✗ Failed on benchmark 2026

Contractive Latent Observer

Implementation & benchmark of arXiv:2607.14885 — Contraction versus Recurrence: An Exponential Separation in Observation-Based Prediction of Deterministic Dynamics

Usefulness8/10

Difficulty5/10

Novelty6/10

Architecture Stability Accuracy Speedup Rnn Ssm World model Training dynamics Dynamical systems Control theory Spectral theory Linear algebra

Source paper: Contraction versus Recurrence: An Exponential Separation in Observation-Based Prediction of Deterministic Dynamics arXiv:2607.14885 ⓘ · analyzed Aug 30, 2026

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 Not peer-reviewed.

Idea description

Replace recurrence or nearest-neighbour analogue lookup with a learned delay-coordinate observer that continuously corrects a latent state using the current observation. Constrain the observer's closed-loop Jacobian or linear state matrix to have spectral radius below one, so prediction error contracts geometrically and required burn-in grows logarithmically with target accuracy.

Formulas

$$\hat z_{t+1}=A\hat z_t+L(y_t-C\hat z_t),\qquad A_{cl}=A-LC.$$

$$e_{t+k}=A_{cl}^{k}e_t,\qquad \|e_{t+k}\|\le K\rho(A_{cl})^k\|e_t\|,\quad \rho(A_{cl})<1.$$

$$k_\varepsilon\approx \frac{\log(K\|e_0\|/\varepsilon)}{-\log\rho(A_{cl})}\simeq \frac{2\log(K\|e_0\|/\varepsilon)}{1-\rho(A_{cl})^2}.$$

$$\mathcal{L}=\mathcal{L}_{pred}+\lambda\max\left(0,\rho_{est}(J_t)-\gamma\right)^2,\qquad 0<\gamma<1.$$

Mathematical statement

Let $y_t$ be the observed signal, $z_t\in\mathbb{R}^m$ the latent state, $A$ the latent transition matrix, $C$ the observation matrix, and $L$ the observer gain. The observer is $\hat z_{t+1}=A\hat z_t+L(y_t-C\hat z_t)$. Its estimation error satisfies $e_{t+1}=A_{cl}e_t$, where $A_{cl}=A-LC$. If the closed-loop spectral radius $\rho(A_{cl})<1$, then $\|e_k\|\le K\rho(A_{cl})^k\|e_0\|$ asymptotically. Therefore the number of steps to reach tolerance $\varepsilon$ is logarithmic in $1/\varepsilon$, with the paper's near-unit-radius approximation proportional to $(1-\rho(A_{cl})^2)^{-1}$.

Implementation notes

Use a delay window $y_{t-p+1:t}$ as input to a recurrent latent model. Implement a prior transition $z^{prior}_{t+1}=f_\theta(z_t)$, decode $\hat y_t=g_\theta(z_t)$, and apply the observer correction $z_{t+1}=z^{prior}_{t+1}+K_\phi(y_t-\hat y_t)$. For the cheapest MVP, use $f_\theta(z)=Az$ and $g_\theta(z)=Cz$, and learn $A,C,K$ with one-step and multi-step prediction loss. At every training step estimate $\rho(A-KC)$ by power iteration and add $\lambda\operatorname{ReLU}(\rho-\gamma)^2$. For a nonlinear transition, estimate Jacobian-vector products with autodiff and penalize the largest sampled singular value. The contraction inequality comes from the paper; parameters, Jacobians, and residual noise floors are estimated from data. First test on Lorenz-63 with scalar observations and the Santa Fe laser benchmark, comparing analogue retrieval, an unconstrained GRU, and the observer. Sweep target tolerance from $10^{-2}$ to $10^{-6}$. The predicted signature is a straight line of burn-in versus $\log(1/\varepsilon)$, with slope increasing approximately as $(1-\rho^2)^{-1}$; measured divergence should occur when $\rho\ge 1$.

Verification

Failed on benchmark

Stage 1 · Toy mechanism gate: Passed ✓

Stage 2 · Mechanism transferred to benchmark: Not reproduced ✗

Stage 2 · Practical benchmark result: No measurable effect

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

Stage 1 — Mechanism check agent confidence 9/10

Built a linear contractive latent observer with spectral-radius penalty, Lorenz-63 scalar-observation forecasting, GRU baseline, and analogue retrieval baseline. The toy checks confirmed the predicted rho=1 contraction boundary exactly, burn-in slopes matched 1/(-log rho) within 0.05 relative error, and scalar gain sweeps matched rho(A-LC) exactly. The observer did not win on Lorenz one-step MSE in this small run: it was worse than both GRU and analogue retrieval, so the mechanism is real but an application-level improvement was not demonstrated.

Agent confidence: 9/10

Baseline: GRU MSE=0.00020846; analogue retrieval MSE=0.00013584
Idea: Observer MSE=0.00175154; learned rho(A-KC)=0.47528. Toy burn-in slope relative errors: rho=[0.20,0.50,0.80,0.95,0.99] -> [4.85%,0.66%,0.18%,0.24%,0.05%]; boundary matched contraction predictions for all tested rho values; gain sweep matched exact scalar prediction.

Limitations: Only a linear one-dimensional observer and one-step Lorenz-63 scalar forecast were tested. Santa Fe laser, nonlinear Jacobian constraints, multi-step rollout loss, repeated seeds, statistical significance, runtime/FLOP comparisons, and tolerance burn-in on noisy learned observers were not tested. CUDA was attempted but cuDNN host allocation failed, so the reported mini-experiment used the required CPU fallback.

How to run: python3 experiment.py

Stage 2 — Benchmark vs. tuned baseline (arXiv:2607.14885) Aug 30, 2026

No significant difference vs. the tuned baseline (+0.0%, p=0.9359, wins 3 / 8 paired seeds).

Benchmark: Dynamics forecasting
Model: rnn_small
Paired seeds: 8
Baseline mean: 0.001
Idea mean: 0.001
Effect (Δ): 0 (+0.0%; negative = idea better)
Wins: 3 / 8 paired seeds
p-value: 0.9359 (permutation test, 20 000 shuffles)
Smallest detectable effect: ±0.3%
Mechanism: Not confirmed ✗
Practical effect: No effect
Baseline tuning: swept over 3 configs

Limitations:

The benchmark used the built-in 8-step actuated-pendulum dynamics track and a linear recurrent-weight spectral-norm proxy rather than an explicit learned A-KC observer with separately parameterized latent observation gain. No Santa Fe laser, Lorenz, nonlinear Jacobian constraint, rollout-horizon sweep, runtime/FLOP comparison, or GPU run was included; CUDA was disabled after the shared-GPU process aborted.

How to run: CUDA_VISIBLE_DEVICES='' /home/maxwelhelp/main/bin/python3 bench_stage2.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

Built a local Contractive Latent Observer intervention on the structurally matched dynamics track using the shared rnn_small-sized GRU and a differentiable spectral-norm proxy penalty. The geometric contraction sanity check passed, and the trained-model proxy averaged 0.8405 versus the target 0.90, confirming the intended stability signal. However, test MSE was very slightly worse than the tuned baseline (delta +9.42e-8, permutation p=0.9359), so there is no observed benchmark improvement.

Baseline: dynamics/rnn_small, tuned lr=0.006; 8-seed test MSE mean=0.001037391914, std=0.000183358837
Idea: contractive penalty, tuned lr=0.006; 8-seed test MSE mean=0.001037486116, std=0.000184032990; delta=+9.42018e-8; p=0.9359; trained rho proxy mean=0.840508 < gamma=0.90

Benchmark result