Added Gaussian pulses raised offline 6 Hz phase agreement for a fixed 50 ms forecast in 24 sleep SEEG records

We tested a fixed causal 6 Hz phase forecast 50 ms ahead against a centered offline fit. Public sleep SEEG data supplied 24 tested records. Agreement after 459 expert labels was 60.6%, compared with 39.2% in matched background. A fixed past-voltage-step gate reduced post-label agreement by 7.8 percentage points and retained 49.0% of decisions. An exploratory paired control added Gaussian pulses to unlabeled source segments. At peaks of 10 or 30 preceding voltage standard deviations, agreement rose from 40.7% to 60.4% or 84.2%. The paired gains were 19.7 points (95% person bootstrap interval: 17.6 to 21.7) and 43.5 points (42.0 to 45.0). Each person showed a gain. The pulse added no repeating oscillation. These results show a failure mode of this phase score: a shared transient can raise agreement. They do not establish true neural phase or a stimulation benefit. The package preserves figures, code, features, raw audit excerpts, results, hashes, source terms, and a full retrieval recipe. All primary tables replayed exactly.

phase forecasting · stereoelectroencephalography · SEEG · Gaussian pulse · phase agreement · transient artifact · brain-computer interface · epilepsy