✗ Failed on benchmark 2026

Full-Rank Local Replay

Implementation & benchmark of arXiv:2607.09187 — Data-driven predictive control of nonlinear systems using weighted regularization

Usefulness7/10

Difficulty4/10

Novelty5/10

Training Accuracy Stability Sample efficiency World model Training dynamics Regularization Optimizer Control theory Optimization Linear algebra Dynamical systems

Source paper: Data-driven predictive control of nonlinear systems using weighted regularization arXiv:2607.09187 ⓘ · analyzed Aug 30, 2026

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 Not peer-reviewed.

Idea description

Train a neural state-space model using all replayed transitions, but assign larger weights to samples near the current operating context rather than discarding distant samples. Add a strictly positive weight floor so local adaptation cannot eliminate global coverage or make the regression problem rank-deficient. This should improve prediction across nonlinear regimes while retaining the numerical robustness of full-data training.

Formulas

lambda_j = lambda_min + (lambda_max-lambda_min) exp(-d_j^2/(2 sigma^2)), d_j^2 = (z_j-z_c)^T M (z_j-z_c), Lambda = diag(lambda_1,...,lambda_N).

theta_plus = theta - eta grad_theta [ (1/N) sum_j lambda_j ||f_theta(z_j)-x_{j+1}||_2^2 + beta ||theta-theta_0||_2^2 ].

W_star = (Phi^T Lambda Phi + beta I)^(-1) Phi^T Lambda Y, Phi^T Lambda Phi + beta I >= beta I > 0.

Mathematical statement

Let z_j = (x_j,u_j) denote replay context j, let z_c be the current operating context, and let d_j be the distance d_j = sqrt((z_j-z_c)^T M (z_j-z_c)), where M is positive semidefinite. Define positive diagonal weights lambda_j = lambda_min + (lambda_max-lambda_min) exp(-d_j^2/(2 sigma^2)), with 0 < lambda_min <= lambda_max and sigma > 0. For a neural transition predictor f_theta, minimize the weighted empirical objective sum_j lambda_j ||f_theta(z_j)-x_{j+1}||^2 plus beta ||theta-theta_0||^2. In a linearized last-layer problem with feature matrix Phi and targets Y, the solution is W = (Phi^T Lambda Phi + beta I)^(-1) Phi^T Lambda Y, where Lambda = diag(lambda_1,...,lambda_N). Because Lambda is positive definite when lambda_min > 0 and beta > 0, the normal matrix is positive definite and has minimum eigenvalue at least beta.

Implementation notes

1. Integration point: use this weighting rule in the replay loss of an online neural state-space model, latent world model, or locally adapted dynamics predictor. Store transitions (x_t,u_t,x_{t+1}) in a replay buffer, normalize state and input coordinates, and define z_t = [x_t;u_t]. At each update, define z_c from the latest observed context or from the mean context of the current rollout. 2. Pseudocode: sample a minibatch or the full buffer; compute d_j^2 = (z_j-z_c)^T M (z_j-z_c); compute lambda_j = lambda_min + (lambda_max-lambda_min) exp(-d_j^2/(2 sigma^2)); normalize weights to have minibatch mean one; compute L = mean_j lambda_j ||f_theta(z_j)-x_{j+1}||^2 + beta ||theta-theta_0||^2; backpropagate and update with SGD or Adam. Do not discard low-weight samples. 3. Taken directly from the paper's mechanism: positive diagonal weighting, operating-point localization, and retention of every data column. Estimated empirically: the metric M, bandwidth sigma, floor lambda_min, and ridge beta. Start with M equal to the identity, lambda_max = 1, lambda_min = 0.05, and sigma equal to the median pairwise context distance. 4. First cheap experiment: compare uniform replay, hard nearest-neighbor replay, and weighted replay on an MLP predicting a two-regime nonlinear dynamical system or the nonlinear two-tank system. Use the same buffer and optimizer, and measure one-step and 20-step rollout error near and far from the current context. The quantitative prediction is a smooth weight decay: at d = sigma, the Gaussian excess weight is exp(-1/2) = 0.6065 of its value at d = 0. The last-layer normal matrix should have minimum eigenvalue at least beta, while hard selection may become singular. Sweeping sigma should produce a U-shaped validation curve: small sigma causes global-regime forgetting, while excessively large sigma removes localization. With lambda_min below approximately 0.01, predict increased far-regime error; with sigma above roughly twice the median context distance, predict behavior close to uniform replay.

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 9/10

Built and ran a fixed-seed full-rank local replay MVP with Gaussian weighting, a positive weight floor, ridge-stabilized linear algebra checks, and a nonlinear MLP replay comparison. The quantitative predictions were confirmed: excess weight at d=sigma was 0.60653066 for every tested sigma, bandwidth decay matched exp(-d^2/(2 sigma^2)), and the ridged normal matrix had minimum eigenvalue 0.0300000000000 for beta=0.03. Localization improved near-context MSE versus uniform, 4.23e-05 versus 1.73e-04, but harmed far-context MSE, 6.63e-03 versus 5.75e-05, so this is a mechanism success and local-adaptation tradeoff rather than an overall accuracy win.

Agent confidence: 9/10

Baseline: Uniform replay: near MSE 1.7282e-04, far MSE 5.7509e-05, all MSE 1.5704e-04
Idea: Full-rank localized replay: near MSE 4.2315e-05, far MSE 6.6287e-03, all MSE 4.3594e-03; hard-nearest reference: near 1.12397e-04, far 2.78484e-01

Limitations: Only one synthetic one-step regression system and one operating context were tested. No online changing-context rollout, multi-step benchmark, speed study, or systematic validation sweep of sigma and lambda_min was performed. A positive weight floor alone does not make Phi^T Lambda Phi full rank if Phi is rank-deficient; the positive-definiteness guarantee here depends on beta>0.

How to run: python3 run_experiment.py

Stage 2 — Benchmark vs. tuned baseline (arXiv:2607.09187) Aug 31, 2026

Significantly WORSE than the tuned baseline (+43.1%, p=0.0224, wins 1 / 8 paired seeds).

Benchmark: Dynamics forecasting
Model: rnn_small
Paired seeds: 8
Baseline mean: 0.0009
Idea mean: 0.0013
Effect (Δ): 0.0004 (+43.1%; negative = idea better)
Wins: 1 / 8 paired seeds
p-value: 0.0224 (permutation test, 20 000 shuffles)
Smallest detectable effect: ±38.9%
Mechanism: Not confirmed ✗
Practical effect: Hurts
Baseline tuning: swept over 3 configs

Limitations:

Only the built-in dynamics track was tested, using 400 training and 400 test examples, 18 epochs, and a three-value sigma sweep. The context center was estimated from the final eight training contexts rather than a changing online rollout. The generated signature had no samples in its narrow sigma bin; its exact Gaussian confirmation is analytical and does not establish a neural benchmark win.

How to run: python3 bench_local_replay.py

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

Stage 2 — Benchmark Aug 31, 2026 latest Failed ✗

Agent confidence: 10/10

Full-Rank Local Replay was implemented as a weighted replay-loss intervention on the registered dynamics/rnn_small benchmark with matched architecture, optimizer, batch, epochs, paired seeds, and parity-compliant sweeps. The tuned uniform baseline achieved mean test MSE 0.00090519, while the best idea configuration achieved 0.00129543; paired delta was +0.00039024 with permutation p=0.0224, indicating a significant loss. The Gaussian and ridge sanity checks passed, but the benchmark effect was negative.

Baseline: mean test MSE 0.0009051897650, best lr=0.006
Idea: mean test MSE 0.0012954341728, best lr=0.003 sigma=2.0 lambda_min=0.05; paired delta +0.0003902444078, p=0.0224

Benchmark result