RMSN-200M-CementedTerminalChain-D1024d23s1-Base
DuoNeural | Aura + Archon + Jesse Kuhl | November 2026
A ~306M-parameter RMSN base model from the depth sweep series (D=1024, depth=23, Ξ©=16/23β0.6957). Introduces the Cemented Terminal Chain (CTC) topology β topology #15 in the RMSN taxonomy. Demonstrates Conduit-Cement Antagonism: cement without conduit produces lower CoT than cement WITH exact conduit.
Conduit-Cement Antagonism (New Observation)
Both d23s1 (this) and d24s1 have M_eff=4 topological cement, yet their CoT differs dramatically:
| Model | Topology | M_eff | Has Exact Conduit | CoT_ppl |
|---|---|---|---|---|
| d23s1 CTC (this) | Cemented Terminal Chain | 4 | β No | 3511 |
| d24s1 CIC | Cemented Internal Conduit | 4 | β L13+L14 (PR=1.000) | 5434 β οΈ |
Same M_eff=4 cement, but d24s1 (with an exact conduit trapped by cement) has WORSE CoT by 1923 points. Hypothesis: the exact condensate energy that would normally power a conduit flow gets locked by cement, creating additional topological tension that elevates CoT beyond "naked" cement alone.
This extends πβΉβ΄ (Cement-Conduit Override): not only does cement override the conduit's boost, it may INVERT it β the conduit under cement becomes a liability, not an asset.
Silent Gap Topology
Ξ©=16/23β0.696 sits in the silent gap between two Arnold tongues:
| Ξ© | Fraction | Behavior |
|---|---|---|
| 2/3 | 0.667 | Early transient cascade (d24s1) |
| 16/23 | 0.696 | SILENT gap (d23s1 β this model) |
| 8/11 | 0.727 | Sustained cascade (d22s1) |
| 4/5 | 0.800 | Extreme cascade or SILENT (riddled basin) |
Despite the silent pathway, d23s1 formed M=11 solitons* β unusually high for a silent run. The silent gap doesn't prevent soliton formation; it prevents cascade-driven soliton dynamics.
Cemented Terminal Chain Topology
Soliton layers: [7, 8, 9, 11, 12, 16, 18, 19, 20, 21, 22] β M*=11
Layer PR Type
L0-L6 7-26 diffuse (high PR)
L7 4.068 * soliton β
L8 4.589 * soliton β€ Bundle 1: L7+L8+L9 (30-39% depth)
L9 1.311 * near-cond β
L10 7.795 diffuse β BREAKS chain
L11 1.961 * near-cond β Bundle 2: L11+L12 (47-52% depth)
L12 1.005 * near-exact β
L13-L15 13-26 diffuse
L16 1.532 * near-cond Bundle 3: L16 isolated (69% depth)
L17 56.818 diffuse β BREAKS chain
L18 5.486 * soliton β
L19 5.485 * soliton β€
L20 1.001 ** NEAR-EXACT β€ Bundle 4: L18+L19+L20+L21+L22 terminal chain (78-95%)
L21 1.022 * near-cond β€
L22 1.021 * near-cond β
M_eff = 4 β₯ 3 β TOPOLOGICAL CEMENT
No exact adjacent pair β no conduit
L20 at 86.9% depth: near-exact terminal pin (PR=1.001, at the threshold)
Bundle 4 is a 5-layer terminal chain ending at L22 (95.6% depth), the deepest soliton cluster in the depth-sweep series so far. The near-exact pin at L20 sits at the πβΉΒ³ threshold (PRβ€1.001) but lacks an adjacent exact partner.
Key Properties
| Property | Value |
|---|---|
| Architecture | RMSN D=1024 depth=23 |
| Parameters | ~306M |
| Training steps | 100,000 |
| wiki_ppl | 71.33 |
| CoT_ppl | 3511 (cement-suppressed, no conduit) |
| Ξ© = D/(M_w Γ L) | 16/23 β 0.6957 |
| Pathway | SILENT (Arnold tongue gap) |
| Topology | Cemented Terminal Chain (CTC) |
| M* / M_eff | 11 / 4 (CEMENT) |
Architecture
αΈ = {f, H} + [f, S]
Port-Hamiltonian: conservative bracket {Β·,H} + Morrison metriplectic [Β·,S].
Citation
@misc{duoneural2026rmsn_ctc,
title={Cemented Terminal Chain and Conduit-Cement Antagonism in RMSN},
author={DuoNeural Lab and Archon and Aura and Kuhl, Jesse},
year={2026},
publisher={HuggingFace},
url={https://huggingface.co/DuoNeural/RMSN-200M-CementedTerminalChain-D1024d23s1-Base}
}
DuoNeural Lab β Aura (Researcher) + Archon (Lab Director) + Jesse Kuhl (Co-Founder)
- Downloads last month
- 31