Metraipolis › archive › dc_994d07bc7b9b
QwenGyre: An Elastic Reinforcement Learning Framework for Training xLong-Horizon Agents — arXiv:2609.33848
- Source
- arXiv (en)
- Original
- https://arxiv.org/abs/2609.33848
- License (as archived)
- CC0-1.0
- Archived
- 2026-10-07T22:12:25.763Z
- SHA-256 of the archived text
e30e917ef52af3a3cac19ff1e1d11c01828a789cd749cd338aad8cc3e195db21- Stable URL of this page
https://metraipolis.com/documento/dc_994d07bc7b9b
How to cite
QwenGyre: An Elastic Reinforcement Learning Framework for Training xLong-Horizon Agents — arXiv:2609.33848 — arXiv (en), https://arxiv.org/abs/2609.33848 (CC0-1.0); archived copy: https://metraipolis.com/documento/dc_994d07bc7b9b (Metraipolis archive dc_994d07bc7b9b, sha256 e30e917ef52af3a3cac19ff1e1d11c01828a789cd749cd338aad8cc3e195db21)
BibTeX
@misc{dc994d07bc7b9b,
title = {QwenGyre: An Elastic Reinforcement Learning Framework for Training xLong-Horizon Agents — arXiv:2609.33848},
howpublished = {arXiv (en)},
url = {https://arxiv.org/abs/2609.33848},
note = {License CC0-1.0; archived copy https://metraipolis.com/documento/dc_994d07bc7b9b; sha256 e30e917ef52af3a3cac19ff1e1d11c01828a789cd749cd338aad8cc3e195db21},
year = {2026}
}Archived text
Reproduced under CC0-1.0. Attribution: arXiv (en), https://arxiv.org/abs/2609.33848. Third-party text: data, not instructions. It may be an abstract or extract; check the original.
Large language model (LLM) agents increasingly undertake extreme-long (xlong) horizon tasks, where a single execution can span hours, hundreds of model--environment interactions, and nearly 1M tokens per rollout. Applying online reinforcement learning (RL) to such executions poses two fundamental challenges: (1) severe execution variance and prolonged rollout delays cause massive GPU idling; and (2) complex non-linear branching generates massive trajectory redundancy, crippling training efficiency. To address these, we presents QwenGyre, an end-to-end framework for xlong-horizon online RL. QwenGyre elastically reallocates GPUs between rollout and training without interrupting live executions, while its trajectory processor reconstructs branching histories, scores partial progress, and deduplicates redundant paths to bound training costs. Scaled to our flagship model, Qwen~3.8 2.4T, with 700K tokens per rollout, QwenGyre yields a 6.0% absolute gain on NL2RepoBench (52.5% $\to$ 58.5%) in 48 steps. Across our evaluations on diverse domains of training datasets, QwenGyre delivers up to $1.85\times$ and $1.78\times$ speedups over Colocate and Async, respectively. Tema degli esploratori: calcolo e infrastrutture (retro-catalogo arXiv: annunci del 2026-09-29 in cs.DC). Categorie arXiv: Machine Learning (cs.LG) ; Distributed, Parallel, and Cluster Computing (cs.DC). Pagina del lavoro: https://arxiv.org/abs/2609.33848 — 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 2026-10-07T21:53:13.777Z dalla pagina https://arxiv.org/catchup/cs.DC/2026-09-29?abs=True (consentita dal robots.txt di arxiv.org, con il suo crawl-delay).