隐藏在模型状态中的困惑信号，能判断何时该相信想象而非现实。
Reading the Room: Implicit Confusion Encoding in Recurrent World Model States
- 用线性探测发现状态中隐含的困惑信号，与预测误差无关。
- 该信号可解释80%的探测结果，由近期高错误步数加权决定。
- 直接编辑状态可改变行为，适合需要判断现实可信度的任务。
基于RSSM架构的世界模型（如DreamerV3）的递归隐藏状态 $h_t$ 仅通过降低预测误差训练。我们发现，该状态还隐含自身困惑信息：方向几乎与方差最大方向正交，任何基于方差的方法都无法检测。该信号不同于集成分歧（标记新输入）和重构误差（标记当前预测失败）。在固定预测误差但变化困惑的测试中，对 $h_t$ 的线性探测获得AUROC 0.72（5次运行），而集成基线低于随机水平。近期高误差步数的折现计数解释了探测输出的80%（$R^2=0.80$）。通过直接编辑 $h_t$ 观察行为变化，包括使用其他轨迹的真实值进行验证，确认信号为因果使用而非单纯存在。其几何结构和闭式表达在三个控制任务中具有泛化性；唯一一个完全分离测试通过，实际应用（决定何时检查现实）仅在两个任务中有效。
原文摘要 · Abstract (English)
World models built on the RSSM architecture, such as DreamerV3, keep a recurrent hidden state $h_t$ trained only to reduce prediction error. We show this state also tracks its own confusion, hiding in plain sight: nearly orthogonal to $h_t$'s directions of greatest variance, invisible to any variance-based method. It is functionally distinct from ensemble disagreement, which flags new inputs, and reconstruction error, which flags bad predictions right now. On a test holding prediction error fixed while confusion varies, a linear probe on $h_t$ finds the signal (AUROC 0.72, 5 runs), while an ensemble baseline scores below chance. A discounted count of recent high-error steps explains 80% of the probe's output ($R^2=0.80$). We confirm the signal is causally used, not merely present, by editing $h_t$ directly and watching behaviour change, including a check using real values from other trajectories instead of synthetic edits. Its geometry and closed form generalize across three control tasks; the decisive dissociation test itself holds cleanly on only one, and its practical use, deciding when to check reality instead of trusting imagination, generalizes to only two of the three tasks.
Thank you to arXiv for use of its open access interoperability. PaperDance 不是 arXiv 官方产品；中文卡片由大模型生成，请以原文为准。