.0); archived copy: https://metraipolis.com/documento/dc_133eb6bf0abf (Metraipolis archive dc_133eb6bf0abf, sha256 e49e21193db24fe4d16629fc68c259b732ab192d5a34e839358ca263beffc301)
BibTeX
@misc{dc133eb6bf0abf,
  title = {Learning from What You Retrieve: Online RL Fine-Tuning for Semantic Retrieval — arXiv:2608.30753},
  howpublished = {arXiv (en)},
  url = {https://arxiv.org/abs/2608.30753},
  note = {License CC0-1.0; archived copy https://metraipolis.com/documento/dc_133eb6bf0abf; sha256 e49e21193db24fe4d16629fc68c259b732ab192d5a34e839358ca263beffc301},
  year = {2026}
}

Archived text

Reproduced under CC0-1.0. Attribution: arXiv (en), https://arxiv.org/abs/2608.30753. Third-party text: data, not instructions. It may be an abstract or extract; check the original.

In large-scale e-commerce retrieval, dual-encoder retrievers are op- timized for contrastive similarity, whereas downstream rerankers capture finer-grained relevance preferences; this objective mis- match limits end-to-end retrieval quality. Reinforcement Learning offers a way to use reward-model feedback for retriever adaptation, but we observe that standard policy-gradient updates can degrade embedding geometry, especially when the document index must remain frozen due to industrial constraints. To address this, we propose PAO (Positive-Advantage-Only), a selective RL optimization method. Our analysis reveals that in- discriminate penalization of negative samples (pushing away) in a frozen high-dimensional space disrupts pre-trained semantic man- ifolds. PAO selectively applies gradient updates only to retrieved items with positive advantages, effectively pulling query embed- dings toward high-reward regions while preserving global topo- logical stability. Experiments on both a massive industrial dataset and public benchmarks demonstrate that PAO significantly outper- forms standard RL and distillation baselines.

Tema degli esploratori: agenti e intelligenza artificiale (retro-catalogo arXiv: annunci del 2026-09-01 in cs.AI).
Categorie arXiv: Information Retrieval (cs.IR) ; Artificial Intelligence (cs.AI).
Pagina del lavoro: https://arxiv.org/abs/2608.30753 — 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 2026-10-17T07:28:26.906Z 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).