EmbodiedSkills：一個用於編排、訓練與部署VLA智能體的統一框架

摘要

視覺-語言-動作（VLA）模型直接將視覺觀察與語言指令映射為機器人動作，然而長程任務所需的不僅是動作預測。智慧體必須在物理狀態演變的過程中，協調感知、規劃、執行、進度驗證與復原機制。單憑動作預測或模型生成的技能決策，並不足以保證所提出的操作在當前狀態下有效，亦無法保證其結果會被驗證。我們提出 EmbodiedSkills，一個統一框架，將每次技能決策視為一項執行提案：執行時期在執行前檢查其前置條件，並在執行後驗證其結果。一個共享的可執行技能介面，將高層級技能選取、受限的低層級 VLA 執行，以及動作後驗證串聯於單一智慧體循環之中。由於此介面保持固定，低層級 VLA 策略得以在不更改智慧體循環的前提下被替換或調整。該介面亦將規劃、執行、驗證與復原事件記錄為結構化軌跡，這些軌跡可為各別元件提供監督訊號，並在具備互動式回饋時支援選擇性的線上調適。我們以 Qwen3-VL 與 OpenPI/pi0.5 為基礎，在 RoboTwin 2.0 與 LIBERO 上實例化 EmbodiedSkills。經任務調適的低層級 VLA 策略在 RoboTwin 2.0 的 50 項任務中達到平均 86.20% 的成功率，並在四個 LIBERO 套件中達到 97.40% 的平均成功率。這些結果確立了 EmbodiedSkills 中所採用之任務調適低層級 VLA 策略的執行效能。在四項依賴記憶的 RMBench 任務中，相同的任務調適執行方法僅達到 12.5% 的平均成功率。本框架提供了一個可訓練且可檢視的智慧體層，用以將這些策略轉化為閉環具身系統。

English

Vision-language-action (VLA) models map visual observations and language instructions directly to robot actions, but long-horizon tasks require more than action prediction. An agent must coordinate perception, planning, execution, progress verification, and recovery as the physical state evolves. An action prediction or a model-generated skill decision does not, by itself, guarantee that the proposed operation is valid in the current state or that its outcome will be verified. We propose EmbodiedSkills, a unified framework that treats each skill decision as an execution proposal: the runtime checks its prerequisites before execution and verifies the outcome afterward. A shared executable-skill interface connects high-level skill selection, bounded low-level VLA execution, and post-action verification within a single agent loop. Because this interface remains fixed, low-level VLA policies can be replaced or adapted without changing the agent loop. The interface also records planning, execution, verification, and recovery events as structured trajectories, which provide supervision for individual components and can support optional online adaptation when interactive feedback is available. We instantiate EmbodiedSkills with Qwen3-VL and OpenPI/pi0.5 on RoboTwin 2.0 and LIBERO. Task-adapted low-level VLA policies achieve an average success rate of 86.20% across 50 RoboTwin 2.0 tasks and 97.40% across the four LIBERO suites. These results establish the execution performance of the task-adapted low-level VLA policies used in EmbodiedSkills. On four memory-dependent RMBench tasks, the same task-adapted execution approach achieves 12.5% average success. The framework provides a trainable and inspectable agent layer for turning these policies into closed-loop embodied systems.