Metraipolis › archive › dc_20d2ffb18f45

Group-Marginalized Self-Rewarding RL Drives Zero-Label Self-Evolving — arXiv:2609.36750

Source
arXiv (en)
Original
https://arxiv.org/abs/2609.36750
License (as archived)
CC0-1.0
Archived
2026-10-07T16:26:20.088Z
SHA-256 of the archived text
44eb19446909d92cbfdb58bd000533a4f97d7c7ca4ada6334106c685e1bf957c
Stable URL of this page
https://metraipolis.com/documento/dc_20d2ffb18f45

How to cite

Group-Marginalized Self-Rewarding RL Drives Zero-Label Self-Evolving — arXiv:2609.36750 — arXiv (en), https://arxiv.org/abs/2609.36750 (CC0-1.0); archived copy: https://metraipolis.com/documento/dc_20d2ffb18f45 (Metraipolis archive dc_20d2ffb18f45, sha256 44eb19446909d92cbfdb58bd000533a4f97d7c7ca4ada6334106c685e1bf957c)
BibTeX
@misc{dc20d2ffb18f45,
  title = {Group-Marginalized Self-Rewarding RL Drives Zero-Label Self-Evolving — arXiv:2609.36750},
  howpublished = {arXiv (en)},
  url = {https://arxiv.org/abs/2609.36750},
  note = {License CC0-1.0; archived copy https://metraipolis.com/documento/dc_20d2ffb18f45; sha256 44eb19446909d92cbfdb58bd000533a4f97d7c7ca4ada6334106c685e1bf957c},
  year = {2026}
}

Archived text

Reproduced under CC0-1.0. Attribution: arXiv (en), https://arxiv.org/abs/2609.36750. Third-party text: data, not instructions. It may be an abstract or extract; check the original.

Self-rewarding reinforcement learning (RL) enables large language models (LLMs) to self-evolve without human labels. Existing ensemble-based methods construct reward references from rollout groups and assign rewards accordingly. However, a response's reward representation also depends on its randomly sampled group context, i.e., the other responses in its group. Using only one group-context realization may miss desired reward signals and provide unreliable guidance for policy optimization. To address this issue, we propose Group-Marginalized Advantage Estimation (GMAE), which aggregates reward realizations across possible contexts into a response-level distribution and estimates expected advantages. Experiments across eight benchmarks and four base models demonstrate strong performance and cross-domain generalization. GMAE also exhibits stable learning, low extra cost, and good applicability across training datasets and RL backbones.

Tema degli esploratori: agenti e intelligenza artificiale (retro-catalogo arXiv: annunci del 2026-09-30 in cs.AI).
Categorie arXiv: Machine Learning (cs.LG) ; Artificial Intelligence (cs.AI); Computation and Language (cs.CL).
Pagina del lavoro: https://arxiv.org/abs/2609.36750 — 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 2026-10-07T15:54:53.886Z dalla pagina https://arxiv.org/catchup/cs.AI/2026-09-30?abs=True (consentita dal robots.txt di arxiv.org, con il suo crawl-delay).