# 预测动力学能否存在于物理世界中？

- 来源：HuggingFace Daily Papers（社区热门论文）
- 发布时间：2026-05-23 08:00
- AIHOT 分数：43
- AIHOT 链接：https://aihot.virxact.com/items/cmpwbvnb101itslsn7qqavab5
- 原文链接：https://arxiv.org/abs/2606.00089

## AI 摘要

预测性物理AI系统的输出（如状态预测、行动规划）即使误差低，也不一定物理可行。本研究提出了“物理可接受性”评估框架，将解码后的提案视为候选动力学，在执行前通过运动学、动力学等条件进行验证。验证不保证任务成功，但能识别违反物理约束的提案并给出组件级原因。在HuggingFace LeRobot PushT基准测试中，该完整验证门的AUC达0.957，残差过滤器能预防87-89%的无效提案，同时保持99.8%的任务正常进展。

## 正文

Predictive Physical AI systems output state rollouts, action chunks, and latent plans, yet a low root-mean-square error (RMSE) does not imply that a particular proposal is physically executable. We formulate physical admissibility as a prediction-control interface: before execution, a decoded proposal is treated as candidate dynamics and evaluated using kinematic, dynamic, and direct-to-composed horizon conditions. Passing is not a certificate of task success; rejection identifies violation of the specified physical envelope and gives a component-level reason. On Hugging Face LeRobot PushT, controlled falsification shows that one-step prediction-RMSE and standardized dynamics residuals reach area under the receiver operating characteristic curve (AUC) 0.982 and 0.972, kinematic-only conditions reach AUC 0.592, and the full gate reaches AUC 0.957 with condition-level attribution. In replay-based intervention experiments, residual-based filters and the full physical-admissibility gate prevent 87-$89% of invalid proposals while preserving mean progress near 0.998.