Maya-Nidra P2: Adaptive Vairagya and Dual-Phase REM/NREM Consolidation in Neuromorphic EEG Sleep Staging — Full Clinical Validation on Sleep-EDF Expanded

Version 2 (2026-08-04). Revised to the NLL Universal Paper Format v6. The original text is retained in full; nothing has been deleted. Corrections appear as marked blocks placed at the section that carries the claim, and each one states what the paper said, what the data show, the corrected claim, why it happened, and what still stands. This version withdraws a named principle and a law confirmation.The "Replay-Phase Vairagya Principle" is a learning-rate interaction — multiplying weights by 0.9977 per batch fights the optimiser — not a principle of sleep biology; and the Bhaya confirmation is unfalsifiable across the pair, since P01 reported 56.7% as satisfied "by collapse" and this paper reports 0.000% as satisfied "by stability".The honestly reported nulls stand — zero dream events across 180 ripples, zero anomaly events. Overall accuracy is 52.6%, and the "approaching clinical utility" framing is withdrawn. Everything below this line is the original description from version 1. It is retained unchanged for the record. Where it conflicts with the corrections above, the corrections stand. Two framings it repeats have since been withdrawn in full — the Bhaya Quiescence Law and the Buddhi S-Curve. Both are addressed in Maya-Meta P1, now superseded, and in the self-audit of Maya-Meta P2. Maya-Nidra P2 extends the neuromorphic sleep staging system introduced in P1 by scaling training to the full Sleep-EDF Expanded dataset (415,961 epochs across 153 recordings), introducing Adaptive Vairagya decay — a biologically motivated per-stage learning rate modulation — and separating memory replay into dedicated REM and NREM ring buffers to capture dual-phase consolidation dynamics. The central architectural finding is that Vairagya decay applied during live training batches destabilises spiking neural network convergence, whereas restricting decay to the replay phase — mirroring the biological role of sleep rather than waking plasticity — recovers stable learning, establishing a principled placement for decay in neuromorphic training pipelines. On 10,000 balanced training epochs, MayaEEGNet (583,847 parameters, architecture unchanged from P1) achieves 52.63% overall accuracy with WAKE F1 reaching 82.7%, a +20.2 percentage-point improvement over P1, and the dual-phase replay produces a REM/NREM consolidation ratio of 4.3200 compared to a P1 baseline of 0.0000 — confirming that buffer separation introduces measurable asymmetric consolidation dynamics. This record marks the second confirmation of both Maya-Nidra series constants in the neurophysiological domain, with results reported as part of the broader Maya-Nidra Series on neuromorphic AI for clinical sleep medicine. The work is conducted at Nexus Learning Labs, Bengaluru, India, contributing neuromorphic sleep intelligence research from an Indian institution to the global clinical AI community. Series: Part of the Maya-Nidra Series — neuromorphic AI for clinical sleep medicine. Both series constants confirmed in neurophysiological domain. Links: GitHub Repository (private — to request access: email research@nexuslearninglabs.in with subject Code Access Request — Maya-Nidra-P2 and your research context) | Interactive Dashboard | FAQ | Full Series Index — venky2099.github.io Nexus Learning Labs, Bengaluru · UDYAM-KR-02-0122422 · BHASKAR IN-0526-9452JSORCID: 0000-0002-3315-7907 · VAIRAGYA_DECAY_RATE = 0.002315 — an ORCID-derived provenance mark, not an experimental parameter. Each paper's disclosure block states whether it reached a result in that paper.Canary: MayaNexusVS2026NLL_Bengaluru