✗ Failed on benchmark 2026

Clustered Small-Gain Certificate for Modular Neural Dynamics

Implementation & benchmark of arXiv:2607.16985 — Cluster-Based Distributed Small-Signal Stability Certificates for Grid-Forming Inverter Networks

Usefulness8/10

Difficulty6/10

Novelty6/10

Dynamics Stability Scalability Optimizer Rnn Transformer Training dynamics Control theory Dynamical systems Linear algebra Graph theory Optimization

Source paper: Cluster-Based Distributed Small-Signal Stability Certificates for Grid-Forming Inverter Networks arXiv:2607.16985 ⓘ · analyzed Aug 30, 2026

AI-generated research hypothesis, automatically tested. Not peer-reviewed.

Idea description

Treat neural modules as interconnected dynamical subsystems and estimate the gain from every module input to every neighboring module output. Replace an expensive global Jacobian spectral-radius calculation by decentralized directed-cycle tests inside clusters and path-gain tests between clusters. Penalizing violations during training should prevent exploding recurrent trajectories while retaining less conservative behavior than constraining every individual block independently.

Formulas

$$z_i^{+}=f_i(z_i,u_i),\qquad u_i=\sum_{j\neq i}G_{ij}z_j,\qquad \|\delta y_i\|\leq\sum_j\gamma_{ij}\|\delta y_j\|,$$

$$\Gamma=[\gamma_{ij}],\qquad \prod_{(j\rightarrow i)\in C}\gamma_{ij}<1\quad\text{for every directed feedback cycle }C,$$

$$\mu(d):=\max_i\frac{1}{d_i}\sum_j\gamma_{ij}d_j<1,\qquad d_i>0,$$

$$\mathcal{L}_{\mathrm{sg}}=\sum_{C}\operatorname{softplus}\!\left(\log\prod_{e\in C}\gamma_e-\log(1-\varepsilon)\right)$$

Mathematical statement

Let modules be indexed by nodes i in V, with perturbation dynamics z_i and stable local input-output maps whose induced gain from module j to module i is gamma_ij >= 0. The gain graph has an edge j -> i when gamma_ij > 0. For a directed cycle C, the cyclic small-gain condition is the product of gamma_ij over all edges in C being less than one. A sufficient global certificate is obtained by finding positive weights d_i such that the scaled gain matrix has maximum row sum below one: mu(d)=max_i d_i^{-1} sum_j gamma_ij d_j < 1. Here gamma_ij is estimated from block Jacobians or empirical perturbation responses, d_i are positive scaling weights, and 1-mu(d) is the certified contraction reserve.

Implementation notes

Use a modular recurrent network, residual network, or transformer in which each block is a subsystem and cross-block communication is an edge. At a minibatch point, compute block Jacobian-vector products J_ij v_j with automatic differentiation, using two to four Hutchinson or power-iteration vectors per source block. Estimate gamma_ij as a high quantile of ||J_ij v_j||/||v_j||, optionally multiplied by a safety factor of 1.2. Construct the sparse nonnegative gain matrix Gamma. Enumerate short directed cycles within each selected cluster; for longer graphs, approximate the worst cycle using log-gain search or optimize positive weights d with gradient descent on mu(d). Add the cycle barrier to the task loss. Pseudocode: sample perturbation vectors; obtain block JVPs; update gamma_ij with an exponential moving maximum; compute cycle products and scaled row sums; add the certificate penalty; perform the ordinary optimizer step. The paper-derived quantities are cycle and path small-gain inequalities; empirical quantities are Jacobian gains, clustering, and d. First experiment: compare a four- to eight-block GRU or residual MLP on sequential MNIST and a synthetic long-horizon copy task against spectral normalization and an unconstrained baseline. Sweep the penalty coefficient and measure trajectory norms over 1,000 steps. The prediction is a sharp divergence boundary: when the estimated worst cycle product crosses 1, long-horizon perturbations should stop decaying and begin growing. Certified runs should have empirical per-step amplification below 1, with the measured boundary agreeing with the gain estimate within approximately 20%.

Verification

Failed on benchmark

Stage 1 · Toy mechanism gate: Passed ✓

Stage 2 · Mechanism transferred to benchmark: Not reproduced ✗

Stage 2 · Practical benchmark result: Significantly worse ✗

Methodology: 8 paired seeds, permutation test (p<0.05) against a learning-rate-tuned baseline. How verification works

Stage 1 — Mechanism check agent confidence 8/10

Built and numerically verified a clustered small-gain certificate for modular dynamics. The symmetric-cycle boundary was observed at q=1.00 versus the predicted 1.00; in the 4-node sweep, positive trajectory growth first appeared at scale 1.05 versus predicted spectral-radius scale 1.012 (3.7% grid/tolerance error). Path gain scaling matched log(q) to machine precision, and optimized positive scaling exactly recovered the 2-cycle spectral radius while unscaled row sums were more conservative. At an unstable scale, clustered attenuation reduced rho to 0.907 and final norm to 7.5e-6, versus rho 0.991 and final norm 0.249 for independent edge clipping, while changing parameters much less on average (0.0152 vs 0.3875 in the analytic comparison).

Agent confidence: 8/10

Baseline: Unconstrained: rho=1.1855, mu_d=1.2021, final norm=6.56e8. Independent edge clip: rho=0.9908, mu_d=0.9972, final norm=0.2488; mean parameter change=0.3875.
Idea: Clustered certificate: rho=0.9072, mu_d=0.9226, final norm=7.50e-6; mean parameter change=0.0152. Predicted cycle boundary q=1.0, observed=1.0; predicted 4-node scale=1.012, observed positive-growth grid point=1.05; predicted path slope log(0.82)=-0.198451, observed=-0.198451.

Limitations: The main experiment uses a linear 4-module system with exact absolute Jacobian gains, not a trained GRU/residual MLP. It does not implement minibatch autograd JVP estimation, EMA gain tracking, cycle enumeration for arbitrary sparse graphs, task-loss optimization, FLOP accounting, or sequential-MNIST/copy-task generalization. The clustered intervention uses common attenuation rather than fully learning per-module scaling weights.

How to run: python3 clustered_small_gain.py && /home/maxwelhelp/main/bin/python3 run_experiment.py

Stage 2 — Benchmark vs. tuned baseline (arXiv:2607.16985) Sep 2, 2026

Significantly WORSE than the tuned baseline (+2.7%, p=0.0149, wins 1 / 8 paired seeds).

Benchmark: Dynamics forecasting
Model: rnn_small
Paired seeds: 8
Baseline mean: 0.0002
Idea mean: 0.0002
Effect (Δ): 0 (+2.7%; negative = idea better)
Wins: 1 / 8 paired seeds
p-value: 0.0149 (permutation test, 20 000 shuffles)
Smallest detectable effect: n/a
Mechanism: Not confirmed ✗
Practical effect: Hurts
Baseline tuning: swept over 0 configs

Limitations:

Only the built-in dynamics track was tested. The gain estimate uses four fixed hidden-state clusters and GRU recurrent-weight block norms rather than minibatch JVPs, EMA tracking, arbitrary sparse cycle enumeration, or learned positive scaling weights. The behavioral probe measures initial-hidden-state perturbation through an 8-step GRU rather than 1,000-step trajectory growth.

How to run: python3 stage2_clustered_gain.py

Verdict computed by deterministic test code from paired-seed statistics — not by the language model.

Stage 2 — Benchmark Aug 30, 2026 latest Failed ✗

Agent confidence: 9/10

Implemented the clustered small-gain certificate as a training-loss intervention for the structurally matched dynamics/rnn_small benchmark, with a spectral-penalty baseline, equal learning-rate/strength sweeps, and 8 paired seeds. The certificate sanity check passed exactly, but the idea significantly worsened test MSE: delta_mean=+6.33625e-06 with permutation p=0.01485. The trained-model mechanism signature was not confirmed because estimated gain averaged 4.79564 while observed hidden-state amplification averaged 0.0284937.

Baseline: Best cfg lr=0.01, strength=0.0; full 8-seed MSE mean=0.0002354654107, std=0.0001008976129.
Comparison: Paired delta idea-baseline=+0.0000063362477; idea wins 1/8 seeds; permutation p=0.01485; verdict=idea worse (significant).
Idea: Best cfg lr=0.01, strength=0.0001; full 8-seed MSE mean=0.0002418016584, std=0.0001006422367.
Mechanism signature: Predicted gain mean=4.7956403; observed amplification mean=0.02849366; relative error=0.9940584; confirmed=false.
Sanity check: Two-cycle optimized scaled row sum=0.8000000000000002 versus predicted sqrt(1.6*0.4)=0.8; path log-gain slope matched log(0.82) to approximately 6e-17.

Benchmark result