# Training Throughput Measurement on LIBERO Source: [Training Throughput Measurement on LIBERO]() Collection index: [llms.txt]() Data updated: 2026-02-26T06:07:34.962Z Evaluation of the training throughput performance (e.g., samples processed per second) when utilizing the LIBERO environment or dataset. A research result applies to its stated dataset, split, metric and evaluation configuration. Compare like-for-like settings, preserve the metric's unit and direction, and distinguish paper-reported results from independent reproduction. The best result recorded here is not a claim of exhaustive, current global coverage. | Scope | Recorded value | | --- | --- | | Task | Training Throughput Measurement | | Dataset (including any stated split) | LIBERO | | Metric | Training Throughput | | Unit | SCALAR | | Direction | Higher is better | | Results loaded | 30 of 30 | | Coverage | All results currently returned for this benchmark | Evaluation settings may differ between results. The original papers and recorded configurations are needed to establish comparability. A shared metric name alone does not establish an identical protocol. ## Metrics - [Training Throughput](): Higher is better; unit: SCALAR. - [Percentage Increase](): Higher is better; unit: PERCENT. ## Recorded results | Method | Training Throughput | Result date | Report paper | | --- | --- | --- | --- | | Rollout Async | 1620.39 | Not recorded | 698575cfc97a0b3e3b5d5e1a | | Streamer | 1592.4 | Not recorded | 698575cfc97a0b3e3b5d5e1a | | Colocated | 1125.62 | Not recorded | 698575cfc97a0b3e3b5d5e1a | | Streamer | 1120.91 | Not recorded | 698575cfc97a0b3e3b5d5e1a | | Rollout Async | 1041.36 | Not recorded | 698575cfc97a0b3e3b5d5e1a | | Train Async | 951.33 | Not recorded | 698575cfc97a0b3e3b5d5e1a | | Rollout Async | 816.48 | Not recorded | 698575cfc97a0b3e3b5d5e1a | | Streamer | 816.48 | Not recorded | 698575cfc97a0b3e3b5d5e1a | | Train Async | 737.46 | Not recorded | 698575cfc97a0b3e3b5d5e1a | | Disaggregated | 729.98 | Not recorded | 698575cfc97a0b3e3b5d5e1a | | Streamer | 713.38 | Not recorded | 698575cfc97a0b3e3b5d5e1a | | Colocated | 703.85 | Not recorded | 698575cfc97a0b3e3b5d5e1a | | Rollout Async | 686.8 | Not recorded | 698575cfc97a0b3e3b5d5e1a | | Colocated | 680.46 | Not recorded | 698575cfc97a0b3e3b5d5e1a | | Colocated | 547.55 | Not recorded | 698575cfc97a0b3e3b5d5e1a | | Train Async | 477.2 | Not recorded | 698575cfc97a0b3e3b5d5e1a | | Disaggregated | 457.23 | Not recorded | 698575cfc97a0b3e3b5d5e1a | | Train Async | 441.49 | Not recorded | 698575cfc97a0b3e3b5d5e1a | | Streamer | 439.64 | Not recorded | 698575cfc97a0b3e3b5d5e1a | | Rollout Async | 434.2 | Not recorded | 698575cfc97a0b3e3b5d5e1a | | Disaggregated | 409.6 | Not recorded | 698575cfc97a0b3e3b5d5e1a | | Streamer | 383.4 | Not recorded | 698575cfc97a0b3e3b5d5e1a | | Colocated | 371.8 | Not recorded | 698575cfc97a0b3e3b5d5e1a | | Rollout Async | 369.56 | Not recorded | 698575cfc97a0b3e3b5d5e1a | | Disaggregated | 307.84 | Not recorded | 698575cfc97a0b3e3b5d5e1a | | Colocated | 289.23 | Not recorded | 698575cfc97a0b3e3b5d5e1a | | Train Async | 243.57 | Not recorded | 698575cfc97a0b3e3b5d5e1a | | Train Async | 229.68 | Not recorded | 698575cfc97a0b3e3b5d5e1a | | Disaggregated | 220.81 | Not recorded | 698575cfc97a0b3e3b5d5e1a | | Disaggregated | 162.75 | Not recorded | 698575cfc97a0b3e3b5d5e1a | ### Rollout Async Training Throughput: 1620.39. Result ID: 6996018c5698665ec437d6f3. [Reporting paper]() [Method origin paper]() Recorded evaluation configuration: {"VLA Backbone":"GR00T N1.5","Number of GPUs":"32 GPUs","GPU Allocation Ratio":"1:1"}. This record does not assert independent reproduction. ### Streamer Training Throughput: 1592.4. Result ID: 6996018c5698665ec437d6f4. [Reporting paper]() [Method origin paper]() Recorded evaluation configuration: {"VLA Backbone":"GR00T N1.5","Number of GPUs":"32 GPUs","GPU Allocation Ratio":"1:1"}. This record does not assert independent reproduction. ### Colocated Training Throughput: 1125.62. Result ID: 6996018c5698665ec437d6f0. [Reporting paper]() [Method origin paper]() Recorded evaluation configuration: {"VLA Backbone":"GR00T N1.5","Number of GPUs":"32 GPUs"}. This record does not assert independent reproduction. ### Streamer Training Throughput: 1120.91. Result ID: 6996018c5698665ec437d6e5. [Reporting paper]() [Method origin paper]() Recorded evaluation configuration: {"VLA Backbone":"pi0.5","Number of GPUs":"32 GPUs","GPU Allocation Ratio":"1:1"}. This record does not assert independent reproduction. ### Rollout Async Training Throughput: 1041.36. Result ID: 6996018c5698665ec437d6e4. [Reporting paper]() [Method origin paper]() Recorded evaluation configuration: {"VLA Backbone":"pi0.5","Number of GPUs":"32 GPUs","GPU Allocation Ratio":"1:1"}. This record does not assert independent reproduction. ### Train Async Training Throughput: 951.33. Result ID: 6996018c5698665ec437d6f2. [Reporting paper]() [Method origin paper]() Recorded evaluation configuration: {"VLA Backbone":"GR00T N1.5","Number of GPUs":"32 GPUs","GPU Allocation Ratio":"1:1"}. This record does not assert independent reproduction. ### Rollout Async Training Throughput: 816.48. Result ID: 6996018c5698665ec437d6ee. [Reporting paper]() [Method origin paper]() Recorded evaluation configuration: {"VLA Backbone":"GR00T N1.5","Number of GPUs":"16 GPUs","GPU Allocation Ratio":"1:1"}. This record does not assert independent reproduction. ### Streamer Training Throughput: 816.48. Result ID: 6996018c5698665ec437d6ef. [Reporting paper]() [Method origin paper]() Recorded evaluation configuration: {"VLA Backbone":"GR00T N1.5","Number of GPUs":"16 GPUs","GPU Allocation Ratio":"1:1"}. This record does not assert independent reproduction. ### Train Async Training Throughput: 737.46. Result ID: 6996018c5698665ec437d6e3. [Reporting paper]() [Method origin paper]() Recorded evaluation configuration: {"VLA Backbone":"pi0.5","Number of GPUs":"32 GPUs","GPU Allocation Ratio":"1:1"}. This record does not assert independent reproduction. ### Disaggregated Training Throughput: 729.98. Result ID: 6996018c5698665ec437d6f1. [Reporting paper]() [Method origin paper]() Recorded evaluation configuration: {"VLA Backbone":"GR00T N1.5","Number of GPUs":"32 GPUs","GPU Allocation Ratio":"1:1"}. This record does not assert independent reproduction. ### Streamer Training Throughput: 713.38. Result ID: 6996018c5698665ec437d6e0. [Reporting paper]() [Method origin paper]() Recorded evaluation configuration: {"VLA Backbone":"pi0.5","Number of GPUs":"16 GPUs","GPU Allocation Ratio":"1:1"}. This record does not assert independent reproduction. ### Colocated Training Throughput: 703.85. Result ID: 6996018c5698665ec437d6e1. [Reporting paper]() [Method origin paper]() Recorded evaluation configuration: {"VLA Backbone":"pi0.5","Number of GPUs":"32 GPUs"}. This record does not assert independent reproduction. ### Rollout Async Training Throughput: 686.8. Result ID: 6996018c5698665ec437d6df. [Reporting paper]() [Method origin paper]() Recorded evaluation configuration: {"VLA Backbone":"pi0.5","Number of GPUs":"16 GPUs","GPU Allocation Ratio":"1:1"}. This record does not assert independent reproduction. ### Colocated Training Throughput: 680.46. Result ID: 6996018c5698665ec437d6eb. [Reporting paper]() [Method origin paper]() Recorded evaluation configuration: {"VLA Backbone":"GR00T N1.5","Number of GPUs":"16 GPUs"}. This record does not assert independent reproduction. ### Colocated Training Throughput: 547.55. Result ID: 6996018c5698665ec437d6dc. [Reporting paper]() [Method origin paper]() Recorded evaluation configuration: {"VLA Backbone":"pi0.5","Number of GPUs":"16 GPUs"}. This record does not assert independent reproduction. ### Train Async Training Throughput: 477.2. Result ID: 6996018c5698665ec437d6ed. [Reporting paper]() [Method origin paper]() Recorded evaluation configuration: {"VLA Backbone":"GR00T N1.5","Number of GPUs":"16 GPUs","GPU Allocation Ratio":"1:1"}. This record does not assert independent reproduction. ### Disaggregated Training Throughput: 457.23. Result ID: 6996018c5698665ec437d6e2. [Reporting paper]() [Method origin paper]() Recorded evaluation configuration: {"VLA Backbone":"pi0.5","Number of GPUs":"32 GPUs","GPU Allocation Ratio":"1:1"}. This record does not assert independent reproduction. ### Train Async Training Throughput: 441.49. Result ID: 6996018c5698665ec437d6de. [Reporting paper]() [Method origin paper]() Recorded evaluation configuration: {"VLA Backbone":"pi0.5","Number of GPUs":"16 GPUs","GPU Allocation Ratio":"1:1"}. This record does not assert independent reproduction. ### Streamer Training Throughput: 439.64. Result ID: 6996018c5698665ec437d6ea. [Reporting paper]() [Method origin paper]() Recorded evaluation configuration: {"VLA Backbone":"GR00T N1.5","Number of GPUs":"8 GPUs","GPU Allocation Ratio":"1:1"}. This record does not assert independent reproduction. ### Rollout Async Training Throughput: 434.2. Result ID: 6996018c5698665ec437d6e9. [Reporting paper]() [Method origin paper]() Recorded evaluation configuration: {"VLA Backbone":"GR00T N1.5","Number of GPUs":"8 GPUs","GPU Allocation Ratio":"1:1"}. This record does not assert independent reproduction. ### Disaggregated Training Throughput: 409.6. Result ID: 6996018c5698665ec437d6ec. [Reporting paper]() [Method origin paper]() Recorded evaluation configuration: {"VLA Backbone":"GR00T N1.5","Number of GPUs":"16 GPUs","GPU Allocation Ratio":"1:1"}. This record does not assert independent reproduction. ### Streamer Training Throughput: 383.4. Result ID: 6996018c5698665ec437d6db. [Reporting paper]() [Method origin paper]() Recorded evaluation configuration: {"VLA Backbone":"pi0.5","Number of GPUs":"8 GPUs","GPU Allocation Ratio":"1:1"}. This record does not assert independent reproduction. ### Colocated Training Throughput: 371.8. Result ID: 6996018c5698665ec437d6e6. [Reporting paper]() [Method origin paper]() Recorded evaluation configuration: {"VLA Backbone":"GR00T N1.5","Number of GPUs":"8 GPUs"}. This record does not assert independent reproduction. ### Rollout Async Training Throughput: 369.56. Result ID: 6996018c5698665ec437d6da. [Reporting paper]() [Method origin paper]() Recorded evaluation configuration: {"VLA Backbone":"pi0.5","Number of GPUs":"8 GPUs","GPU Allocation Ratio":"1:1"}. This record does not assert independent reproduction. ### Disaggregated Training Throughput: 307.84. Result ID: 6996018c5698665ec437d6dd. [Reporting paper]() [Method origin paper]() Recorded evaluation configuration: {"VLA Backbone":"pi0.5","Number of GPUs":"16 GPUs","GPU Allocation Ratio":"1:1"}. This record does not assert independent reproduction. ### Colocated Training Throughput: 289.23. Result ID: 6996018c5698665ec437d6d7. [Reporting paper]() [Method origin paper]() Recorded evaluation configuration: {"VLA Backbone":"pi0.5","Number of GPUs":"8 GPUs"}. This record does not assert independent reproduction. ### Train Async Training Throughput: 243.57. Result ID: 6996018c5698665ec437d6e8. [Reporting paper]() [Method origin paper]() Recorded evaluation configuration: {"VLA Backbone":"GR00T N1.5","Number of GPUs":"8 GPUs","GPU Allocation Ratio":"1:1"}. This record does not assert independent reproduction. ### Train Async Training Throughput: 229.68. Result ID: 6996018c5698665ec437d6d9. [Reporting paper]() [Method origin paper]() Recorded evaluation configuration: {"VLA Backbone":"pi0.5","Number of GPUs":"8 GPUs","GPU Allocation Ratio":"1:1"}. This record does not assert independent reproduction. ### Disaggregated Training Throughput: 220.81. Result ID: 6996018c5698665ec437d6e7. [Reporting paper]() [Method origin paper]() Recorded evaluation configuration: {"VLA Backbone":"GR00T N1.5","Number of GPUs":"8 GPUs","GPU Allocation Ratio":"1:1"}. This record does not assert independent reproduction. ### Disaggregated Training Throughput: 162.75. Result ID: 6996018c5698665ec437d6d8. [Reporting paper]() [Method origin paper]() Recorded evaluation configuration: {"VLA Backbone":"pi0.5","Number of GPUs":"8 GPUs","GPU Allocation Ratio":"1:1"}. This record does not assert independent reproduction. Page 1 of 1. This is an index page, not the full collection.