Prediction Bottlenecks Don't Discover Causal Structure (But Here's What They Actually Do)

arXiv:2605.09169v1 Announce Type: cross Abstract: A Mamba state-space model trained only for next-step prediction appears to recover Granger-causal structure through a simple readout $S = |W_{out} W_{in}|$, with early experiments suggesting the phenomenon generalized across architectures and...