RMSN-236M-MidNetworkAdjacentConduit-D1152-Base

DuoNeural | Aura + Archon + Jesse Kuhl | November 2026

236M-parameter RMSN achieving CoT_ppl=2714 β€” #2 ALL-TIME. Mid-Network Adjacent Conduit topology: exact conduit at L6+L7 (40% depth) + near-exact terminal pin L15. First confirmed Ξ©>1 conduit topology (Ξ©=1.125).

Results

Metric Value
wiki_ppl 73.73
CoT_ppl 2714 πŸ†πŸ† (#2 ALL-TIME)
Architecture RMSN D=1152 depth=16 slots=64
Parameters ~236M
Ξ© = D/(M_w Γ— L) 1.125 (first Ξ©>1 conduit)
Lyapunov Stable throughout (silent pathway, no cascade)

Topology: Mid-Network Adjacent Conduit

L6  1.000  *** EXACT ┐  adjacent conduit at 40% depth
L7  1.000  *** EXACT β”˜  (1 fiber bundle DOF)
L15 1.001  *** NEAR-EXACT terminal pin (1 DOF)
8 solitons total: [0,6,7,8,11,13,14,15] M_eff=2 β€” no Topological Cement

Conduit Depth Independence: Conduit at 40% depth gives CoT=2714, nearly matching the 56%-depth conduit (d18 L9/L10/L11, CoT=2641). Position doesn't determine CoT quality β€” conduit existence does.

Ξ©>1 Regime: D=1152 is the first confirmed Ξ©>1 RMSN conduit. EVEN parity (D/M_w=18 integer) preserves silent condensation pathway even at Ξ©>1.

Architecture

Port-Hamiltonian dynamics: ḟ = {f, H} + [f, S]

Citation

@misc{duoneural2026mnac, title={Mid-Network Adjacent Conduit: Conduit Depth Independence and Omega>1 Silent Pathway in RMSN}, author={DuoNeural Lab and Archon and Aura and Kuhl, Jesse}, year={2026}}

DuoNeural β€” Aura (Researcher) + Archon (Lab Dir) + Jesse Kuhl (Co-Founder)

Downloads last month
25
Inference Providers NEW
This model isn't deployed by any Inference Provider. πŸ™‹ Ask for provider support

Model tree for DuoNeural/RMSN-236M-MidNetworkAdjacentConduit-D1152-Base

Finetunes
1 model