基於技能框架進化的自我進化具身智能體

摘要

具身智能體日益以基礎模型為核心來構建系統，其效能不僅取決於模型權重，也取決於環繞模型的技能、情境、動作介面與執行框架。雖然監督式微調與強化學習能使智能體適應新環境，但它們需要額外的資料、獎勵與訓練運行；與此同時，許多免訓練且以程式碼為中心的方法依賴可程式化機器人API，而這些API在固定介面設定中可能無法取得。我們提出SHAPER，一個用於免訓練具身適應的自演化框架：它保持模型參數凍結，並透過在目標環境中的展開，演化可重用的技能與情境-程式碼執行框架，從而改進非參數化代理系統。在SHAPER中，同一個凍結模型既可充當規劃器也可充當優化器，在無需更新參數的情況下精煉其外部技能與情境-程式碼執行框架。我們在VLABench與ESI-Bench上評估SHAPER，涵蓋具備不同低階動作介面的具身智能體，並與純執行、監督式微調以及測試時擴展基線（如無驗證器選擇與投票）進行比較。結果表明，當模型訓練成本高昂、不可得或不理想時，技能與執行框架優化是實現自演化具身智能體的一條務實途徑。

English

Embodied agents are increasingly built as systems around foundation models, where performance depends not only on model weights but also on the skills, context, action interfaces, and execution harness surrounding the model. While supervised fine-tuning and reinforcement learning can adapt agents to new environments, they require additional data, rewards, and training runs; meanwhile, many train-free code-centric approaches rely on programmable robot APIs that may be unavailable in fixed-interface settings. We propose SHAPER, a self-evolving framework for train-free embodied adaptation that keeps model parameters frozen and improves the non-parametric agent system by evolving reusable skills and a context-code harness through target-environment rollouts. In SHAPER, the same frozen model can serve as both planner and optimizer, refining its external skills and context-code harness without parameter updates. We evaluate SHAPER on VLABench and ESI-Bench, covering embodied agents with different low-level action interfaces, and compare against pure execution, supervised fine-tuning, and test-time-scaling baselines such as verifier-free selection and voting. Our results suggest that skill-and-harness optimization is a practical route to self-evolving embodied agents when model training is expensive, unavailable, or undesirable.