@inproceedings{sheng2024hybridflow, title = {{HybridFlow}: A Flexible and Efficient {RLHF} Framework}, author = {Sheng, Guangming and Zhang, Chi and Ye, Zilingfeng and Wu, Xibin and Zhang, Wang and Zhang, Ru and Peng, Yanghua and Lin, Haibin and Wu, Chuan}, booktitle = {Twentieth European Conference on Computer Systems}, pages = {1279--1297}, year = {2025}, publisher = {ACM}, doi = {10.1145/3689031.3696075}, url = {https://doi.org/10.1145/3689031.3696075} } @article{yao2023deepspeedchat, title = {{DeepSpeed-Chat}: Easy, Fast and Affordable {RLHF} Training of {ChatGPT}-like Models at All Scales}, author = {Yao, Zhewei and Aminabadi, Reza Yazdani and Ruwase, Olatunji and Rajbhandari, Samyam and Wu, Xiaoxia and Awan, Ammar Ahmad and Rasley, Jeff and Zhang, Minjia and Li, Conglong and Holmes, Connor and Zhou, Zhongzhu and Wyatt, Michael and Smith, Molly and Kurilenko, Lev and Qin, Heyang and Tanaka, Masahiro and Che, Shuai and Song, Shuaiwen Leon and He, Yuxiong}, journal = {arXiv preprint arXiv:2308.01320}, year = {2023}, eprint = {2308.01320}, archiveprefix = {arXiv}, primaryclass = {cs.LG}, doi = {10.48550/arXiv.2308.01320} } @inproceedings{zhong2024rlhfuse, title = {Optimizing {RLHF} Training for Large Language Models with Stage Fusion}, author = {Zhong, Yinmin and Zhang, Zili and Wu, Bingyang and Liu, Shengyu and Chen, Yukun and Wan, Changyi and Hu, Hanpeng and Xia, Lei and Ming, Ranchen and Zhu, Yibo and Jin, Xin}, booktitle = {22nd USENIX Symposium on Networked Systems Design and Implementation (NSDI 25)}, pages = {489--503}, year = {2025}, publisher = {USENIX Association}, url = {https://www.usenix.org/conference/nsdi25/presentation/zhong} } @inproceedings{hu2024openrlhf, title = {{OpenRLHF}: A Ray-based Easy-to-use, Scalable and High-performance {RLHF} Framework}, author = {Hu, Jian and Wu, Xibin and Shen, Wei and Liu, Jason Klein and Wang, Weixun and Jiang, Songlin and Wang, Haoran and Chen, Hao and Chen, Bin and Fang, Wenkai and {Xianyu} and Cao, Yu and Xu, Haotian and Liu, Yiming}, booktitle = {2025 Conference on Empirical Methods in Natural Language Processing: System Demonstrations}, pages = {656--666}, year = {2025}, month = nov, address = {Suzhou, China}, publisher = {Association for Computational Linguistics}, doi = {10.18653/v1/2025.emnlp-demos.48}, url = {https://aclanthology.org/2025.emnlp-demos.48/}, isbn = {979-8-89176-334-0} } @inproceedings{mei2024realhf, title = {{ReaL}: Efficient {RLHF} Training of Large Language Models with Parameter Reallocation}, author = {Mei, Zhiyu and Fu, Wei and Li, Kaiwei and Wang, Guangju and Zhang, Huanchen and Wu, Yi}, booktitle = {Machine Learning and Systems}, volume = {7}, year = {2025}, url = {https://proceedings.mlsys.org/paper_files/paper/2025/file/3b3889d313ba9476c12c2d77ea66b24f-Paper-Conference.pdf} } @inproceedings{fu2025areal, title = {{AReaL}: A Large-Scale Asynchronous Reinforcement Learning System for Language Reasoning}, author = {Fu, Wei and Gao, Jiaxuan and Shen, Xujie and Zhu, Chen and Mei, Zhiyu and He, Chuyi and Xu, Shusheng and Wei, Guo and Mei, Jun and Wang, Jiashu and Yang, Tongkai and Yuan, Binhang and Wu, Yi}, booktitle = {Advances in Neural Information Processing Systems}, volume = {38}, pages = {36256--36282}, year = {2025}, publisher = {Curran Associates, Inc.}, doi = {10.52202/085713-1218}, url = {https://proceedings.neurips.cc/paper_files/paper/2025/hash/33c00862bfa29ac72ecf630a41e19352-Abstract-Conference.html} } @article{zhong2025streamrl, title = {{StreamRL}: Scalable, Heterogeneous, and Elastic {RL} for {LLMs} with Disaggregated Stream Generation}, author = {Zhong, Yinmin and Zhang, Zili and Song, Xiaoniu and Hu, Hanpeng and Jin, Chao and Wu, Bingyang and Chen, Nuo and Chen, Yukun and Zhou, Yu and Wan, Changyi and Zhou, Hongyu and Jiang, Yimin and Zhu, Yibo and Jiang, Daxin}, journal = {arXiv preprint arXiv:2504.15930}, year = {2025}, eprint = {2504.15930}, archiveprefix = {arXiv}, primaryclass = {cs.LG}, doi = {10.48550/arXiv.2504.15930} } @inproceedings{he2026hetrl, title = {{HetRL}: Efficient Reinforcement Learning for {LLMs} in Heterogeneous Environments}, author = {He, Yongjun and Zhang, Shuai and Gai, Jiading and Zhang, Xiyuan and Han, Boran and Wang, Bernie and Rangwala, Huzefa and Karypis, George}, booktitle = {Machine Learning and Systems}, volume = {8}, pages = {975--996}, year = {2026}, url = {https://proceedings.mlsys.org/paper_files/paper/2026/file/5321b1dabcd2be188d796c21b733e8c7-Paper-Conference.pdf} } @article{hu2026jigsawrl, title = {{JigsawRL}: Assembling {RL} Pipelines for Efficient {LLM} Post-Training}, author = {Hu, Zhengding and Ouyang, Hehua and Chen, Chang and Pan, Zaifeng and Guan, Yue and Yu, Zhongkai and Wang, Zhen and Swanson, Steven and Ding, Yufei}, journal = {arXiv preprint arXiv:2604.23838}, year = {2026}, eprint = {2604.23838}, archiveprefix = {arXiv}, primaryclass = {cs.LG}, doi = {10.48550/arXiv.2604.23838} } @article{lu2026rollplex, title = {{Rollplex}: Cross-Phase {GPU} Spatial Sharing for Vision Language Model Post-Training}, author = {Lu, Hanfeng and Feng, Tianyu and Li, Suyi and Zhao, Yuheng and Gao, Wei and Xiong, Shaopan and Huang, Ju and Yang, Siran and Wang, Jiamang and Qu, Lin and Wang, Wei}, journal = {arXiv preprint arXiv:2608.14498}, year = {2026}, eprint = {2608.14498}, archiveprefix = {arXiv}, primaryclass = {cs.LG}, doi = {10.48550/arXiv.2608.14498} } @article{santacroce2023hydra, title = {Efficient {RLHF}: Reducing the Memory Usage of {PPO}}, author = {Santacroce, Michael and Lu, Yadong and Yu, Han and Li, Yuanzhi and Shen, Yelong}, journal = {arXiv preprint arXiv:2309.00754}, year = {2023}, eprint = {2309.00754}, archiveprefix = {arXiv}, primaryclass = {cs.LG}, doi = {10.48550/arXiv.2309.00754} } @inproceedings{kwon2023pagedattention, title = {Efficient Memory Management for Large Language Model Serving with {PagedAttention}}, author = {Kwon, Woosuk and Li, Zhuohan and Zhuang, Siyuan and Sheng, Ying and Zheng, Lianmin and Yu, Cody Hao and Gonzalez, Joseph E. and Zhang, Hao and Stoica, Ion}, booktitle = {29th Symposium on Operating Systems Principles}, pages = {611--626}, year = {2023}, doi = {10.1145/3600006.3613165} } @misc{vllm2026sleepmode, title = {{vLLM} Sleep Mode}, author = {{vLLM Project}}, howpublished = {vLLM 0.18.0 documentation}, url = {https://docs.vllm.ai/en/v0.18.0/features/sleep_mode/}, note = {Accessed September 14, 2026} } @inproceedings{zhou2024rlhfmemory, title = {Understanding and Alleviating Memory Consumption in {RLHF} for {LLMs}}, author = {Zhou, Jin and Yang, Hanmei and Tang, Steven (Jiaxun) and Xiang, Mingcan and Guan, Hui and Liu, Tongping}, booktitle = {Machine Learning for Systems Workshop at NeurIPS 2024}, year = {2024}, url = {https://mlforsystems.org/assets/papers/neurips2024/paper8.pdf} } @article{lysenstoen2026sloguard, title = {{SLO-Guard}: Crash-Aware, Budget-Consistent Autotuning for {SLO}-Constrained {LLM} Serving}, author = {Lysenst{\o}en, Christian}, journal = {arXiv preprint arXiv:2604.17627}, year = {2026}, eprint = {2604.17627}, archiveprefix = {arXiv}, primaryclass = {cs.LG}, doi = {10.48550/arXiv.2604.17627} } @inproceedings{henderson2018matters, title = {Deep Reinforcement Learning that Matters}, author = {Henderson, Peter and Islam, Riashat and Bachman, Philip and Pineau, Joelle and Precup, Doina and Meger, David}, booktitle = {Thirty-Second AAAI Conference on Artificial Intelligence}, year = {2018}, doi = {10.1609/aaai.v32i1.11694}, url = {https://ojs.aaai.org/index.php/AAAI/article/view/11694} } @inproceedings{agarwal2021precipice, title = {Deep Reinforcement Learning at the Edge of the Statistical Precipice}, author = {Agarwal, Rishabh and Schwarzer, Max and Castro, Pablo Samuel and Courville, Aaron and Bellemare, Marc G.}, booktitle = {Advances in Neural Information Processing Systems}, volume = {34}, pages = {29304--29320}, year = {2021}, url = {https://proceedings.neurips.cc/paper/2021/hash/f514cec81cb148559cf475e7426eed5e-Abstract.html} } @article{shao2024deepseekmath, title = {{DeepSeekMath}: Pushing the Limits of Mathematical Reasoning in Open Language Models}, author = {Shao, Zhihong and Wang, Peiyi and Zhu, Qihao and Xu, Runxin and Song, Junxiao and Bi, Xiao and Zhang, Haowei and Zhang, Mingchuan and Li, Y. K. and Wu, Y. and Guo, Daya}, journal = {arXiv preprint arXiv:2402.03300}, year = {2024}, eprint = {2402.03300}, archiveprefix = {arXiv}, primaryclass = {cs.CL}, doi = {10.48550/arXiv.2402.03300} } @inproceedings{wu2026weave, title = {Weave: Efficient Co-Scheduling for Disaggregated {RL} Post-Training}, author = {Wu, Tianyuan and Cao, Lunxi and Wei, Yining and Gao, Wei and Zhao, Yuheng and An, Dakai and Xiong, Shaopan and Lv, Zhiqiang and Huang, Ju and Yang, Siran and Yu, Yinghao and Wang, Jiamang and Qu, Lin and Wang, Wei}, booktitle = {20th USENIX Symposium on Operating Systems Design and Implementation}, pages = {809--827}, year = {2026}, publisher = {USENIX Association}, url = {https://www.usenix.org/conference/osdi26/presentation/wu-tianyuan} } @inproceedings{wu2026rlboost, title = {{RLBoost}: Harvesting Preemptible Cloud Resources for Cost-Efficient Reinforcement Learning on {LLMs}}, author = {Wu, Yongji and Liu, Xueshen and Zheng, Haizhong and Gu, Juncheng and Chen, Beidi and Mao, Z. Morley and Krishnamurthy, Arvind and Stoica, Ion}, booktitle = {23rd USENIX Symposium on Networked Systems Design and Implementation}, pages = {1323--1339}, year = {2026}, publisher = {USENIX Association}, url = {https://www.usenix.org/conference/nsdi26/presentation/wu-yongji} } @inproceedings{wang2026dynarl, title = {{DynaRL}: Flexible and Dynamic Scheduling of Large-Scale Reinforcement Learning Training}, author = {Wang, Yuanqing and Lin, Hao and Hu, Junhao and Zhu, Chunyang and Zhang, Quanlu and Guo, Zhen and Zhang, Yuchen and Fu, Xu and Xu, Si and Dai, Bo and Huang, Zixiao and Yu, Chao and Li, Boxun and Dai, Guohao and Yang, Zhi and Wang, Yu}, booktitle = {20th USENIX Symposium on Operating Systems Design and Implementation}, pages = {847--862}, year = {2026}, publisher = {USENIX Association}, url = {https://www.usenix.org/conference/osdi26/presentation/wang-yuanqing} } @inproceedings{tan2026rollart, title = {{RollArt}: Disaggregated Multi-Task Agentic {RL} Training at Scale}, author = {Gao, Wei and Zhao, Yuheng and Wu, Tianyuan and Xiong, Shaopan and Wang, Weixun and An, Dakai and Cao, Lunxi and Muhtar, Dilxat and Liu, Zichen and Zhao, Haizhou and Huang, Ju and Yang, Siran and Li, Yongbin and Su, Wenbo and Wang, Jiamang and Qu, Lin and Zheng, Bo and Wang, Wei}, booktitle = {20th USENIX Symposium on Operating Systems Design and Implementation (OSDI 26)}, pages = {863--881}, year = {2026}, publisher = {USENIX Association}, url = {https://www.usenix.org/conference/osdi26/presentation/gao} } @inproceedings{chen2026robustrl, title = {{RobustRL}: Role-Based Fault Tolerance System for {RL} Post-Training}, author = {Chen, Zhenqian and Zhong, Baoquan and Li, Xiang and Dai, Qing and Zhao, Xinkui and Ye, Miao and Cheng, Ren and Zhang, Lufei and Yin, Jianwei}, booktitle = {20th USENIX Symposium on Operating Systems Design and Implementation}, pages = {1407--1424}, year = {2026}, publisher = {USENIX Association}, url = {https://www.usenix.org/conference/osdi26/presentation/chen-zhenqian} } @article{gao2026rose, author = {Wei Gao and Yuheng Zhao and Dilxat Muhtar and Dakai An and Xuchun Shang and Tianyuan Wu and Lunxi Cao and Shaopan Xiong and Weixun Wang and Ju Huang and Teng Ma and Siran Yang and Jiamang Wang and Lin Qu and Bo Zheng and Wei Wang}, title = {{ROSE}: Rollout On Serving {GPUs} via Cooperative Elasticity for Agentic {RL}}, journal = {arXiv preprint arXiv:2605.06534}, year = {2026}, eprint = {2605.06534}, archiveprefix = {arXiv}, primaryclass = {cs.DC}, doi = {10.48550/arXiv.2605.06534} } @inproceedings{wang2025sirius, title = {Colocating {ML} Inference and Training with Fast {GPU} Memory Handover}, author = {Wang, Jiali and Wang, Yankui and Han, Mingcong and Chen, Rong}, booktitle = {2025 USENIX Annual Technical Conference (USENIX ATC 25)}, pages = {1657--1675}, year = {2025}, month = jul, publisher = {USENIX Association}, address = {Boston, MA}, isbn = {978-1-939133-48-9}, url = {https://www.usenix.org/conference/atc25/presentation/wang-jiali} } @inproceedings{hu2022lora, title = {{LoRA}: Low-Rank Adaptation of Large Language Models}, author = {Hu, Edward J. and Shen, Yelong and Wallis, Phillip and Allen-Zhu, Zeyuan and Li, Yuanzhi and Wang, Shean and Wang, Lu and Chen, Weizhu}, booktitle = {International Conference on Learning Representations}, year = {2022}, url = {https://openreview.net/forum?id=nZeVKeeFYf9} } @article{cobbe2021gsm8k, title = {Training Verifiers to Solve Math Word Problems}, author = {Cobbe, Karl and Kosaraju, Vineet and Bavarian, Mohammad and Chen, Mark and Jun, Heewoo and Kaiser, Lukasz and Plappert, Matthias and Tworek, Jerry and Hilton, Jacob and Nakano, Reiichiro and Hesse, Christopher and Schulman, John}, journal = {arXiv preprint arXiv:2110.14168}, year = {2021}, eprint = {2110.14168}, archiveprefix = {arXiv}, doi = {10.48550/arXiv.2110.14168} } @inproceedings{hendrycks2021math, title = {Measuring Mathematical Problem Solving With the {MATH} Dataset}, author = {Hendrycks, Dan and Burns, Collin and Kadavath, Saurav and Arora, Akul and Basart, Steven and Tang, Eric and Song, Dawn and Steinhardt, Jacob}, booktitle = {Neural Information Processing Systems Track on Datasets and Benchmarks}, volume = {1}, year = {2021}, url = {https://datasets-benchmarks-proceedings.neurips.cc/paper_files/paper/2021/hash/be83ab3ecd0db773eb2dc1b0a17836a1-Abstract-round2.html} } @article{li2022alphacode, title = {Competition-level code generation with {AlphaCode}}, author = {Li, Yujia and Choi, David and Chung, Junyoung and Kushman, Nate and Schrittwieser, Julian and Leblond, Rémi and Eccles, Tom and Keeling, James and Gimeno, Felix and Dal Lago, Agustin and Hubert, Thomas and Choy, Peter and de Masson d'Autume, Cyprien and Babuschkin, Igor and Chen, Xinyun and Huang, Po-Sen and Welbl, Johannes and Gowal, Sven and Cherepanov, Alexey and Molloy, James and Mankowitz, Daniel J. and Sutherland Robson, Esme and Kohli, Pushmeet and de Freitas, Nando and Kavukcuoglu, Koray and Vinyals, Oriol}, journal = {Science}, volume = {378}, number = {6624}, pages = {1092--1097}, year = {2022}, doi = {10.1126/science.abq1158} } @misc{lumi2026, author = {{LUMI Consortium}}, title = {{GPU} Nodes -- {LUMI-G}, {LUMI} Documentation}, year = {2026}, url = {https://docs.lumi-supercomputer.eu/hardware/lumig/}, note = {Accessed October 21, 2026} } @misc{it4i2026karolina, author = {{IT4Innovations National Supercomputing Center}}, title = {{Karolina}: System Description}, year = {2026}, url = {https://www.it4i.cz/en/infrastructure/karolina}, note = {Accessed September 14, 2026} } @misc{kobza2026benchmark, author = {Kobza, Ond{\v r}ej and {\v S}ediv{\'y}, Jan}, title = {{RLVRAMBench}: Benchmark Evidence Release}, year = {2026}, howpublished = {Hugging Face dataset}, url = {https://huggingface.co/datasets/kobzaond/RLVRAMBench}, doi = {10.57967/hf/10780}, note = {Version electronics-submission-2026-10-23-r2; released October 23, 2026} } @misc{kobza2026reproduction, author = {Kobza, Ond{\v r}ej and {\v S}ediv{\'y}, Jan}, title = {{RLVRAMBench}: Resource Transfer Benchmark and Reproduction Code}, year = {2026}, howpublished = {GitHub repository}, url = {https://github.com/kobzaond/RLVRAMBench-reproduce}, note = {Tag electronics-submission-2026-10-23-r2; released October 23, 2026} }