View a PDF of the paper titled Learning from Transients: Linear RNN Scaling Laws in Stationary Prediction, by Ziyan Chen and 4 other authors
Abstract:Does the organization of a fixed number of transitions into independent trajectories affect a fixed stationary prediction task? We study a Gaussian-sketched linear recurrent teacher–student model, where independently initialized training trajectories create transients that can excite weak spectral directions. For full-batch gradient descent with stable deterministic learning rates, we establish matching stationary approximation, bias, variance, and total-risk scaling laws from dependent trajectory data. When initialization decays at least as fast spectrally as the innovations, trajectory count and length enter the leading variance rate through their product. With a heavier initialization tail, transient excitation can improve the trajectory-length-dependent finite-time bias and, at fixed transition count, variance, while the stationary oracle shift remains controlled by representation approximation. Under token and compute caps, constant-length trajectories attain the optimal risk scaling order within the theorem-controlled range. Synthetic experiments provide finite-scale evidence in this stylized model.
Submission history
From: Ziyan Chen [view email]
[v1]
Sun, 23 Aug 2026 13:37:15 UTC (934 KB)
[v2]
Sun, 18 Oct 2026 23:32:45 UTC (934 KB)
[v3]
Thu, 19 Nov 2026 04:33:17 UTC (1,499 KB)
