No clear winner. The survival gap is under 10 percentage points and the balanced-score gap is under 0.05, so neither SolvSRK nor the best baseline clears the win threshold. Either works — choose on cost, licensing, or integration effort. All verdicts →
10D Wendling model with fast/slow inhibition with additive noise σ=0.0001. Noise regime: sub-clinical. Base problem: NMM.4. Runner injects noise via acceptance_criteria.noise_sigma.
Wendling et al. (2002); Hebbink et al. (2020)
Canonical RHS excerpt from the registered callable used for this benchmark cell. Expand it to verify the state equations; it is not a standalone runnable fixture.
def _stable_expit(x):
"""Numerically stable sigmoid 1/(1+exp(-x)), scalar or array."""
x = np.asarray(x, dtype=float)
return np.where(
x >= 0,
1.0 / (1.0 + np.exp(-x)),
np.exp(x) / (1.0 + np.exp(x)),
)
def _jr_sigmoid(v, v_max=5.0, r=0.56, v0=6.0):
"""Jansen-Rit population sigmoid: S(v) = v_max / (1 + exp(r*(v0 - v)))."""
return v_max * _stable_expit(r * (v - v0))
def rhs(t: float, y: np.ndarray) -> np.ndarray:
y0v, y1v, y2v, y3v, y4v = y[0], y[1], y[2], y[3], y[4]
y5v, y6v, y7v, y8v, y9v = y[5], y[6], y[7], y[8], y[9]
S_pyr = _jr_sigmoid(y1v - y2v - y3v)
S_exc = _jr_sigmoid(C1 * y0v)
S_slow = _jr_sigmoid(C3 * y0v)
S_fast = _jr_sigmoid(C5 * y0v - C6 * y4v)
S_self = _jr_sigmoid(C3 * y0v)
return np.array([
y5v, # ẏ₀
y6v, # ẏ₁
y7v, # ẏ₂
y8v, # ẏ₃
y9v, # ẏ₄
A * a * S_pyr - 2.0 * a * y5v - a2 * y0v, # ẏ₅
A * a * (p + C2 * S_exc) - 2.0 * a * y6v - a2 * y1v, # ẏ₆
B * b * C4 * S_slow - 2.0 * b * y7v - b2 * y2v, # ẏ₇
G * g * C7 * S_fast - 2.0 * g * y8v - g2 * y3v, # ẏ₈
B * b * S_self - 2.0 * b * y9v - b2 * y4v, # ẏ₉
])Canonical RHS excerpt captured from the same registered callable used for the published benchmark. Frozen closure values are summarized below; helper imports and solver settings are intentionally omitted.
Fingerprint
Spread: high
Default noise: medium
Recommendation snapshot
Clean best: SolvSRK
Noisy best: SciPy BDF
Coverage
14 solver arms · clean + 5 noise levels
Ranked on survival, precision, and speed
20 seeds/cell default · 14 arms · TRL 4–5 · simulation-lab validated · this page: Wendling extended — noise σ=0.0001 (sub-clinical) (wendling-extended-noise-0-0001-sub-clinical)
Governed SolvTune benchmark freeze; per-arm medians only. RHS definitions and raw trial rows are not published.
Self-reported by Resonix Labs · not independently verified
Pick an objective and a noise level to rank all arms on survival, median SCD, median nfev, and median wall time. Medians across seeds.
Objective
Best overall trade-off of survival, precision, and speed.
Noise level
| # | Solver | Survival | SCD | nfev | Wall | Score |
|---|---|---|---|---|---|---|
| 1 | SolvSRK | 100% | 9.0 | 23,399 | 368 ms | 0.833 |
| 2 | Vern7external | 100% | 4.4 | 798,062 | 28.69 s | 0.725 |
| 3 | SciPy RK23SciPy | 100% | 4.2 | 91,313 | 2.00 s | 0.718 |
| 4 | SciPy RK45SciPy | 100% | 4.2 | 65,756 | 1.58 s | 0.718 |
| 5 | Tsit5external | 100% | 3.3 | 142,044 | 11.68 s | 0.698 |
| — | SciPy BDFSciPy | 0% | — | — | — | — |
| — | SciPy RadauSciPy | 0% | — | — | — | — |
| — | SciPy LSODASciPy | 0% | — | — | — | — |
| — | SciPy DOP853SciPy | 0% | — | — | — | — |
| — | CVODE BDFexternal | 0% | — | — | — | — |
| — | CVODE Adamsexternal | 0% | — | — | — | — |
| — | Vern9external | 0% | — | — | — | — |
| — | TRBDF2external | 0% | — | — | — | — |
| — | FBDFexternal | 0% | — | — | — | — |
At Clean, best balanced arm is SolvSRK.
Values are medians across seeds, measured by Resonix Labs on Resonix hardware and not independently verified; nfev and wall are on reference lab hardware (indicative). Under injected noise only SolvSRK and the SciPy arms are run. How we measure accuracy → · Verification status →
SolvScout · free
This page shows one published benchmark cell. SolvScout fingerprints your ODE, compares it to the full corpus, and recommends a solver with the same survival / precision / speed ranking you see here — including when a SciPy arm wins.
SolvSRK · 30-day trial
SolvSRK is the stiffness-adaptive integrator behind the SolvSRK column in these tables. Create an account, activate a machine, and take a 30-day trial — same binary you'd ship after purchase.
Replace the access date. Pin the freeze ID and library versions when comparing against a later export. Cite it as what it is — a self-reported vendor benchmark, not an independently verified result. The note field says so; please keep it.
@misc{resonix_evidence_wendling_extended_noise_0_0001_sub_clinical_2026,
title = {Resonix Evidence Portal: Wendling extended — noise σ=0.0001 (sub-clinical)},
author = {{Resonix Labs (Canada) Inc.}},
year = {2026},
howpublished = {\url{https://resonix.tech/evidence/problems/wendling-extended-noise-0-0001-sub-clinical}},
note = {Self-reported vendor benchmark; internally generated by Resonix Labs and not independently verified. Accessed YYYY-MM-DD. Freeze 2026-08-13; libsolvsrk 2.3.0; SciPy 1.14.}
}TRL 4–5 · simulation-lab validated · 398 problems · 14 solver arms · clean + 5 noise levels
Freeze: 2026-08-13 · scipy 1.14 · libsolvsrk 2.3.0 · Methodology
Self-reported by Resonix Labs · not independently verified · Verification status