外观
论文HF Daily Papers · arXiv 2610.285898 天前
AgenticRoboRL框架：9B机器人规划器后训练成功率达71%
本文由 AI 根据公开内容整理，可能有误，请以原文为准：AgenticRoboRL: Bringing Coding-Agent Post-Training to Agentic Robotics（HF Daily Papers · arXiv 2610.28589）
核心内容
针对现有具身智能系统过度依赖闭源大模型作为规划器且缺乏自身能力迭代的问题，研究团队提出AgenticRoboRL框架，旨在通过后训练提升机器人规划器本体能力。实验数据显示，在LIBERO-PRO基准测试中，该框架将9B参数规模规划器的任务成功率从基线18.1%显著提升至71.0%，在未使用外部记忆的情况下，性能已接近Codex（GPT-5.5）配合外部记忆所达到的72.1%成功率，证实了后训练技术路线的有效性。
技术方案采用监督微调结合多轮强化学习的组合策略，核心创新在于设计了Harness-specific Adapters机制。该适配器能够将VLA技能调用与控制代码生成等不同执行机制下的规划器交互记录统一映射至任务结果，从而构建通用的RL训练管线，同时保持底层执行组件固定不变。这种解耦设计解决了跨平台适配难题，使同一套后训练流程可复用于多种机器人硬件架构，团队计划开源该框架及训练后的规划器权重检查点。
关键看点
- 性能跃升：9B参数规划器在LIBERO-PRO基准成功率由18.1%升至71.0%，无外部记忆下逼近GPT-5.5的72.1%水平
- 训练范式：采用SFT+多轮RL组合策略，通过Harness-specific Adapters统一VLA技能调用与代码生成两种执行反馈
- 架构解耦：适配器机制实现通用RL管线与底层执行组件解耦，支持跨不同机器人Harness复用后训练流程
- 开源资产：团队承诺发布AgenticRoboRL框架源码及训练后的9B规划器Checkpoints，降低社区复现门槛
工程与产业观察
该工作证明了具身规划器无需盲目追求超大参数或闭源模型，通过针对性的后训练与RL对齐，9B级开源模型即可达到SOTA性能。其提出的Adapter解耦架构为异构机器人本体的统一算法迭代提供了工程范式，对降低具身智能研发成本、推动端侧小模型部署具有重要参考价值。
英文摘要 (Abstract)
Agentic robotics systems combine planners with robot harnesses to perform complex manipulation tasks. Recent work improves these systems through better harnesses and external memory, but relies on powerful closed-source models as planners and does not improve the planners' own capabilities. In contrast, coding agents have improved substantially by post-training the model itself. We therefore investigate whether post-training can also improve robot planners. However, applying it across robot harnesses requires accommodating different execution mechanisms, including VLA skill invocation and control-code generation. To this end, we introduce AgenticRoboRL, a framework for post-training robot planners across different harnesses. The training recipe combines supervised fine-tuning with multi-turn reinforcement learning. To accommodate different execution mechanisms, harness-specific adapters connect recorded planner interactions with task outcomes, enabling a common RL training pipeline while keeping execution components fixed. On LIBERO-PRO, AgenticRoboRL raises the success rate of a 9B-parameter planner from 18.1% to 71.0% without external memory, approaching the 72.1% achieved by Codex (GPT-5.5) with external memory. These results demonstrate post-training as an effective approach for improving robot planners. We will release the framework and trained planner checkpoints.
Hugging Face Daily Papers 热度：18 upvotes 作者：Xinxin Liu, Bingxin Xu, Wenrui Bao, Ming Li, Wentao Bao, Chen Chen, Yuzhang Shang
来源
本页事实对应的公开页面，以原文为准。