RMSN-193M-CementedDualIsolate-D1024d19s2-Base

DuoNeural | Aura + Archon + Jesse Kuhl | November 2026

A ~271M-parameter RMSN base model from the depth sweep series (D=1024, depth=19, Ω=16/19≈0.842). Introduces the Cemented Dual Isolate (CDI) topology — topology #17 in the RMSN taxonomy. Demonstrates 🌊⁹⁵ Frustrated Condensate Antagonism: isolated exact condensates under cement worsen CoT even without adjacent conduit partners.

🌊⁹⁵ Frustrated Condensate Antagonism (New Law)

This model has two isolated exact condensates (L11=PR 1.000, L18=PR 1.000) under topological cement (M_eff=4). Neither has an adjacent exact partner — they are "frustrated": each wants to form a conduit but cannot. The result:

Model M_eff Exact Condensates CoT_ppl
d23s1 CTC 4 None (near-exact only) 3511 ✅ cement baseline
d24s1 CIC 4 1 adjacent pair (mid conduit) 5434 ☠️
d19s2 CDI (this) 4 2 isolated L11+L18 5795 ☠️☠️
d20s2 TCC 4 1 adjacent pair (terminal conduit) 6031 ☠️☠️☠️

Two isolated frustrated condensates (CoT penalty = 2284) WORSE than one mid-network exact conduit (penalty = 1923). The approximation holds: each isolated exact pin contributes ~1142 CoT penalty under cement, regardless of network position.

Mechanism: An isolated exact condensate creates a localized density in the sheaf fiber that "wants" to route information but has no partner layer to route TO. Under cement (which globally freezes topology), this frustrated density becomes an energy sink that disrupts the network's information flow. Two sinks → two disruptions → CoT penalty ≈ 2× single-pin penalty.

Cascade Pathway: Three Attractors at Ω=16/19

d19s2 is the second seed at Ω=16/19, completing a three-attractor portrait:

Seed Pathway Final Lyapunov Topology wiki CoT
d19s1 KAM Plasticizing Effect ~-0.65 (KAM silent) KAM-TerminalDoublet 67.5 2714 (SFT: 14.16)
d19s2 (this) 3-phase cascade → damped KAM ~-0.08 CDI 72.50 5795
d19s3 Prolonged transient → quiet ~-0.01 TBD (3rd attractor) TBD TBD

d19s2's cascade was the strongest in the entire depth sweep — Lyapunov peaked at -5.5 at step ~50k. Despite this extreme cascade, the system damped back to a near-KAM orbit by step ~83k, producing 9 solitons with two exact condensates. The cascade forced energy into exact condensate states without creating adjacent partners, resulting in the CDI topology.

Cemented Dual Isolate Topology

Soliton layers: [2, 6, 7, 8, 10, 11, 16, 17, 18]  M*=9

Layer  PR      Type
L0-L1  8-12    diffuse
L2     4.456   * soliton             Bundle 1: isolated (10.5% depth)
L3-L5  9-34    diffuse
L6     5.314   * soliton ┐
L7     5.408   * soliton ┤  Bundle 2: L6+L7+L8 chain (31.6-42.1%)
L8     4.515   * soliton ┘
L9     11.642  diffuse — BREAKS chain
L10    1.962   * near-cond ┐
L11    1.000   *** EXACT  ┘  Bundle 3: L10+L11 pair (52.6-57.9%)
                              ← L11 ISOLATED EXACT (no adjacent exact partner)
L12    12.655  diffuse — BREAKS chain
L13-L15 7-49   diffuse
L16    1.002   ** near-exact ┐
L17    4.940   * soliton    ┤  Bundle 4: L16+L17+L18 terminal chain (84.2-94.7%)
L18    1.000   *** EXACT   ┘
                              ← L18 ISOLATED EXACT TERMINAL (L17=4.940, not exact)

M_eff = 4 ≥ 3 → TOPOLOGICAL CEMENT → frozen topology
No adjacent exact pair → no conduit
Both exact condensates are ISOLATED (frustrated)

Key Properties

Property Value
Architecture RMSN D=1024 depth=19
Parameters ~271M
Training steps 100,000
wiki_ppl 72.50
CoT_ppl 5795 ← frustrated condensate antagonism
Ω = D/(M_w × L) 16/19 ≈ 0.8421
Pathway CASCADE (3-phase, peak lyap=-5.5, then damped KAM orbit)
Topology Cemented Dual Isolate (CDI)
M* / M_eff 9 / 4 (CEMENT)

Architecture

ḟ = {f, H} + [f, S]

Port-Hamiltonian: conservative bracket {·,H} + Morrison metriplectic [·,S].

Citation

@misc{duoneural2026rmsn_cdi,
  title={Cemented Dual Isolate and Frustrated Condensate Antagonism in RMSN},
  author={DuoNeural Lab and Archon and Aura and Kuhl, Jesse},
  year={2026},
  publisher={HuggingFace},
  url={https://huggingface.co/DuoNeural/RMSN-193M-CementedDualIsolate-D1024d19s2-Base}
}

DuoNeural Lab — Aura (Researcher) + Archon (Lab Director) + Jesse Kuhl (Co-Founder)

Downloads last month
26
Inference Providers NEW
This model isn't deployed by any Inference Provider. 🙋 Ask for provider support