Many of the series that generative time-series models are benchmarked on place a large probability mass on a single value --- it does not rain, no ride is requested, no part is ordered. We report what happens when such data is evaluated carefully. First, the standard rolling-origin protocol can score a model on a window whose atom structure bears no resemblance to the dataset: on one benchmark the dataset is $42\%$ zeros and the evaluation windows are $13\%$, on another $47\%$ against $5\%$. This is not a cosmetic problem --- it reversed one of our own conclusions, turning the strongest occurrence model in our study into what looked like a cautionary tale. Second, we give a control in which CRPS is invariant \emph{by construction} while the temporal coupling is destroyed, which measures exactly how much that coupling contributes to a chosen statistic. Third, benchmarking seven models on a matched protocol over five seeds, an autoregressive hurdle beats a conditional flow on five of six datasets, by up to a factor of $153$, while the flow's own occurrence statistics vary by up to $62\%$ across training seeds and every baseline is deterministic. Finally, the model ordering is not the same under five different occurrence statistics, and the two that do not share a construction agree with each other least.
Comparisons between AutoML systems at short time budgets -- tens of seconds rather than hours -- are common in tool READMEs and workshop papers, and they are easy to get wrong. We report a case study in which a simple AutoML engine, Orcetra, appeared to beat FLAML and AutoGluon on 513 OpenML datasets, winning 57.1% of them at a nominal 60-second budget and 78.4% of datasets against FLAML alone at 30 seconds. Both margins came from protocol defects that a results table cannot show. The search loop scored every candidate on the test split and reported the best, making the headline metric a maximum over dozens of noisy estimates while the baselines selected on training data and touched the test set once; and the budget was checked before launching a candidate but never enforced during one, so the system consumed a median of 120 s against a 60-second budget, 2.24x the wall-clock AutoGluon used. Re-running with selection moved to a validation split, the deadline enforced externally and every framework pinned to an equal share of the machine, Orcetra's win rate on the re-run subset falls from 59.4% to 34.3% and no pairwise difference against either competitor remains significant. Recording both estimands inside a single search lets us attribute the collapse: the selection rule accounts for 4.8 percentage points and unequal compute for most of the rest. The same traces give the selection bias as a function of budget, measured rather than assumed: it grows with $K$ but reaches only 0.27 accuracy points, about five times below the $σ\sqrt{2\ln K}$ bound a marginal-standard-error argument predicts, because candidates scored on shared test rows cancel most of the noise. We close with a checklist for short-budget comparisons. Code, per-dataset results and the scripts that regenerate every number and figure in the paper are released with it.
Delay-risk models are usually judged by predictive accuracy. What matters in practice is narrower: with capacity to review only a few shipments, which ones should a manager check first? We evaluate whether machine learning clears a demanding no-model baseline: inspect the highest-value shipments first. Across three real supply-chain contexts: SCMS procurement, DataCo logistics, and Olist e-commerce, we use leakage-controlled rolling-origin evaluation and 1000-sample paired bootstrap confidence intervals. Ranking by predicted delay severity times known value (M1) beats severity-only ranking in all three datasets, yet it does not generally beat value sorting. At a 10% review budget, M1 minus VALUE_ONLY is -5.5 percentage points (pp) for SCMS, +10.1 pp for DataCo, and -4.9 pp for Olist. The divide is consistent with severity learnability: DataCo has R^2 = 0.27 and calibration bias of +0.01 days, whereas SCMS and Olist have R^2 of approximately -0.02 and negative calibration bias. Nested-CV cost-sensitive retraining does not deliver a stable improvement over M1. Rather than proposing a new learning algorithm, this paper presents a deployment diagnostic and evaluation protocol. Value sorting should remain a permanent benchmark, and ML should be deployed only after severity learnability and calibration have been audited and the model clears that gate under leakage-controlled rolling-origin evaluation.
In many operational time-series forecasting applications, such as crowd demand forecasting, the risk related to under-prediction is substantially higher than that of over-prediction. Accurate prediction of rare demand spikes plays a critical role in downstream tasks. Yet most time-series forecasters are trained with symmetric objectives (e.g., MSE, MAE) and evaluated primarily on aggregate error, which can mask failures in extreme-values and peak-timing predictions. We introduce Asymmetric Peak-Aware Loss (APAL), a simple, model-agnostic objective that (i) penalizes under-predictions more heavily and (ii) increases the training weight of peak regions within each forecast window. We further propose a peak-critical evaluation protocol that complements MAE/MSE with channel-wise tail error (Top-10% and Top-1%) and peak metrics (precision, recall, F1 under timing tolerance, and peak timing error). We evaluate APAL on long-horizon multivariate forecasting across five state-of-the-art backbones, with a focus on pedestrian demand forecasting using (i) a production-ready subset of the City of Melbourne pedestrian hourly count dataset and (ii) a beach visitor count dataset. The generality of the loss function for time-series forecasting is tested on additional benchmarks. Across peak-critical datasets and settings, APAL improves tail accuracy and peak-prediction quality while exposing a controllable trade-off with aggregate error, making it a practical solution when peak-prediction failures are the dominant operational concern.
Faults on a cyber-physical system (CPS) are too rare and unrepresentative to characterise, or even to select a model on, so detection must instead model normal behaviour; the standard point-adjusted evaluation, however, rewards detectors that never do. CPS normal behaviour is the union of many imbalanced, curved, thin-fringed operating regimes rather than a single blob; we state this structure as ten assumptions (A1-A10), abbreviated Massive, Implicit, Imbalanced Multimodality (MIIM). We model the normal law with a jointly learned latent representation plus explicit Gaussian-mixture mode clustering, scored in the latent rather than by a global density or a reconstruction residual, and evaluate under a deliberately fair protocol: raw point-wise metrics with no point adjustment, a trivial-detector difficulty split, prevalence-matched F1, and train-normal-only calibration. On three real CPS datasets (WADI, HAI, SKAB), the detector wins both the combined column and the difficult correlation/dynamics-fault column on all three, reaching difficult-subset AUROC 0.831 on HAI, 0.726 on WADI, and 0.610 on SKAB. The margin is largest on the two multimodal datasets the MIIM assumptions target and slimmest on the near-unimodal one, tracking multimodality as the thesis predicts, and it holds against three deep detectors (USAD, TranAD, GDN) re-computed with the same raw metrics, all of which collapse on the difficult subset. The methodological contributions are the MIIM assumption set, the difficulty-stratified fair protocol, and a latent-only score that drops reconstruction because a flexible decoder rebuilds the hard faults faithfully.
While auditing a perturbation-based OOD detector on a document benchmark, we recorded an AUROC of 0.326 -- well below the 0.5 chance level. The cause is a benchmark leak: the designated "OOD" class is one the model was trained on, so its examples sit inside the in-distribution fit set and the detector is penalized for correctly ranking them as familiar. Deleting the class and retraining 35 models across two domains raises the score to 0.911. We distill the contamination into a leak fingerprint -- near-perfect supervised decodability (AUROC approximately 1) coupled with unsupervised detection collapsed below 0.65 -- and validate it on a controlled battery of 52 settings (20 leaked, 32 clean) across ResNet-50 and ViT-B/16 on CIFAR-10/100, achieving sensitivity 18/20 and specificity 31/32 in embedding space; the matched fit-set-exclusion controls are perfect at 20/20. An in-the-wild audit of 24 standard near/far OOD benchmark pairs fires on exactly one (the intrinsically hard CIFAR-100 vs CIFAR-10 pair) and on no far-OOD pair, confirming specificity and that standard cross-dataset construction is clean. Under the corrected protocol, perturbation signals are decodable but not detectable: a supervised reader recovers the OOD signal (AUROC 0.87-1.00) while no unsupervised detector does, and the perturbation method does not improve on plain Mahalanobis distance. We provide a theoretical account of why and, for transparency, retract an earlier circular correlation. The contributions are a corrected protocol and a validated leak diagnostic, not a new OOD method.
Despite the success of deep learning, training deep networks in biologically plausible and hardware-efficient ways remains an open challenge. Feedback alignment (FA) methods address this by replacing backpropagation's symmetric backward weights with fixed random matrices, but their effectiveness depends critically on whether they can be accurately evaluated. The standard evaluation relies on two quantities: task accuracy and cosine similarity between the method's credit signal and the backpropagation gradient. We show that this reporting pair is insufficient by identifying two independent failure modes, both silent under current reporting: (1) measurement degeneracy, where the BP reference gradient collapses to the numerical floor in terminal-LayerNorm residual architectures, rendering cosine uninterpretable; and (2) aggregation collapse, where the aggregate cosine masks layerwise heterogeneity that concentrates credit at one end of the network. To address these limitations, we propose a diagnostic evaluation protocol based on three checks -- scale stability, reference validity, and depth utility -- together with per-layer rather than aggregate cosine reporting. Across multiple architectures and methods, the standard reporting pair gives no signal of failure in any audited case, while our protocol identifies all failures with wide calibration margins. The two failure modes are causally independent: a per-block scale penalty alleviates Mode 1 (residual scale explosion driving reference collapse) without affecting Mode 2 (cosine ranking that contradicts every functional metric we measured). Identifying these silent failures prevents researchers from building on non-functional credit assignment and provides actionable guidance for developing FA methods that genuinely train deep layers.
Reconstruction-based methods are widely used for time series anomaly detection, where models are trained to reconstruct subsequences, and anomalies are identified through reconstruction errors. However, reported results are often hard to compare due to heterogeneous evaluation practices and underspecified inference procedures. In this paper, we revisit reconstruction-based anomaly detection in the univariate offline setting and study the role of the inference stride, which controls whether subsequences are processed as disjoint windows or with overlap. We propose a unified training, tuning, and multi-seed evaluation protocol on the curated TSB-AD benchmark, and study how overlapping inference affects anomaly detection performance for a range of reconstruction models, including PCA-based baselines, DLinear, an AutoEncoder, TimesNet, and Transformer variants. The results show that across all models, overlapping windows yield consistent improvements, with average relative gain up to +28%, and can alter method rankings. We further analyze variability across datasets, random seeds, and hyperparameter configurations. Finally, we complement the benchmark study with an evaluation on the full UCR archive using localization criteria aligned with sliding-window reconstruction. Overall, our results highlight that reconstruction-based anomaly detection performance depends not only on model architecture and training, but also on inference choices, motivating a clear and reproducible protocol. Our results show that reconstructionbased baselines achieve strong performance on both TSB-AD and UCR benchmarks, supporting them as competitive and practical approaches for univariate time series anomaly detection.