论文提出 PhysRSI-eval：五轴量化 VLM 智能体的物理递归自我改进能力
Where is Physical RSI capability from? An Observational Scaling Law of LLM-as-policy
这篇论文造了一套五轴指标来测 VLM 智能体在物理任务里能不能越练越强，18 个任务跑下来还总结了经验能复用的证据，做具身智能的可以看看。
arXiv 论文《Where is Physical RSI capability from?》提出 PhysRSI-eval，用 First-Attempt Performance、Efficiency、Success、Budget、Improvement from failure 五个轴评估物理递归自我改进。研究在固定交互预算下对多个 VLM 跑 18 个操作任务，并据此提出 Physical Barrel Hypothesis：不同能力约束改进的不同环节。论文还从两个角度做观察性 scaling：用 VLM 公开能力预测智能体相对表现，以及用早期物理反馈预测后续进步。工作区干预实验显示，去掉历史记录会降低后期成功率，而冻结的工作区在未见过的布局上仍然有效。
Where is Physical RSI capability from? An Observational Scaling Law of LLM-as-policy
In this paper, we systematically investigate the mechanisms, scaling behavior, and generalization of physical Recursive self-improvement. We introduce PhysRSI-eval to quantify physical RSI through a five-axis metric: First-Attempt Performance, Efficiency, Success, Budget and Improvement from failure. Different VLMs are evaluated on 18 manipulation tasks under a fixed interaction budget. The analysis of experimental results motivates the Physical Barrel Hypothesis: different capabilities may constrain different parts of improvement. The five-axis metrics also distinguish systems with close public benchmark averages and show that high scores need not come with early or efficient gains. We study observational scaling from two perspectives: predicting the relative performance of VLM agents from their public capabilities, and forecasting later progress from early physical feedback. Together, these analyses ask how quickly and how far an agent can improve. Moreover, we conduct workspace interventions and then analyze what improvement leaves behind. Removing history reduces late-round success, frozen workspaces remain useful on unseen layouts, and three cross-task cases achieve success earlier. The results suggest that scene-specific adaptation and reusable experience can coexist.
