✗ Mechanism failed 2026

Closure-Decorrelation Memory Scheduler

Implementation & benchmark of arXiv:2608.18148 — Flux-form spatiotemporal neural operators for coarse-grained dynamics of multiscale PDEs

Usefulness7/10

Difficulty4/10

Novelty7/10

Training Accuracy Stability Speedup Ssm Rnn Attention Training dynamics Scheduler Pde Statistics Stochastic processes Dynamical systems

Source paper: Flux-form spatiotemporal neural operators for coarse-grained dynamics of multiscale PDEs arXiv:2608.18148 ⓘ · analyzed Sep 1, 2026

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 Not peer-reviewed.

Idea description

Choose the neural operator's input-history length from the measured correlation time of the unresolved closure signal produced by coarse-graining. This avoids under-memory, which causes systematic closure error, and over-memory, which increases attention cost and can destabilize training. The same diagnostic can drive adaptive memory truncation across physical regimes.

Formulas

$$r(t)=\mathcal C\mathcal L(u(t))-\mathcal L_{\mathrm{res}}(\bar u(t)),\qquad\bar u(t)=\mathcal C u(t).$$

$$\rho(\tau)=\frac{\mathbb E_t[\langle r(t),r(t+\tau)\rangle_\Omega]}{\mathbb E_t[\|r(t)\|_\Omega^2]},\qquad\|r\|_\Omega^2=\int_\Omega\|r(x)\|^2dx.$$

$$T_{\mathrm{in}}=\min\{\tau:|\rho(s)|\leq\varepsilon\text{ for all }s\in[\tau,\tau+\Delta]\}.$$

$$\tau_c=\int_0^\infty|\rho(\tau)|d\tau,\qquad T_{\mathrm{in}}=\alpha\tau_c.$$

Mathematical statement

Let $\mathcal C$ be a linear coarse-graining map, $\bar u=\mathcal C u$ the resolved field, and $\mathcal L$ the fine-scale PDE generator. Define the filtered closure-injection signal $r(t)=\mathcal C\mathcal L(u(t))-\mathcal L_{\mathrm{res}}(\bar u(t))$, where $\mathcal L_{\mathrm{res}}$ is the resolved-only drift obtained by inserting $\bar u$ into the coarse PDE. The unresolved contribution $r(t)$ is the forcing that a memory model must represent. Estimate its normalized autocorrelation $\rho(\tau)=\mathbb E[\langle r(t),r(t+\tau)\rangle]/\mathbb E[\|r(t)\|^2]$. Select $T_{\mathrm{in}}$ as the first lag at which $|\rho(\tau)|\leq\varepsilon$ and remains below the threshold over a persistence interval, or use the integrated correlation time $\tau_c=\int_0^\infty|\rho(\tau)|d\tau$ with safety factor $\alpha$. The mechanism predicts diminishing returns once the history exceeds the closure decorrelation time.

Implementation notes

Use the selected window length to set the number of lag tokens in a causal attention operator, temporal convolution, recurrent state, or neural state-space model. For each high-resolution trajectory, compute the resolved field and the closure injection. Pseudocode is: u_bar[t] = coarse(u_fine[t]); r[t] = coarse(fine_rhs(u_fine[t])) - resolved_rhs(u_bar[t]); rho[k] = mean(dot(r[t], r[t+k])) / mean(norm(r[t])**2); choose the first persistent threshold crossing; m = ceil(T_in / dt); train the model using the previous $m$ slices only. The closure-residual definition and correlation-time criterion are obtained from the paper's coarse-graining mechanism; the threshold $\varepsilon$, persistence interval $\Delta$, and safety factor $\alpha$ must be estimated or cross-validated. If fine-scale right-hand sides are unavailable, estimate them by finite differences of high-resolution trajectories and fit the resolved drift separately. The first cheap experiment should use filtered Kuramoto–Sivashinsky or Burgers data and compare windows $m\in\{1,2,4,8,16,32\}$ with the diagnostic-predicted $m_*$. Measure one-step loss, 100-step rollout error, long-run spectra, and inference cost. The predicted signature is an elbow: error decreases while $T_{\mathrm{in}}<T_c$, then plateaus once $|\rho(T_{\mathrm{in}})|<\varepsilon$; the optimal window should lie within about 20 percent of the first persistent correlation crossing.

Verification

Mechanism failed

Stage 1 · Toy mechanism gate: Failed ✗

Stage 2 · Mechanism transferred to benchmark: Not tested

Stage 2 · Practical benchmark result: Not run

Methodology: Toy-system gate first; the benchmark stage runs only after a pass. How verification works

Stage 1 — Mechanism check agent confidence 9/10

Built and numerically verified a closure-decorrelation scheduler on a synthetic unresolved AR(1) closure process with noisy resolved observations. The persistent threshold-crossing prediction matched theory exactly across all five correlation parameters, with 0-step MAE. However, scheduler-selected memory did not reliably match the forecasting elbow, so the claimed mechanism was not sufficiently confirmed.

Agent confidence: 9/10

Baseline: Fixed m=1: mean one-step MSE 1.10295 and mean 30-step rollout MSE 1.40463.
Idea: Diagnostic-selected memory had mean m=9.2, mean one-step MSE 1.07141, a 2.86% improvement over m=1, and mean rollout MSE 1.38282. Fixed m=32 achieved 1.06710 one-step MSE, only 0.40% better than the scheduler. Crossing MAE was 0 steps; integrated-correlation-time relative MAE was 20.7%.

Limitations: This was a synthetic scalar AR(1) closure experiment rather than Burgers, Kuramoto–Sivashinsky, or a neural operator. It used a linear ridge history predictor, did not measure attention/FLOPs or wall-clock savings, did not test regime adaptation, and finite trajectory length biased the integrated-correlation estimate.

How to run: python3 closure_scheduler_experiment.py

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

Artifacts

Implementation overview ⬇ Download all as ZIP 3 files · code, reports and structured results

🐍 closure_scheduler_experiment.py 4.8 KB View Raw
📄 report.md 1.4 KB View
📊 results.json 5.9 KB View Raw