mlsanityData to 7 Oct 2026

JoyNexus: Service-Oriented Multi-Tenant Post-Training for VLA Models

arXiv 2607.16074 ↗First posted 17 Jul 2026cs.DC cs.AI cs.SEcited by 0

On the map: Vision-language-action models → Efficient VLA inference and adaptation

Abstract (arXiv)

The post-training of Vision-Language-Action (VLA) models is essential due to the diversity of simulators, robot embodiments, and task objectives. Existing compute services, whether offered as direct accelerator rental or batch-workload submission, typically allocate an exclusive set of GPU and CPU resources to a single tenant. While this paradigm maximizes client flexibility, it burdens users with infrastructure adaptation, and the fixed card-hour accounting model renders short or bursty workloads both expensive for tenants and inefficient for the service provider. To address these challenges, we present JoyNexus, a unified service for multi-tenant VLA supervised fine-tuning, reinforcement learning, and evaluation. JoyNexus decouples the Training Model Service, Inference Model Service, and Environment Service, each accessed through APIs and backed by resident shared base models with tenant-specific slots. Tenants can directly invoke high-level semantic APIs for training, rollout, and evaluation, or compose custom algorithms using lower-level APIs and their assigned endpoints. Multiple tenants submit workloads concurrently; their action modules, optimizers, rollout records, and policy versions remain isolated, and the service is scheduled by the global Training Queue and Inference Queue. To further improve multi-tenant training efficiency, JoyNexus introduces group batching for heterogeneous VLA data schemas that share a compatible model-facing prefix, enabling a single shared backbone forward pass over grouped samples. Finally, we evaluate JoyNexus through workload simulation and a group-batching pipeline in a realistic embodied scenario. Results show that, compared with isolated single-tenant execution, JoyNexus reduces aggregate GPU time and improves service utilization via cross-tenant scheduling on shared resources.

Key facts. No citing papers in the map yet (data through 7 Oct 2026). Placed in Efficient VLA inference and adaptation, one of 13 directions in Vision-language-action models. Builds on 42 of its 59 references that are in the map. First posted on arXiv 17 Jul 2026.

Builds on

42 of 59 references.

References · 42

  1. 2023Efficient Memory Management for Large Language Model Serving with PagedAttentioncited by 5,767
  2. 2023RT-2: Vision-Language-Action Models Transfer Web Knowledge to Robotic Controlcited by 2,993
  3. 2024OpenVLA: An Open-Source Vision-Language-Action Modelcited by 2,777
  4. 2023Diffusion Policy: Visuomotor Policy Learning via Action Diffusioncited by 2,730
  5. 2024π0: A Vision-Language-Action Flow Model for General Robot Controlcited by 2,297
  6. 2023PaLM-E: An Embodied Multimodal Language Modelcited by 2,037
  7. 2022RT-1: Robotics Transformer for Real-World Control at Scalecited by 2,007
  8. 2024HybridFlow: A Flexible and Efficient RLHF Frameworkcited by 1,980
  9. 2019Megatron-LM: Training Multi-Billion Parameter Language Models Using Model Parallelismcited by 1,829
  10. 2025π0.5: a Vision-Language-Action Model with Open-World Generalizationcited by 1,653
  11. 2024Octo: An Open-Source Generalist Robot Policycited by 1,427
  12. 2023LIBERO: Benchmarking Knowledge Transfer for Lifelong Robot Learningcited by 1,338
30 more
  1. 2019ZeRO: Memory Optimizations Toward Training Trillion Parameter Modelscited by 1,278
  2. 2025GR00T N1: An Open Foundation Model for Generalist Humanoid Robotscited by 1,167
  3. 2023SGLang: Efficient Execution of Structured Language Model Programscited by 1,107
  4. 2025Fine-Tuning Vision-Language-Action Models: Optimizing Speed and Successcited by 844
  5. 2019RLBench: The Robot Learning Benchmark & Learning Environmentcited by 669
  6. 2021CALVIN: A Benchmark for Language-Conditioned Policy Learning for Long-Horizon Robot Manipulation Taskscited by 564
  7. 2017Ray: A Distributed Framework for Emerging AI Applicationscited by 516
  8. 2023ManiSkill2: A Unified Benchmark for Generalizable Manipulation Skillscited by 313
  9. 2025X-VLA: Soft-Prompted Transformer as Scalable Cross-Embodiment Vision-Language-Action Modelcited by 285
  10. 2025AReaL: A Large-Scale Asynchronous Reinforcement Learning System for Language Reasoningcited by 234
  11. 2016Clipper: A Low-Latency Online Prediction Serving Systemcited by 191
  12. 2023SARATHI: Efficient LLM Inference by Piggybacking Decodes with Chunked Prefillscited by 187
  13. 2023S-LoRA: Serving Thousands of Concurrent LoRA Adapterscited by 146
  14. 2024Scaling Proprioceptive-Visual Learning with Heterogeneous Pre-trained Transformerscited by 145
  15. 2025SimpleVLA-RL: Scaling VLA Training via Reinforcement Learningcited by 141
  16. 2023DeepSpeed-Chat: Easy, Fast and Affordable RLHF Training of ChatGPT-like Models at All Scalescited by 107
  17. 2026StarVLA: A Lego-like Codebase for Vision-Language-Action Model Developingcited by 103
  18. 2021Colossal-AI: A Unified Deep Learning System For Large-Scale Parallel Trainingcited by 96
  19. 2023Punica: Multi-Tenant LoRA Servingcited by 80
  20. 2025A Vision-Language-Action-Critic Model for Robotic Real-World Reinforcement Learningcited by 66
  21. 2025VLA-RFT: Vision-Language-Action Reinforcement Fine-tuning with Verified Rewards in World Simulatorscited by 65
  22. 2025RLinf: Flexible and Efficient Large-scale Reinforcement Learning via Macro-to-Micro Flow Transformationcited by 51
  23. 2021RLDS: an Ecosystem to Generate, Share and Use Datasets in Reinforcement Learningcited by 31
  24. 2023AlpaServe: Statistical Multiplexing with Model Parallelism for Deep Learning Servingcited by 31
  25. 2026ProRL Agent: Rollout-as-a-Service for RL Training of Multi-Turn LLM Agentscited by 17
  26. 2026RL-VLA3: A Flexible and Asynchronous Reinforcement Learning Framework for VLA Trainingcited by 7
  27. 2025LobRA: Multi-tenant Fine-tuning over Heterogeneous Datacited by 4
  28. 2026OpenTinker: Separating Concerns in Agentic Reinforcement Learningcited by 3
  29. 2026MARLaaS: Multi-Tenant Asynchronous Reinforcement Learning as a Servicecited by 1
  30. 2026MinT: Managed Infrastructure for Training and Serving Millions of LLMscited by 1

Built on by

Not cited yet.

Similar, not cited

Close in wording, no citation either way

  1. 2026Look Before You Leap: Distilling Tree Search into Action Evaluation for Frozen VLA Models
  2. 2026vla.cpp: A Unified Inference Runtime for Vision-Language-Action Models
  3. 2025RLinf-VLA: A Unified and Efficient Framework for Reinforcement Learning of Vision-Language-Action Models
  4. 2026Efficient Vision-Language-Action Management and Serving for Robot Factories
  5. 2026vla-eval: A Unified Evaluation Harness for Vision-Language-Action Models
  6. 2025SP-VLA: A Joint Model Scheduling and Token Pruning Approach for VLA Model Acceleration
  7. 2026Characterizing Vision-Language-Action Models across XPUs: Constraints and Acceleration for On-Robot Deployment
  8. 2026SOP: A Scalable Online Post-Training System for Vision-Language-Action Models