Pre-registration (frozen 5 Nov 2026): replication of the Time-o1 loss improvement for Fredformer on ETTh1 and ETTh2 (arXiv 2505.17847), as published and with a matched learning-rate search for the baseline
Year: 2026
Type: article
Abstract: Frozen protocol, decision code, job design, harness, planted-effect table and kernel for a private pre-registered replication, deposited before any confirmatory run. Clauses: the median per-cell test-MSE improvement of the Time-o1 loss over the MSE-trained Fredformer has a 95% bootstrap lower end above 1.8% (about half the paper's mean) and at least 7 of 8 dataset-horizon cells improve, for the as-published arm and, separately, an arm in which the baseline receives a three-value learning-rate search. Outcome wording: Replicated / Did not replicate / Inconclusive under protocol X.
Author: Goodman, John
Language: English
Subfield: Signal Processing
Field: Computer Science
Domain: Physical Sciences
Open Access status: green
OpenAlex ID: W7225328347