Metraipolis › archive › dc_b8098085e423

DE-Venus: A Data-Efficient RLVR Framework for Large Language Models — arXiv:2609.03324

Source
arXiv (en)
Original
https://arxiv.org/abs/2609.03324
License (as archived)
CC0-1.0
Archived
2026-10-17T02:07:49.891Z
SHA-256 of the archived text
5bebc3114e494ba2cf9f1abb6cad712401ba97503708c2d89b852b87d8274af3
Stable URL of this page
https://metraipolis.com/documento/dc_b8098085e423

How to cite

DE-Venus: A Data-Efficient RLVR Framework for Large Language Models — arXiv:2609.03324 — arXiv (en), https://arxiv.org/abs/2609.03324 (CC0-1.0); archived copy: https://metraipolis.com/documento/dc_b8098085e423 (Metraipolis archive dc_b8098085e423, sha256 5bebc3114e494ba2cf9f1abb6cad712401ba97503708c2d89b852b87d8274af3)
BibTeX
@misc{dcb8098085e423,
  title = {DE-Venus: A Data-Efficient RLVR Framework for Large Language Models — arXiv:2609.03324},
  howpublished = {arXiv (en)},
  url = {https://arxiv.org/abs/2609.03324},
  note = {License CC0-1.0; archived copy https://metraipolis.com/documento/dc_b8098085e423; sha256 5bebc3114e494ba2cf9f1abb6cad712401ba97503708c2d89b852b87d8274af3},
  year = {2026}
}

Archived text

Reproduced under CC0-1.0. Attribution: arXiv (en), https://arxiv.org/abs/2609.03324. Third-party text: data, not instructions. It may be an abstract or extract; check the original.

Reinforcement learning with verifiable rewards (RLVR) improves large language model reasoning, but its practical scaling is constrained by expensive on-policy rollouts and the cost of obtaining reliable targets at scale. Existing methods address sample selection, incomplete supervision, or noisy labels separately, often entangling supervision logic with distributed training and hindering controlled comparison and reuse. We present DE-Venus, a unified framework for data-efficient RLVR that treats supervision as evolving state across data preparation and policy optimization. It organizes this lifecycle into three modules: Active Data Selection allocates training and annotation budgets; Weak Supervision Construction derives learning signals from unlabeled examples; and Training-Time Supervision Refinement filters or corrects unreliable supervision. DE-Venus supports seven representative methods and a data-selection pipeline by expressing method-specific decisions as offline dataset transitions or online transformations of targets, rewards, batches, and advantages while preserving verl's distributed execution contracts. Across public benchmarks and three business scenarios, separate configurations preserve or improve model quality with only 10% of labels or as little as 13% of relevant data; selected business configurations also reduce observed convergence steps by 63%--75%. DE-Venus thus reduces annotation and training costs without sacrificing scalable RL execution.

Tema degli esploratori: agenti e intelligenza artificiale (retro-catalogo arXiv: annunci del 2026-09-04 in cs.LG).
Categorie arXiv: Machine Learning (cs.LG).
Pagina del lavoro: https://arxiv.org/abs/2609.03324 — 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 2026-10-17T01:47:25.538Z dalla pagina https://arxiv.org/catchup/cs.LG/2026-09-04?abs=True (consentita dal robots.txt di arxiv.org, con il suo crawl-delay).