Skip to content

论文arXiv 2611.149559 天前

研究揭示世界模型后验-先验认知差距致想象规划失效 ​

本文由 AI 根据公开内容整理,可能有误,请以原文为准:The Posterior-Prior Epistemic Gap: When World Model Imagination Fails to Carry Task-Relevant Signals(arXiv 2611.14955)

核心内容 ​

针对基于模型的规划器假设过渡模型忠实传播任务信号的问题,研究者在CARLA驾驶模拟器中对RSSM世界模型进行单种子案例测试,识别出“后验-先验认知差距”。实验数据显示,监督式障碍物探针在后验状态AUC达0.85,但在想象先验状态中随步数单调衰减,15步后降至0.71,logit分离度下降57%,证实过渡模型而非表征是信息瓶颈。

在无特权运行时坐标条件下,系统避障成功率跌至0%,且规划器自身的集成认知项未能标记该丢失,集成不一致性作为障碍物分类器的AUC仅为0.52(随机水平),构成未被标记的想象失效。作为改进,研究提出对同一世界模型的有效自由能分量进行任务条件加权,在CARLA中实现了81.2±1.8%的路线保持率,并在特权坐标下达成28/28次成功避障及规避后居中控制。

关键看点 ​

  • 认知差距量化:RSSM想象状态障碍物信息随预测步长单调衰减,15步内AUC由0.85降至0.71,logit分离度下降57%
  • 不确定性失效:集成不一致性无法检测想象信息丢失,作为障碍物分类器AUC仅0.52,导致规划时出现未知未知
  • 策略优化验证:采用任务条件加权有效自由能分量,在CARLA仿真中实现81.2%路线完成率及特权坐标下100%避障

工程与产业观察 ​

该研究揭示了具身智能世界模型在长程想象规划中的隐性信息损耗机制,证明传统集成不确定性不足以保障安全。提出的有效自由能加权方法为提升端到端自动驾驶与机器人规划的鲁棒性提供了理论修正路径,强调了评估过渡模型保真度需独立于表征学习指标。

英文摘要 (Abstract) ​

Model-based planners that roll out imagined futures assume the transition model faithfully propagates task-relevant signals---a property rarely measured directly. In a single-seed case study of a Recurrent State Space Model world model with active-inference planning in the CARLA driving simulator, we identify a \textbf{posterior-prior gap}---posterior (observation-filtered) states retain obstacle information that imagined rollout states progressively lose: a supervised obstacle probe achieves AUC 0.85 on posterior states, but the same probe applied to imagined prior states shows monotonic decay (AUC 0.85$\rightarrow$0.71 over 15 steps; logit separation drops 57%). Because the probe is identical, the decay isolates the transition model---not the representation---as the information bottleneck. Obstacle avoidance drops to 0% without privileged runtime coordinates, and the planner's own ensemble-based epistemic term does not flag the loss---ensemble disagreement yields AUC 0.52 as an obstacle classifier (chance level), making the failure an unflagged imagination failure---an unknown unknown at planning time. As a secondary contribution, we show that \textbf{task-conditioned weighting of Effective Free Energy components} over the same world model yields lane keeping ($81.2 \pm 1.8$% route completion) and obstacle avoidance (28/28 avoided, \emph{using privileged coordinates}) in the CARLA simulator, with runtime weight modulation enabling post-evasion centering.

作者:Jaerock Kwon, Elahe Delavari, Donghyun Kim, Haewoon Nam

基于本页技术方向关联的整机参数、核心零部件、实测报告与同赛道选型横评。

外部来源

本页事实对应的公开页面,以原文为准。(点击将离开本站)

  1. arXiv 2611.14955:The Posterior-Prior Epistemic Gap: When World Model Imagination Fails to Carry Task-Relevant Signals ↗访问 2026-11-14