The Predictive-Causal Gap: An Impossibility Theorem and Large-Scale Neural Evidence

ArXi:2605.05029v1 Announce Type: new We report a systematic failure mode in predictive representation learning. Across 2695 neural network configurations trained to predict linear-Gaussian dynamics, the optimal encoder tracks the environment rather than the system it is meant to model. The mean causal fidelity -- the fraction of encoder sensitivity allocated to system degrees of freedom -- is 0.49, and only 2.5% of configurations exceed 0.70.