Area Under The Receiver Operating Characteristic Curve

Video anomaly detection (VAD) is an actively studied task, having wide applications in typical scenarios such as public surveillance and road traffic safety. The task is also relevant for robotic arm interactions, where it has several downstream applications, including learning better interaction and manipulation abilities, triggering recovery procedures when anomalies occur, etc. Despite its relevance, the exploration of anomaly detection in robotic arm manipulation videos is limited by the low number of available resources. To this end, we introduce RoboVAD, a large-scale benchmark for video anomaly detection that comprises challenging cross-domain evaluation scenarios, where certain actions (tasks executed by a robotic arm) and anomaly types (mistakes that occur while performing certain tasks) remain unseen during training. RoboVAD is designed to benchmark VAD methods in realistic scenarios, where robotic arms can perform unforeseen tasks, and thereby encounter new anomaly types. We train and evaluate several state-of-the-art VAD methods, including a novel method specifically adapted for robotic arm manipulation. While the proposed method outperforms many state-of-the-art competitors, all methods remain below a micro-averaged frame-level AUC threshold of 70% in the most challenging evaluation setup, confirming the difficulty of the proposed benchmark. We publicly release our dataset and code at https://zenodo.org/records/22754659.

The clinical management of autism spectrum disorder (ASD) faces a bottleneck in early screening, mainly because trained specialists are scarce and conventional assessment tools are subjective. Here, we introduce ASDchat, a multimodal large language model designed for evidence-based ASD screening, which takes video, audio, and dialogue as input. ASDchat adopts a dual-branch architecture, where the decision branch generates screening probabilities and the evidence branch generates traceable, timestamped behavioral evidence aligned with standardized clinical criteria (ADOS-2). The model was trained and evaluated on a dataset of 1,035 participants from 27 sites in China, which covered typically developing (TD) children, children with ASD, and children with other disorders. For ASD versus TD, ASDchat reached an area under the receiver operating characteristic curve (AUC) of 0.953 $\pm$ 0.021. On 9 held-out sites that were not used for training, the mean AUC was 0.932. Furthermore, unsupervised clustering of the behavioral dimensions split the ASD cases into six subtypes with different phenotypic profiles, and ASDchat suggests an intervention for each subtype. ASDchat provides a feasible path for large-scale, evidence-based early ASD screening in clinical practice.

Maximizing the area under the receiver operating characteristic curve (AUC) is a standard approach to imbalanced binary classification. Although positive and negative data are required for maximizing the AUC, negative data are often difficult to collect in some real-world applications due to privacy concerns or the need for specialized expertise to annotate them. Thus, AUC maximization from positive and unlabeled (PU) data has been attracting attention. Existing methods assume that labeled positive data are unbiased samples from the true positive distribution. However, this ideal assumption is often violated in practice. In this paper, we propose a method to maximize the AUC from biased PU data. To address the bias, our key idea is to exploit {\it confidence}, i.e., the probability that an instance is positive, associated with the small number of labeled positive data. We derive an estimator of the AUC risk using biased PU data with confidence, enabling AUC maximization under such bias. We further show that the rewritten AUC risk induces a Bayes-optimal AUC ranking even when the available confidence is any strictly increasing transformation of the true posterior probability. We experimentally show the effectiveness of our method on eight real-world datasets.

Cardiovascular screening models trained on national health surveys routinely report areas under the receiver operating characteristic curve (AUROC) near 0.89. We asked whether that accuracy reflects learning or target leakage, whether tabular foundation models change the answer, and whether the properties deployment requires survive joint examination. We benchmarked ten classifiers spanning linear, tree-ensemble, neural, glass-box, and tabular foundation classes for prevalent myocardial infarction in 442,067 respondents of the 2022 Behavioral Risk Factor Surveillance System across five feature tiers of decreasing leakage risk. Each was audited for discrimination, calibration, fairness at an explicit screening threshold, conformal coverage, explanation faithfulness, and inference cost, then applied -- models and thresholds frozen -- to 430,755 respondents of 2023. Removing two post-diagnostic features cost every model 0.049-0.051 AUROC, collapsing the field into a 0.0045-wide band. The glass-box explainable boosting machine was non-inferior to every alternative within a pre-specified 0.005 margin while scoring the cohort roughly 104 times faster than the strongest foundation model. One threshold detected 75.4% of women's infarctions against 89.0% of men's; editing the model's shape functions reduced the gap to 0.010. Marginal conformal prediction gave 0.86 coverage to men and 0.82 to adults over 60; Mondrian calibration repaired every stratum. Frozen models transported within 0.002 AUROC. Reported headroom in this literature is a property of the feature set, not the learner. Transparency cost nothing measurable and made fairness repair and uncertainty conditioning directly auditable. Evaluation practice, not model capacity, is the binding constraint.

Most Deepfake detectors report near-perfect AUC scores on their reference benchmarks. However, a recent ICML position paper argues that these evaluations collectively neglect the impact of modern smartphone photography: the widely used on-device neural image-signal processing pipelines (like multi-sensor fusion or noise and motion-blur suppression) increasingly shift the imaging paradigm from simple lens projections towards computational photography. Hence, devices actually generate, rather than record photos. This increases the risk that deepfake detectors may flag ordinary phone photos as fake. Due to the lack of large-scale datasets containing images from modern smartphones, this hypothesis has so far only been tested in small proof-of-concept studies. The aim of this paper is to close this gap. We introduce LAION-Mobile, an open dataset containing about 1 million smartphone images with EXIF metadata distilled from re-LAION-5B. Evaluating twelve state-of-the-art deepfake detectors with their original paper checkpoints on a 9,115-image evaluation sample of this pool (DIRE on 738), we report three key findings: (i) On modern AI content no detector exceeds AUC 0.624, and five of twelve fall below chance. (ii) Real-photo false-alarm rates are an artefact of threshold calibration: thresholds fitted on legacy GAN data make several detectors look deployable (less than 11 percent FPR), yet the same detectors flag 17-91 percent of real photos once the identical criterion is refit on modern content. (iii) Consequently, no detector both beats chance on modern AI content and keeps a deployable real-photo false-alarm rate. Mirroring the device mix of web collections, the corpus probes the first neural-ISP generation (2018-2020); current flagships are essentially absent, leaving the modern-ISP regime as the open gap.

Clinical diagnosis agents must decide not only what test to request next, but also when to diagnose or defer. Existing agent benchmarks largely evaluate accuracy after fixed or unconstrained interaction, leaving autonomous stopping reliability implicit. We present Cros, a risk-constrained stopping layer combining state-wise error ranking, policy design on disjoint development splits, and LTT-style exact tests of selective diagnostic error and minimum autonomous coverage for complete sequential policies. Its finite-sample guarantee requires the candidate family, testing rule, and any randomization to be frozen before calibration labels are accessed. On a 1,834-episode MIMIC-derived abdominal-pain benchmark, the full ranker achieves exploratory state-error AUROC 0.853, compared with 0.715 for maximum class probability and 0.552 for the backbone's native stop score. On the previously viewed 367-episode evaluation split, analytically averaging over the frozen Cros weights yields 16.9% selective error at 78.8% coverage, cost 5.57, and 0.68 tests, versus 30.8% error at 100% coverage, cost 8.14, and 1.53 tests under native stopping. Forced continuation is non-monotone: error is 28.3% with HPI alone and 34.3% after full workup. However, the uniform-weight mixture ablation is cheaper on this viewed split despite missing the locked development margins, and Cros nominally satisfies the joint criterion in only 6 of 20 development resplits. Because evaluation labels were inspected during earlier development, these findings provide exploratory feasibility and audit evidence, not a confirmatory safety certificate.

Chagas disease is a major cause of cardiomyopathy in Latin America. Cardiac magnetic resonance (CMR) imaging can characterize its structural abnormalities, but scanners and expert readers remain scarce in endemic regions. Electrocardiography (ECG) is inexpensive and widely available, yet structural disease must be inferred indirectly from electrical signals. We propose to transfer CMR-derived structural knowledge to ECG through contrastive pre-training. Using 63,193 paired ECG-CMR examinations from the UK Biobank, we align an ECG encoder with a clinically grounded CMR embedding space using an asymmetric InfoNCE objective. Despite seeing no Chagas cases during pre-training, the resulting representation improves ECG-based Chagas detection. Across CODE-15% and SaMi-Trop, a frozen linear probe achieves an AUROC of 0.851 and sensitivity at the top 5% of predicted risk (Top5%-TPR) of 0.427 in five-fold cross-validation, compared with 0.827 and 0.377 for an unaligned ECG-FM baseline. On the PhysioNet/CinC 2025 Challenge test set, our model obtains the highest AUROC on SaMi-Trop-3 and the best ELSA-Brasil challenge score among the three top-performing methods, indicating that imaging-supervised ECG representations can generalize to populations and settings beyond the pre-training distribution.

Shortcut learning denotes the widespread situation in which a classifier exploits spurious correlations rather than diagnostic features. Existing mitigation strategies mostly aim to learn shortcut-invariant representations; their empirical success is limited and they cannot be applied to classifiers using frozen foundation model encoders. We propose to reframe shortcut learning as fundamentally a calibration problem: unconstrained learning implicitly calibrates each shortcut group to its training set disease prevalence, rendering the resulting classifier necessarily over-confident in one group and under-confident in the other. Building on this insight, we prevalence-equalize calibration between shortcut groups through two encoder-agnostic methods, an in-processing regularizer and a post-hoc prevalence-equalized recalibration step. Across chest-drain-pneumothorax benchmarks on CheXpert and SIIM-ACR, spanning fine-tuned CNNs and frozen foundation-model backbones, both methods substantially outperform all baselines. Post-hoc recalibration of a standard ERM-trained DenseNet raises misaligned-group AUROC from 0.23 to 0.73, indicating that shortcut reliance degrades the classification head rather than the underlying representation. Besides two new state-of-the-art shortcut mitigation approaches, our findings more fundamentally connect shortcut learning to calibration theory and algorithmic fairness.

How can vision-language models help video anomaly detection (VAD) when surveillance data remain distributed, weakly labeled, and resource-constrained? Most weakly supervised VAD methods assume centralized training; recent VLM-based extensions further rely on dense inference, generated explanations, or additional adaptation. We introduce a lightweight federated MIL-VLM cascade in which only a compact MIL scorer is trained across clients, while a frozen VLM verifies high-scoring suspect segments post hoc. We study two VLM feedback interfaces: parsed text-generation decisions and a logit-based interface that extracts a continuous anomaly score from next-token Yes/No probabilities. Experiments on UCF-Crime with InternVL3.5-2B and Qwen3-VL-2B-Instruct show that text-generation verification can improve frame-level AUC after diagnostic temporal post-processing, but remains sensitive to prompts, parsers, model choice, and smoothing. In contrast, the logit interface provides a fixed parser-free signal that improves both frame-level AUC and frame-level AP over the MIL baseline across both VLMs, without temporal post-processing in its main configuration. Since suspect segments are updated independently once available, next-token logit feedback provides a simple segment-local alternative to text-generation verification.

Text-to-Speech (TTS) foundation models are increasingly fine-tuned on private datasets to synthesize highly personalized voices, introducing severe privacy risks by exposing both biometric identities and sensitive speech content. Existing black-box membership inference attacks (MIAs) follow a two-stage pipeline of query generation and representation engineering, both of which face unique challenges when adapted to TTS. For query generation, dual conditioning on synthesis text and reference speech creates a large and underexplored query design space with no established criterion for identifying an effective query. For representation engineering, the multi-level speech characteristics and temporal variability of speech make low-level representations and direct comparisons inadequate for capturing membership signals. To address these challenges, we present the first black-box MIA framework explicitly tailored to TTS models at both the speaker and record levels. For query generation, we characterize the feasible query space and establish two criteria, scorable extent and memorization elicitation, for evaluating five representative queries, identifying recitation as the strongest. For representation engineering, we obtain multi-level speech representations from embedding models and temporally align the generated and target audio for fine-grained comparison. Evaluations across three state-of-the-art TTS models (CosyVoice2, F5-TTS, and XTTS-v2) fine-tuned on two benchmark datasets (VCTK and British Dialect) reveal severe privacy leakage: speaker-level AUC remains above 0.80 and approaches 1.0 in the strongest settings, while record-level AUC ranges from 0.80 to 0.90 and remains effective even in challenging scenarios where both members and non-members are of the same speakers. We further identify speech characteristics associated with disproportionate vulnerability to memorization.

Following arXiv:2607.25507, this report examines whether phase-derived features improve endpoint prediction over a combined baseline in two settings: a sealed contradiction comparison and a retrospective analysis of answer changes across matched pressure prompts. Study 1 froze a contradiction-category pipeline before sealed scoring. On 1,136 eligible primary cases, adding PC-2 produced a paired AUROC difference of +0.00087. The 99% bias-corrected accelerated interval included zero, and the prespecified +0.05 threshold was not met. A replication role with 1,063 cases showed a same-direction increment of +0.00019. The replication-direction condition passed, but both primary conditions failed, so advancement failed. Study 2 developed fifteen treatments on blocks b0-b4 using 1,415 eligible answer-change comparisons and twenty-repeat, five-fold grouped cross-validation. An execution on 9 September 2026 recomputed development statistics from saved prediction units and applied a three-condition gate; no treatment advanced. Layer 25 had the only positive mean-repeat PC-2 increment, approximately +0.00027, with a favorable sign in three of five seed blocks. These statistics measure development consistency, not whole-seed-block holdout performance. Agreement with earlier selection records does not establish an earlier notebook execution or that the rule was fixed before inspecting development results. The planned b5 selection and b6 evaluation were not completed through this gate. Neither study demonstrated the incremental benefit required by its applied advancement rule. The findings limit support for the evaluated feature constructions; they do not test the rotary score identity, its local pre-softmax bound, or the effectiveness of execution-boundary governance.

We propose a lightweight two-stage framework for real-time video anomaly detection. The first stage employs YOLO v11n-pose to detect persons and extract seventeen skeletal keypoints in a single forward pass. The second stage encodes each cropped person region through CLIP ViT-B/32 and computes cosine similarity against predefined textual descriptions of anomalous behaviors. This architecture eliminates the need for optical flow, standalone pose estimators, and density-based scoring modules. Experiments on CUHK Avenue, ShanghaiTech Campus, and a custom indoor dataset collected at Chulalongkorn University demonstrate an end-to-end throughput of approximately 51 FPS on an NVIDIA Titan XP GPU, a 3.36x speedup over the multi-feature baseline, while maintaining frame-level AUROC values of 89.26%, 70.26%, and 84.13%, respectively.

Background: Population level assessments of predictive artificial intelligence (AI) can conceal performance disparities across subgroups. Fairness evaluations commonly rely on performance analyses across subgroups. However, some performance metrics are non-collapsible, meaning that the overall population performance value does not equal the weighted average of subgroup specific values. Objective: To examine the collapsibility properties of commonly reported performance metrics in predictive AI, with a focus on the area under the receiver operating characteristic curve (AUC, also known as c-statistic). Methods: We investigate the collapsibility of 15 performance metrics, either by expressing each metric as a linear combination of its stratum specific values or, where non-collapsible, by providing a counterexample inspired by Simpson's paradox as a formal disproof. Results: Five performance metrics (AUC, calibration intercept, calibration slope, expected calibration error, and Nagelkerke R^2) are shown to be non-collapsible, and ten (O:E ratio, logloss, Brier score, accuracy, F1-score, true positive rate, true negative rate, positive predictive value, negative predictive value, and net benefit) are shown to be collapsible. The AUC is shown to be non-collapsible because it decomposes into within- and cross-group AUC terms when subpopulations coexist, such that its overall value may fall outside the range of subgroup specific AUCs. Conclusions: Non-collapsibility of performance metrics has important consequences for reporting, model appraisal, and fairness evaluation. It can generate spurious differences between subgroup and overall performance, which may mislead fairness evaluations. Explicitly acknowledging and reporting the collapsibility properties of performance metrics improves both the interpretability and transparency of fairness assessments.

Accurate bladder tumor segmentation and assessment of mus- cle invasion from T2-weighted MRI are important for treatment plan- ning, but developing robust models across institutions is challenging be- cause patient data cannot be centrally pooled and imaging characteristics vary across scanners and acquisition protocols. We propose a federated multi-task learning framework for joint bladder tumor segmentation and MIBC/NMIBC classification across four clinical centers. The proposed Swin Hybrid model combines a ResNet-34 branch for local texture and boundary information with a Swin-Tiny Transformer for global anatomi- cal context. A segmentation-guided classification mechanism further uses tumor localization information to support MIBC prediction. We also investigate several augmentation strategies under both centralized and federated training to improve robustness to multi-center variability. Ex- periments on the FedBCa dataset show that the Swin Hybrid provides the best overall balance between segmentation and classification among the evaluated architectures. Under federated training, Geo+Elastic aug- mentation achieved a DSC of 0.8100 and a patient-level AUC of 0.8931, yielding the highest combined score of 0.8474. These results demonstrate that joint segmentation and classification can be effectively performed across multiple institutions using federated training without centralizing patient data.

We compare six self-supervised pretext tasks for infant cry analysis under a fixed budget, meaning the same compact encoder of 1.17M parameters, the same 115 hours of license-verified public pretraining audio, and the same evaluation protocol for every candidate. On cry detection the reconstructive objectives dominate, and a linear probe over a masked-spectrogram encoder reaches 0.988 AUC with subject-wise splits even though the encoder never observed a cry during pretraining. On cry-reason classification over donateacry, the de facto public benchmark for cry reasons, every encoder performs at chance (0.38 to 0.54 macro AUC over 5 classes), and neither domain adaptation on 1.8 hours of real cries nor end-to-end fine-tuning moves the result. Since a frozen HuBERT-base with 80 times more parameters shows the same pattern, the bottleneck must sit in the labels and not in model capacity. We then reproduce the 90%+ accuracies of the donateacry literature on our own system by changing nothing but the evaluation protocol: clip-wise splits raise accuracy to 85.2% (barely above the 83.8% majority-class baseline), and applying augmentation before splitting raises it to 97.9%, matching the reported state of the art, from the same model that measures 0.49 macro AUC under subject-wise splits. Under leakage-free splits, a twentyfold augmentation of the labeled set (vocoder speaker perturbation and noise mixing, 21 hours) leaves cross-subject AUC unchanged: for this task the effective sample size is the number of infants. We release code, seeds and per-clip license manifests.

Click-through rate (CTR) models vary in feature-interaction design, yet their top networks usually remain a single multilayer perceptron shared by all examples. Heterogeneous user, item, and context subgroups therefore update the same parameters; weakly aligned learning signals make the aggregate gradient a compromise among competing directions. We study the competition on Avazu with 4 models and 4 semantic fields. Across all architectures, semantic subgroups show lower Top-NN gradient cosine similarity than random groups matched by sample size and label ratio, with reductions of 0.23-0.37. This competition motivates input-conditioned experts, but directly replacing an established Dense mapping changes its initial function, sharing pattern, and capacity, obscuring the source of gains. We introduce PRIME (Plug-in Residual Input-conditioned Mixture of Experts), a Dense-anchored mixture of low-rank residual experts. PRIME anchors the original prediction and uses zero-residual initialization to match the Dense baseline exactly at training onset. Input-dependent routing weights low-rank experts for example-specific logit corrections; multi-bag aggregation and EMA load biases stabilize conditional estimation. We evaluate PRIME on held-out Avazu and Criteo test sets across 13 CTR architectures and five paired seeds. Median paired AUC gains are +0.0022 and +0.0066, with LogLoss reductions of 0.0011 and 0.0081, respectively. On FiBiNET and DCNv2, PRIME outperforms APG in all ten seed-level AUC comparisons while using fewer parameters and lower inference latency on both backbones. These results show that function-preserving conditional residuals add input-dependent capacity while preserving the Dense path and its optimization stability. Code is available at https://github.com/YH-learning/PRIME.

Concept bottleneck models (CBMs) can improve the transparency of cancer image diagnostic prediction by expressing predictions through radiological concepts. However, their dependence on instance-level concept annotations limits practical applicability. We propose a prior-guided hybrid CBM that integrates limited concept annotations, class-conditional concept distribution matching on unannotated patients, and prior initialization of the concept-to-diagnosis head. We evaluate the method on CBIS-DDSM mammographic masses and calcifications and LIDC-IDRI pulmonary nodules across 0-100% concept annotation. In the clinically relevant 0-20% annotation regime, the hybrid CBM consistently improves mean concept AUC over a matched standard CBM, while maintaining diagnostic performance close to black-box models. At 10% annotation specifically, concept AUC increases from 0.619 to 0.741 for masses, from 0.650 to 0.787 for calcifications, and from 0.597 to 0.642 for pulmonary nodules. Ablation experiments identify prior initialization as the main component contributing to improved concept detection, likely by stabilizing the concept-to-diagnosis head. Zero-shot VLMs remain insufficient for reliable fine-grained tumor-level concept prediction. These findings suggest that structured priors can substantially reduce the annotation burden of interpretable cancer imaging models.

Access to clinical data is essential for developing reliable healthcare machine learning systems, but direct use of electronic health records is constrained by privacy regulation, institutional review, data-use agreements, and the risk of re-identification. Synthetic data promises a practical alternative: it can preserve useful statistical and clinical structure while reducing exposure of sensitive patient records. Prior studies often evaluate a single generator, one dataset, or a narrow downstream task, making it difficult to know when synthetic data can support model development and when it fails to preserve task-critical signal. We introduce CoMedBench, a reproducible benchmark that evaluates a family of generators under a common clinical-validity framework and one shared training and evaluation engine, spanning static tabular and temporal downstream tasks on established critical-care datasets. In total the benchmark spans 37 dataset-task pairs across two modalities consists of 20 static tabular and 17 temporal ICU time-series-drawn from seven public data sources: three intensive-care databases (MIMIC-III, MIMIC-IV, and eICU) together with the UCI Machine Learning Repository, the CDC BRFSS diabetes cohort (2015), NHANES (1999-2014), and the pycox survival datasets (GBSG and METABRIC). The benchmark evaluates both statistical fidelity and task utility by comparing models trained and tested across real and synthetic data. In these settings, synthetic training data preserves most of the downstream signal: on tabular tasks the reference generator CoMed-CTGAN retains a mean AUROC utility (the synthetic-to-real performance ratio) of 90.6%, rising to 97.3% for the strongest generator, CoMed-TVAE. Temporal ICU tasks are harder and more generator-sensitive: CoMed-CTGAN retains 81.6% (AUROC) and only 64.0% under the imbalance-sensitive AUPRC, whereas CoMed-TVAE still retains ~95% (AUROC).

Professionals across medicine, engineering, finance, manufacturing, and the sciences often make consequential decisions from charts. Existing chart benchmarks do not sufficiently measure this ability: they are dominated by bar, line, and pie formats, rely on shorter reasoning chains, and are nearing saturation, with frontier models already scoring 80-90%. We introduce Chartography, a benchmark of 100 tasks that pair charts drawn from professional practice, in domain-specific formats that standard chart benchmarks rarely include, with questions written by professionals who read these charts for a living and independently verified by three additional experts. In an evaluation of 30 frontier-model configurations (20 scored trials per task), the best configuration reaches only 45.0% mean pass@1; the remainder span 9.0-39.5%. Failures concentrate in visual perception: models can miss nuanced features, misread values along sparsely labeled axes, mishandle projected 3D geometry, and violate domain conventions encoded in the chart. We release all tasks, images, provenance metadata, and evaluation code.

Automated radiology report generation is advancing rapidly in response to the shortage of radiologists, yet unlike a perception model, existing generation models offer no control over the sensitivity-specificity trade-off of their diagnostic content. Such control is essential because clinical scenarios diverge: emergency triage prioritizes sensitivity to reduce missed findings, whereas confirmatory interpretation emphasizes specificity to limit unnecessary interventions. A single fixed report can neither adapt to these scenarios nor support the ROC-based validation widely expected for regulatory clearance. We introduce RadFusion, a framework that equips report generation with threshold controllability. Our method fuses a multi-label classifier, which provides per-disease confidence scores, with a VQA-based report generator, which describes medical findings in detail; an LLM then rewrites the report so that its stated diagnoses follow the classifier's decisions at the selected threshold while staying grounded in the generator's descriptions. On MIMIC-CXR, the performance of RadFusion conforms to the classifier's ROC curve: sweeping the threshold and mapping the reports back to class labels reproduces the classifier's validated ROC performance. This conformance makes generated reports quantitatively evaluable through ROC analysis, strengthening the case for regulatory clearance, and enables operating-point selection that matches report behavior to clinical context. Moreover, combining the two model types improves diagnostic accuracy over uncontrolled generation: sensitivity increases by 6.9% at matched specificity, and specificity by 20.7% at matched sensitivity. These results show that RadFusion makes report generation clinically adaptable, quantitatively verifiable, and diagnostically more reliable.

Agent frameworks ship quality gates that compare text blocks by embedding-cosine similarity and decide at a fixed cutoff. Deduplication filters, semantic caches, drift guards, and answer grader gates deploy to answer the question: "Does this text still mean the same thing?" But the score answers a different question: "How much did the wording change?" We audit this gate class as a measurement instrument. In the cases these gates exist to catch, the two can run in opposite ways. Many times, reversing an instruction is a single word edit, while agreement often rephrases a sentence. The consequence is a safety check that fires backwards. The production drift guard we audited caught 0 of 56 meaning-breaking mutations, and one approved item, "withhold the study drug" -> "administer the study drug", came in at cosine 0.9608. We observed five shipped operating points, and balanced accuracy across 90 configuration-threshold-task cells never exceeded 0.700 (median 0.525). The same confounder also corrupted evaluations. A naively built corpus inherits this confounder and can return an inverted verdict, with a decision AUROC exactly 0.000 in 13 of 18 configuration-task cells (at most 0.040 in all 18) against 0.440-0.815 for the same nine configurations under a balanced 2x2 design. Twice in the effort it captured our own headline claims. Obvious repairs fail: an encoder swap and an overlap-conditioned gate (0.750 in-sample, 0.533 held-out) land at chance on separately authored held-out data, and an NLI drop-in did no better. Embeddings do still bear hope here, as the strongest two of nine configurations separated reversal from paraphrase at matched overlap (AUROC 0.79-0.90), but only a matched-pair audit reveals the deployment regime. We release the corpus method, harness, and frozen results, and contend that scores gated this way measure the wrong thing. We believe a valid instrument is buildable.

Internal safety scores judge a prompt before any text is generated, and they are validated by how well they separate harmful prompts from benign ones. That separation is then read as evidence that the score will also catch the attacks that succeed. Harmful intent is a property of the prompt. Jailbreak success is an outcome produced later by a particular target model, decoding policy, and judge. A filter tuned on a score that measures the wrong quantity spends its false positive budget on attacks that would have failed anyway. In this paper we audit that inference. Attention based measurements are usually read from prompt dependent locations, so a wrapper changes both the content being judged and the place the signal is taken from. We therefore introduce Active Attention Probing, which supplies a fixed content independent measurement coordinate. We pair every base goal with a plain and a wrapped version and generate real completions from the target models. On Llama, wrapping raises harmful generation from 0.05 to 0.27 while harmful intent AUROC falls from 0.936 to 0.803, so the attacks grow more dangerous while the prompts look safer to the score. Among wrapped harmful prompts the outcome AUROC is 0.220, which places the attacks that succeeded below the attacks that failed. Rare token, passive, and detector derived channels reproduce the reversal on the same matched design, and the reversal itself persists across three target models, seven attack families, and two independent judges. Distribution shift then degrades calibration and threshold transfer before it degrades ranking.

Outlier detection in decentralized data environments is a challenging task for many machine learning implementations, particularly in settings where data cannot be shared. Recently, there have been advances in federated outlier detection, some of which are based on the use of autoencoder networks. The introduction of attention mechanisms to autoencoders boosts their efficiency. However, the application of attention-based models in federated learning remains underdeveloped due to the absence of proper aggregation functions for these types of networks. In our work, we propose two novel aggregation functions tailored for attention-based autoencoders, which better preserve the learned information stored within the memory modules of these networks. We evaluated our approach on the KDDCUP10 dataset, and we showed that the proposed methods achieve up to 2.9% and 5.1% better results for F1 score and AUC ROC respectively when compared to traditional autoencoders.

Conformal changepoint localization turns any score into a confidence set for the changepoint with finite-sample coverage. Coverage is universal; efficiency is not. The oracle score is a likelihood ratio, so practical scores estimate density ratios, and set length deteriorates under heavy tails, skewness, and distribution shift, where no length guarantee applies. We propose ARC (Augmented-Rank Conformalization), a family of scores depending on the data only through within-segment ranks: rank-CUSUM location and scale channels, their fixed combinations, and a lightweight neural score frozen after synthetic training. Every ARC score inherits finite-sample coverage for every frozen weight configuration, including random initialization and mistraining. The main result is an efficiency transfer theorem: the entire ARC confidence set is almost surely invariant under strictly increasing marginal transforms, so the set length distribution depends on the data pair only through its rank structure, and lengths certified once hold verbatim across its monotone orbit, whereas a plug-in score's length changes with every re-expression. Across different rank structures lengths do change, and are reported as such. Classical rank-test theory positions ARC as targeting the optimal invariant score at bounded cost. Simulations confirm nominal coverage for all scores, including sabotaged networks, identical sets under monotone transforms where plug-in scores inflate, and smooth degradation where plug-in sets become vacuous; on the well-log benchmark ARC localizes annotated shifts to three to five candidates and flags misfit by an empty set. Two boundaries are stated rather than hidden: serial dependence destroys exactness, and trend-type alternatives lie outside the piecewise-exchangeable model.

Joint Energy-Based Models (JEM) unify classification and generation within a single network and support out-of-distribution (OOD) detection. Canonical JEM training relies on stochastic gradient Langevin dynamics (SGLD); a theoretically motivated alternative, the Predictor-Corrector (PC) sampler, has not previously undergone a systematic replication test on the canonical model. We reproduce canonical JEM on WideResNet-28-10 without normalisation layers on two independent runs and test a fixed-noise PC adaptation - with the degenerate annealed-noise predictor replaced by a deterministic gradient step - across three protocols: the adapted sampler replacing SGLD throughout the full training trajectories (115-132 epochs); cold-start generation (FID); and refinement-style multi-OOD detection (AUROC). The reconstruction reaches 92.88% test accuracy and buffer-FID 44.46 (canonical: 92.9% and 38.40). We document two failure modes: catastrophic late-training divergence with the signature of the canonical outlier-buffer mechanism (all four runs), and run-dependent SVHN OOD-discrimination dynamics. No consistent method-level advantage of the adaptation over SGLD is observed on any protocol: refinement AUROC differences stay below 0.007 across ten checkpoint-OOD pairs; seeded cold-start generation favours SGLD by about five FID points; on the training protocol a hierarchical seed-by-image bootstrap gives a 95% confidence interval on the macro-averaged AUROC difference that contains zero, while a seed-level equivalence test with two runs per method cannot establish formal equivalence. The training-protocol data are consistent both with equivalence and with a small directional effect. This outcome is consistent with theory: the guarantees of the annealed-noise PC framework do not transfer to the constant-noise regime of canonical JEM.

Zero-shot visual anomaly detection has achieved remarkable progress, with recent vision-only approaches further improving performance while simplifying the inference pipeline. However, existing methods typically perform dense computation over all images and spatial tokens, despite the fact that normal samples dominate real-world scenarios and anomalies usually occupy only small regions. Token pruning offers a promising solution, but introduces an asymmetric pruning risk in anomaly detection: retaining normal tokens mainly incurs redundant computation, whereas removing anomalous tokens may eliminate the only evidence for detection and localization. This risk is particularly severe in early layers, where pruning provides the greatest computational benefit but anomaly semantics remain unreliable. We propose KeepAD, a defect-preserving token pruning framework that formulates token selection as high-recall, anomaly-aware routing. In shallow layers, KeepAD combines coverage-preserving selection over local $2\times2$ patch neighborhoods with deterministic anomaly rescue to reduce the risk of discarding subtle defects. In deeper layers, frozen normal and abnormal prototypes guide pruning under an image-adaptive token budget, aggressively removing low-risk normal tokens while preserving local anomaly evidence. Dense-to-sparse self-distillation further supervises early token routing without introducing additional inference overhead. Experiments on six industrial and seven medical zero-shot anomaly detection benchmarks show that KeepAD reduces the token retention ratio to below $20\%$, while limiting the average degradation in image-level and pixel-level AUROC to within $2.7$ percentage points. At the most aggressive operating point, KeepAD achieves a $7.9\times$ speedup over the strongest CLIP-based baseline.

Detecting a change in a multivariate series answers only the first of two questions; the operational question is which coordinates changed. Existing answers are incomplete. Block-level procedures certify predefined groups of coordinates under an additive union bound, high-dimensional variable-selection methods return interpretable rankings without error guarantees, and the post-detection inference literature controls error along the time axis rather than across coordinates. We propose ARM (Attribution by Rank Maxima), a wrapper that accepts a changepoint located by an arbitrary detector and returns the set of coordinates certified to have changed, each carrying a location or scale type label. ARM scores each coordinate by a max-over-splits rank statistic. Because this statistic dominates the corresponding statistic at the estimated split, the resulting certificate is invariant to the manner, and to the accuracy, of the changepoint estimate. Three finite-sample guarantees follow from within-coordinate ranks alone: per-coordinate validity under any detector; exact family-wise error control through a Westfall--Young joint permutation that preserves cross-coordinate dependence, with a fully distribution-free Holm fallback; and false discovery rate control under arbitrary coordinate dependence in high dimensions through Benjamini--Yekutieli and e-BH. In simulations, naive per-coordinate testing at the estimated changepoint inflates its family-wise error beyond $0.66$ as the dimension grows, whereas ARM maintains the nominal level while retaining validity under heavy tails, power in high dimensions, and accurate type labels. On five financial series surrounding the 2008 collapse, ARM attributes a scale change to every asset class and excludes injected control coordinates.

Medical anomaly detection identifies abnormal images and localizes lesions under scarce supervision while generalizing across organs and modalities. Existing CLIP-based methods reduce annotation requirements through vision--language alignment, but their normal and abnormal references, whether text prompts or learned visual tokens, remain fixed across test images. Such static references may not transfer reliably to unseen targets in a cross-domain medical imaging scenario. To address this, we propose ReCAP, a language-free framework that replaces static anchors with input-conditioned visual prototypes. ReCAP re-centers separated normal and abnormal prototypes for each image through a bounded gated modulation, enabling query-adaptive anomaly scoring while constraining context-induced prototype drift. For the few-shot setting, we introduce a non-parametric normal-reference memory to preserve instance-level target-domain variation and complement the conditional prototype branch. Across six medical benchmarks, ReCAP achieves the best image-level AUROC on all zero-shot and 23 of 24 few-shot settings, and the best zero-shot pixel-level AUROC on all three segmentation datasets. Particularly, it reduces inference latency by over 70% compared to the fastest baseline, without text prompts or test-time gradient updates.

A marketplace review photograph is a document: platforms approve refunds on it, and generative models drove the cost of forging one to zero. We study that detection problem, so we build a detector and attach an attribution map as its evidence, then measure what that pair delivers on 186,527 images under controls designed to change our conclusions when something is wrong. Compression history, not synthesis, drives naive evaluation: our strongest model reaches 0.9999 PR-AUC (area under the precision-recall curve) on a product-disjoint split, yet falls to 0.7254 once we re-encode synthetics into the real class's format, while five public detectors move by at most 0.07. Aligning one class relocates the cue rather than removing it, and the repaired model then assigns native files a median probability of synthesis of 0.0004. One identical final encode for both classes repairs that, and a three-seed factorial credits the encoding change with the whole gain (+0.176 +- 0.009 PR-AUC). That encode equalises the last stage only: forensic features alone still separate the classes at 0.7145 against a base rate of 0.254. For evidence we test maps causally, against controls that never consult the detector. Whether an attribution ranking exists at all depends on whether the detector reacts to the image. On our first-fix detector, which calls 96 of 100 edited frames real, no map beats a random one. On the detector we selected, twelve of seventeen maps clear that control on edited images and eight on generated ones; perturbation leads both axes and no gradient-CAM variant shows a positive advantage. The trivial controls never clear it, and on generated images the centre prior is worse than random. Our ensembled regional map clears both axes and takes the top pixel AP at 12.4 s per map against 44.9 for occlusion. Clearing a detector-blind control is not yet a faithful explanation, and we demonstrate none.

Glaucoma is a group of eye diseases that damage the optic nerve, often caused by elevated intraocular pressure. It is a leading cause of irreversible vision loss and is typically developed slowly and painlessly, making it difficult to notice until significant damage has occurred. Therefore, early detection is crucial to prevent or slow the progression of vision loss. In recent years, deep learning based uni-modal models have improved the accuracy and efficiency of glaucoma detection, empowering doctors with tools for earlier diagnosis, better monitoring, and timely treatment. Building on this, multi-modal models have emerged, leveraging the strengths of different imaging modalities to learn richer and more robust representations, further enhancing glaucoma detection accuracy. However, multi-modal learning faces challenges such as imbalanced and under-optimized uni-modal representations due to joint learning objectives. To address this, we propose a balanced soft mixture-experts model with three experts and load balancing loss. The performance is measured by AUC, our proposed method surpasses the performance of all uni-modal baselines, conventional multi-modal models, and current stateof- the-art balanced multi-modal models. The proposed model can be generalized to other disease detections such as diabetic retinopathy.

Static malware detectors are commonly evaluated using clean-sample metrics such as accuracy, F1, ROC AUC, and PR AUC. However, these metrics provide limited insight into how learned malware representations behave when feature vectors are perturbed, how close samples move toward uncertain decision regions, or whether compressed representations preserve security-relevant structure. This paper presents a latent-stability analysis pipeline for malware perturbation assessment in EMBER feature space. The pipeline compares full EMBER features, PCA-based compression, beta/denoising variational autoencoder representations, Mandelbrot-inspired escape-time descriptors, and a PINN-style latent-flow module. We define Latent Escape Divergence (LED) to measure changes in escape-time profiles under perturbation, and use PINNFlow-derived residual, velocity, risk, and gradient-shift metrics to characterize latent movement. Experiments are conducted on EMBER static PE feature vectors using 180,000 training samples, 180,000 test samples, and 240,000 holdout samples. Full EMBER features achieve the strongest clean classification performance with ROC AUC of 0.9962 and F1 of 0.9713, while PCA-64 is the strongest compressed baseline with ROC AUC of 0.9846 and F1 of 0.9347. The proposed VAE+Mandelbrot+PINNFlow representation does not outperform these baselines for clean classification, but it provides additional diagnostic value under controlled feature-space perturbation probes.

Vision-language models (VLMs) deployed on consumer hardware must decide when to answer and when to defer, and that decision depends on having a confidence signal that tracks correctness. A practitioner with a fixed memory budget faces a choice between a small model at full precision, the same small model quantized, and a larger model quantized into the same footprint -- three configurations that push the confidence signal in opposing directions. We measure, on identical inputs, how model scale and 4-bit quantization affect two confidence signals in the Qwen2-VL family: the confidence a model states in natural language, and its own mean token probability over the answer it generates. Across 5,700 predictions spanning six realistic photographic degradations at three severities, we find that scale sharply improves the model's internal uncertainty signal (mean error-detection AUROC 0.80 to 0.98 from 2B to 7B) while its verbalized confidence stays weak and often at chance (mean 0.61 to 0.69): the gap between what the model knows and what it says widens rather than closes with size. We find that 4-bit quantization is nearly free for accuracy (-1.6 points) but expensive for the confidence signal (internal AUROC 0.95 to 0.80, and the verbalized-confidence parse rate collapses from 99% to 64%). For a fixed memory budget the recommendation is therefore to prefer a larger quantized model over a smaller full-precision one: 7B-4bit gives both the best accuracy and the best uncertainty signal (internal AUROC 0.98) of the three configurations that fit. We frame the results as selective-prediction operating points so they translate directly into a deployment recommendation, and we argue that error-detection AUROC, not calibration error, is the metric that exposes the difference between the two signals.

Electrocardiograms (ECGs) are widely used for cardiovascular risk prediction, yet models often fail to transfer across hospitals because of protocol, population, and measurement differences. We benchmark cross-dataset generalization on three tasks - heart failure classification, 30-day all-cause mortality, and 30-day mortality among sinus-rhythm ECGs - using two large cohorts (MIMIC-IV and the Alberta Cohort). To reduce vendor-specific measurement mismatch, we build a harmonized, interpretable feature representation computed directly from raw waveforms: FeatureDB morphology/heart-rate-variability summaries plus compact time-frequency descriptors (autoregressive and wavelet features). We train XGBoost models on this unified feature space and evaluate with patient-disjoint internal and bidirectional external testing. We pre-specify two hypotheses: (H1) external AUROC retains at least 90% of source-site internal AUROC under transfer, and (H2) internal AUROC of the harmonized feature set stays within 10% of dataset-native machine-measurement models. Across tasks, internal AUROC is 0.79-0.82 and cross-dataset AUROC is 0.74-0.78, with larger and direction-dependent AUPRC shifts under transfer. As an exploratory benchmark, an end-to-end ConvNeXt model trained directly on raw ECG waveforms with age and sex achieves higher internal AUROC, while the harmonized representation remains competitive in relative cross-dataset transfer stability. These findings show that a consistent waveform-derived feature interface preserves performance, supports realistic external validation, and provides a transparent alternative for cross-site clinical prediction.

Machine learning models for clinical prediction tasks, such as in-hospital mortality and sepsis onset, routinely achieve high AUROC scores. However, AUROC measures ranking performance rather than clinical sensibility. A model may rank patients correctly overall while predicting a lower mortality risk when a patient's SOFA score worsens, contradicting established medical knowledge. This paper proposes applying metamorphic testing (MT) to clinical machine learning models to evaluate behavioral correctness without requiring ground-truth labels for individual predictions. We design a catalog of 12 candidate metamorphic relations (MRs) for three ICU prediction tasks using the MIMIC-III and MIMIC-IV datasets, with each MR grounded in an authoritative clinical guideline. We further propose a five-layer validation strategy to ensure that MRs are clinically sound before deployment. As a feasibility study, we evaluate the approach on the UCI Heart Disease dataset. Although the three clinical models achieve strong predictive performance (AUROC = 0.849-0.900), they exhibit MT violation rates ranging from 27% to 87% across five pilot MRs. An injected-fault experiment further shows that a sign-negation error in a blood pressure feature remains undetected by AUROC but increases the MT violation rate by 31-67 percentage points. These findings suggest that metamorphic testing provides a valuable complement to conventional performance metrics for assessing the behavioral correctness of clinical prediction models.

Anomaly detection is inherently characterised by severe class imbalance, making the interpretation of evaluation metrics challenging. Although metrics such as AUROC, AUPR, F1-score, and MCC are widely used, their values convey different meanings depending on the anomaly ratio. In this work, we analyse the behaviour of those four common anomaly detection metrics under varying levels of imbalance. We focus on the study of metric landscapes, visualisations that relate metric values to true positive and true negative rates, providing an intuitive view of metric preferences and stability. Our analysis offers practical guidance for interpreting and comparing anomaly detection results across datasets with different imbalance ratios.

High task performance does not show whether a model retains prediction-relevant structural information in its internal representation. Temporal graph models, for example, can achieve high future-link AUC while basic graph statistics remain difficult to recover from the same representation. We identify one source of this gap in the weighted averaging used by standard attention: when an evidence pattern is repeated, the numerator and denominator grow at the same rate, so inputs with different amounts of accumulated evidence can produce the same aggregate. We propose Mass-Aware Attention (MAA), which generalizes standard L1 normalization to an Lp family. Under repetition, MAA makes the numerator and denominator scale at different rates, retaining the effective number of contributing inputs in the representation magnitude. It adds no supervision, parameters, hidden dimensions, or explicit count features, and recovers standard attention at p=1. Across four continuous-time dynamic graph models and three datasets, MAA improves future-link AUC in 11 of 12 model-dataset cells. Linear recovery from the same hidden representation increases by 4.49% on average, and preferential-attachment recovery improves in all 12 cells after family-wise correction. We also observe consistent evidence in marked temporal point processes, temporal knowledge graphs, retrieval-augmented generation, and spatio-temporal point processes. Information accessibility and task utility remain distinct: NLL improves in MTPP, ranking is largely preserved in TKG, additional information in RAG does not improve the diagnostic head, and downstream LayerNorm can erase the signal in STPP. These results position MAA as a general normalization principle for improving predictor-facing representation informativeness by controlling repetition invariance in standard attention.

Concept Bottleneck Models provide interpretable-by-design predictions by mediating diagnosis through human-understandable concepts, but in medical imaging, their trustworthiness is often limited by the quality and granularity of available supervision. In particular, predicted concept activations can be driven by irrelevant regions, leading to spatially unfaithful explanations. We study a data-centric spatially grounded Concept Bottleneck Model (SG-CBM) that leverages coarse lesion delineations as weak supervision to encourage anatomically plausible concept evidence. For breast ultrasound, we derive two clinically motivated zones from each lesion mask: (i) an in-lesion region of interest for morphology-related concepts and (ii) a posterior acoustic band for posterior phenomena. We train concept maps using a grouped spatial grounding objective and preserve semantic faithfulness with a linear bottleneck classifier. Across five-fold stratified group cross-validation, the proposed SG-CBM improves diagnostic AUROC and concept macro-AUROC while markedly increasing spatial alignment of concept evidence. We also perform a Train-corrupt/Test-clean annotation-quality stress test to quantify the impact of supervision quality on diagnosis and spatial faithfulness. Overall, the results underscore the need for data-quality-aware supervision design and systematic trustworthiness validation for deployable healthcare AI systems.

Machine learning models for medical image analysis typically lack a reliable measure of confidence, limiting their use in ambiguous or atypical cases. Here we show that Monte Carlo dropout, applied to a multi-task chest-radiograph classifier (eight thoracic findings, 137,593 training images), provides an epistemic uncertainty signal that tracks generalisation across training-set scales and flags confident yet error-prone predictions. Adding this signal to the point prediction raised error-detection AUROC from 0.74 to 0.77 ($Δ$AUROC +0.023, 95% CI [+0.014, +0.033]). In a controlled 2x2 factorial experiment, a clinical-decision-support agent exploited this uncertainty only when it was delivered as a binary error-risk flag rather than as raw scores, cutting confident misdiagnoses on unreliable findings from 8.5% to 2.7%. Epistemic uncertainty estimation thus carries decision-relevant information beyond point predictions, but its value for downstream agents depends on how it is communicated.

Accurate emergency triage decision is critical to avoid clinical deterioration, morbidity, and mortality. Machine learning-based triage system involves acquiring the main presenting complaint in text form and assessing vital signs in numerical data, enabling an automated and efficient analysis of patient information for timely and accurate prioritization of medical attention. However, modelling the intricacies of both data types requires a comprehensive understanding of the temporal structure and dependencies within the data. Thus, the aim of this study is to propose a multimodal deep learning architecture that can effectively handle both tabular and textual data. Furthermore, the proposed model exploits self-attention to to capture both local and global relationships between the features. A dataset consisting of 11,102 triage data collected from emergency department of Hospital Universiti Sains Malaysia is used for model development and validation. The proposed model demonstrated an increase of 1.95% in accuracy, 2.49% in F1-score, and 1.41% in ROC AUC compared to the baseline model. The experimental results demonstrated the potential of the proposed model in predicting triage decisions.

As automated decision-making and data-driven technologies pervade society and are used to manage consequential outcomes, understanding the technology's capabilities, limitations, and attendant risks in context requires analysis of full sociotechnical systems. Sociotechnical analysis of risks in highly complex systems provides clear lessons for the design and evaluation of AI systems, transcending a technical focus on reliable or "responsibly designed" components to understand risks at a systems level. Human-made catastrophes have been studied for decades because of the severity of these events: consider Chernobyl, Three Mile Island, Fukushima-Daiichi, Bhopal, the Challenger disaster. A common misconception is that these kinds of events are freak accidents, resulting from the inherently unforeseeable interactions in complex systems. Closer examination reveals that the risks and hazards were well-known beforehand but not acted upon due to social structural, political and economic factors. We outline several areas where the development and use of AI can benefit from learning these unlearned lessons: improved risk perception, communication, and analysis at the organizational level; traceability of requirements and responsibilities; and holistic approaches to responsibility and safety that include social and organizational dynamics as first-order engineering concerns. For each area, we offer concrete unlearned lessons and exemplify how they led to failure in prior accidents as well as examples of how these lessons remain unlearned for modern computing systems, particularly AI.

Deepfake detectors that achieve near-perfect scores on academic benchmarks collapse on real-world content: recent in-the-wild evaluations report AUC drops of 45-50% for state-of-the-art open-source models. We argue this gap is structural: static detectors are trained once against a moving generative frontier. We present BitMind Forensics (BMF), trained through Bittensor SN34, an open adversarial competition that continually refreshes the training distribution. We evaluate one dated export comprising image, general-video, and human-video checkpoints across nineteen public datasets: the canonical face-swap suites (FaceForensics++, Celeb-DF v1/v2/++, DFDC, DFD, UADFV, DF40) and recent in-the-wild and AI-generated-media benchmarks (Sumsub, Deepfake-Eval-2024, WildRF, Community Forensics, AIGCDetectBench, GenImage, AI-GenBench, AIGIBench, RAID, GenVidBench, GenVideo-100K). BMF reaches 0.936 AUC on Sumsub's original images and 0.872 pooled AUC over its full four-condition manipulation battery (1.4M images), staying robust under perturbation (0.855 JPEG, 0.799 downscaled), while GPEN enhancement improves detection (0.996). On Deepfake-Eval-2024, it matches the best commercial detector on images (0.915 vs 0.90) and exceeds it on video (0.822 vs 0.79), far above the best open-source detectors (0.56 and 0.63). It reaches 0.991 AUC on a 21-generator AI-image panel and 0.918 on GenVidBench, and exceeds the FF++-trained frontier on DFDC (0.947 vs 0.843) and Celeb-DF v2 (0.9985 vs 0.956), both contamination-audited, with statistical parity on Celeb-DF++. In a temporal study, successive dated exports improve on held-out media from generators absent from the static baseline's training (image 0.842 to 0.902; video 0.864 to 0.936). Our evaluation harness is public, and at publication the production API serves the exact evaluated snapshot for independent verification.

Perineural invasion (PNI) is a critical prognostic factor in cholangiocarcinoma. Non-invasive prediction from 3D MRI is challenging, demanding models that efficiently capture both fine-grained details and global context. We propose the Multi-window Mixture-of-Head Attention Transformer (MMA-Former), a novel end-to-end 3D architecture featuring a Coarse-Fine Transformer (CFT) structure for parallel multi-scale feature extraction. We advance this structure by integrating a novel Window-Specific Mixture-of-Head attention (WS-MoH) mechanism. Unlike standard Multi-Head Self Attention (MSA), WS-MoH generates a representation for each 3D window and dynamically routes the entire window to specialized or common attention heads. This enables spatially adaptive feature extraction tailored to the local context of each window, enhancing specialization and reducing redundancy without increasing parameters. Evaluated on a retrospective dataset of 168 T1-weighted MRI scans, MMA-Former achieved an AUC of 0.752, outperforming other 3D architectures, including the best CNN (AUC of 0.708) and Transformer baselines (AUC of 0.681).

The Shapley value is a widely used concept in attribution problems, as it uniquely satisfies the axioms of linearity, consistency, equal treatment, and efficiency. Often, the inclusion AUC metric is used to evaluate the quality of player rankings, in order to identify positively participating players. However, it can be established that the Shapley value is not always reliable for this purpose. The core issue lies in its linearity: the Shapley value acts as a linear operator with an excessively large null space, which is likely to contain non-negligible perturbations that remain indistinguishable to the operator. To address this limitation, we explore the design of nonlinear axiomatic attribution methods. Inspired by the least core, which is a popular nonlinear substitute for the Shapley value, we introduce a class of nonlinear attribution methods that retain the remaining necessary axioms. Each method yields a contribution vector that is the unique optimal solution to a minimization problem, which aims to approximate utility functions as faithfully as possible. In terms of the inclusion AUC metric, our experiments demonstrate the potential effectiveness of these methods compared to Shapley value variants that relax only the efficiency axiom. Our code is available at https://github.com/watml/nonlinear-axiom.

Automated "suspicious behavior" flagging is a headline promise of AI surveillance, and the field reports high frame-level ROC-AUC on standard video anomaly detection benchmarks. Those numbers are measured by training and testing on the same camera and scene. We audit what happens when that assumption is dropped. We build an unsupervised normality model from the all-normal training frames of one dataset, using frozen off-the-shelf embeddings (CLIP, DINOv2, ResNet-50, EfficientNet-B0) and a nearest-neighbour distance, and score the test frames of the same and of other datasets. Across 4 real datasets (UCSD Ped1, UCSD Ped2, CUHK Avenue, ShanghaiTech) and 4 backbones, same-dataset AUC averages 0.704 but cross-dataset AUC averages 0.499, which is chance: a detector calibrated on one scene is no better than a coin flip on another, and in several pairs it is below chance. The strongest backbone makes this worse, not better: DINOv2 has the best same-dataset AUC (up to 0.901 on Ped2) and the largest cross-dataset drop. The collapse is not an artefact of the scoring rule: replacing the nearest-neighbour detector with a PaDiM-style Mahalanobis detector reproduces it almost exactly (cross-dataset gap 0.202 versus 0.208). Even at a favourable operating point the false-alarm rate is on the order of 31,931 per hour. We conclude that the benchmark numbers quoted for surveillance anomaly detection describe a calibrated laboratory setting and overstate deployable reliability by a wide margin, and we release the code that reproduces every number.

Unsupervised tabular anomaly detection requires methods that are accurate, robust across heterogeneous datasets, and computationally efficient. Classical statistical detectors are often efficient, but they usually rely on a fixed data view and a single notion of abnormality. Deep anomaly detectors can learn more flexible scoring functions, but they are substantially slower and difficult to tune in unsupervised settings due to the lack of a reliable supervisory signal. We propose RGLD, a randomized global-local density estimator for efficient unsupervised tabular anomaly detection. RGLD combines a global random-feature density branch, which identifies samples in broadly low-density regions, with a local neighbor branch, which detects samples that are weakly supported by nearby observations. Both branches operate over feature-bagged randomized views, allowing RGLD to expose anomaly evidence that may be hidden in any single representation. We conduct experiments on 47 tabular datasets against 23 statistical and deep anomaly detection baselines under fully unsupervised setting. RGLD achieves the strongest dataset-level AUROC performance, ranking 1st in dataset wins, and ranks 2nd in AUPRC wins. RGLD is also faster than all evaluated deep detectors, achieving 50x-580x speedups, and remains competitive with statistical methods in runtime, yielding a favorable accuracy-efficiency tradeoff.

Financial fraud detection in digital banking requires reasoning over multiple heterogeneous event streams -- transactions, login sessions, risk signals -- that individually appear benign but collectively reveal fraudulent patterns. We propose the Multi-Stream Fraud Transformer (MSFT), a unified architecture that encodes each event stream with independent Transformer encoders and fuses their representations through configurable mechanisms. We conduct a systematic ablation study comparing five fusion strategies: concatenation, gated fusion, time-aware positional encoding, cross-stream attention, and a full combination. On a large-scale dataset (10M users, 1.5% fraud rate) with 85M parameter models, we demonstrate that (1) sequence models significantly outperform gradient-boosted trees operating on aggregated features (0.74 vs. 0.99 AUROC), (2) per-stream encoding is essential -- a single-stream Transformer baseline with matched parameter budget reaches only 0.82 AUROC, an 18-point gap that confirms the multi-stream inductive bias is necessary, (3) time-aware positional encoding achieves the highest discrimination (0.9961 AUROC), (4) gated fusion yields the best precision (0.989) suitable for production deployment, and (5) the risk event stream provides the strongest individual signal contribution. We further validate on proprietary production data from a digital banking platform, showing over 22% relative AUROC improvement over the XGBoost baseline.

Headline accuracies on the Tuebingen cause-effect pairs are routinely compared across papers even though each is measured under its authors' own protocol -- different pair subsets, weightings, model-selection, and decision rates. We argue this is the wrong comparison and run the right one: a same-hands re-evaluation in which every method is run by us on the identical 102 pairs, with one strict rule -- no tuning and a decision forced on every pair. As a clean reference point we introduce a deliberately minimal baseline: sorted-conditional compression, which feeds quantized, sorted, first-differenced data to an off-the-shelf compressor (bz2) and has zero fitted parameters. Under the common ruler the ranking differs sharply from the literature. Our baseline reaches 74.7% weighted accuracy (p = 3.7e-7); on the same 100 pairs that SLOPE is evaluated on it scores 76.0%, a 1.2-point gap below the authors' own forced-decision SLOPE (77.2%) that is well inside noise (McNemar p = 0.39). A faithful re-run of RECI lands at 70.7% -- inside the original authors' reported error bar, not the 77.5% often quoted (which we trace to a mis-copied cell). SLOPE's published 82.4% is a decided-subset figure: scoring the authors' own stored output only on the pairs its significance test chose to answer reproduces 81.7%. Under the common ruler the methods cluster in the low-to-mid 70s and the zero-parameter compressor ties the strongest of them. We document the mechanisms that inflate published figures (test-set model selection, significance-gated abstention) and contribute two further results: compression score magnitude is a model-free confounding flag (p = 2.8e-68), and a pre-registered falsification test fails in an instructive way that bounds the method's theoretical interpretation. Code, pre-registrations, and per-pair outputs are released.

Hidden-state probing -- a linear classifier on a frozen vision-language model's internal activations -- has emerged as an attractive evaluation tool for flagging indirect prompt injection (IPI) in multimodal computer-use agents before the agent emits a corrupted action. We argue, on a single-backbone cautionary case study (Qwen2.5-VL-7B on Mind2Web, teacher-forced replay), that a high probing AUC on a clean-vs-attack split is not, on its own, evidence of malicious-content detection. Two post-hoc diagnostics -- a paired-construction scalar baseline on text-side injections, and same-step nuisance-matched visual controls on the overlay surface -- do not license an unqualified malicious-content interpretation of the headline while leaving room for partly-semantic readings. We package the diagnostics as a candidate control set with reporting heuristics for what a high clean-vs-attack AUC does and does not license. Labels are injection-surface-present, not attack success; generalisation beyond this backbone and benchmark is a conjecture.

Detectors for GNSS radio-frequency impairments (jamming, spoofing, multipath) are usually reported with a single AUC measured on the distribution they were tuned on. That number falls once conditions move, and the size of the drop is rarely known in advance because labelled field data is scarce. We ask whether this optimism can be predicted before any out-of-distribution data is seen. On an open, parameter-grounded synthetic testbed with a tunable severity shift, we evaluate thirteen detectors (five physics baselines, full-feature logistic regression and multilayer perceptrons, and single-feature learned controls) across four impairment classes. The optimism gap, the difference between in-distribution and shifted AUC, grows monotonically as the shift deepens (mean Spearman correlation 0.50). It is driven by how many observables a detector uses rather than by whether it is learned, and it varies systematically by class. Centrally, a ridge model built only from in-distribution score statistics predicts the gap for a detector it has never seen (R^2 = 0.47) and for an impairment class it has never seen (R^2 = 0.46); both are significant against a 2000-fold permutation null (p < 0.001) and survive removing the feature that is, by construction, part of the target. The headline findings are synthetic. We then run the pre-registered protocol on three open field corpora: on Jammertest 2024 the cross-detector prediction holds (R^2 = 0.11, p = 0.009), and on SatGrid, whose spoofer power sweep gives a calibrated severity axis, in-distribution AUC overstates higher-severity AUC by up to 0.22 and to the point of sign inversion, with in-distribution AUC and realised gap perfectly rank-correlated (Spearman rho = 1.0). The mechanism survives contact with real data, at smaller magnitude than in simulation. We release the testbed, a software-receiver front end, the ingest adapters and the protocol.

Training-free detectors of AI-generated images promise generator-agnostic deployment without classifier training, yet their reported numbers are rarely compared under a single controlled protocol. We audit two representative training-free scores -- an autoencoder-reconstruction score (AEROBLADE-style) and a noise-perturbation feature-similarity score (RIGID-style) -- plus a naive feature-kNN control, on a common 1,500-image GenImage-derived benchmark spanning seven generators and JPEG compression at quality 70 and 50. The audit yields three cautionary findings. (i) Implementation details masquerade as method differences: replacing the LPIPS backbone (AlexNet -> VGG-16) changes overall AUROC by +0.085, and switching between resize-to-512 and native-resolution preprocessing flips per-generator conclusions by up to 0.38 AUROC. (ii) Score direction is not a property of the method but of its hyperparameters: the RIGID-style score is inverted (AUROC < 0.5) on SD1.5 and Wukong at noise level sigma=0.05, recovers to >0.5 for every generator at sigma=0.01, and collapses to 0.15 at sigma=0.3. (iii) Dataset format bias inflates robustness claims: without unified re-encoding, AUROC under JPEG-50 exceeds the clean condition for the AlexNet-backbone reconstruction score; after bias correction the residual anomaly localizes to a single generator (BigGAN). The audited scores have complementary per-generator failure sets, but naive z-score fusion does not beat the best single score, indicating that exploiting complementarity requires direction-aware combination.

Pose-flow video anomaly detectors are attractive for one-class surveillance because they provide likelihood-based rankings for tracked skeleton windows. However, a single likelihood score may hide multimodal normal behavior and be sensitive to pose-observation noise. We study a frozen-detector setting in which the pose-flow backbone, cached skeleton tracks, and evaluation pipeline are fixed. Reliability-Aware Prototype Calibration (RPC) is a post-hoc score calibration method for this setting. It adds a standardized nearest-prototype deviation in the frozen latent space to the standardized flow score, and uses keypoint confidence only to gate this added geometric evidence. Thus, RPC preserves the original density signal while correcting the ranking with empirical normal-mode structure under pose reliability. Across two frozen pose-flow backbones and four datasets, RPC improves frame-level AUROC in all eight backbone-dataset pairs, with gains ranging from 0.34 to 4.49 percentage points and averaging 2.03 points. Ablation and reliability analyses show that prototype deviation is the main corrective signal, while reliability gating is most useful when pose observations are less trustworthy. These results suggest that lightweight post-hoc calibration can strengthen cached pose-flow systems when retraining or reproducing the full pose pipeline is impractical.

Training-free anomalous sound detection (ASD) scores a test clip against a memory bank of normal embeddings from a frozen pretrained audio encoder. Recent work attributes domain-shift robustness mainly to how frame-level features are pooled over time; the scoring backend applied on top of the pooled embedding has received far less systematic attention. Using a single frozen BEATs encoder on the DCASE 2023 Task 2 development set (all seven machine types), we cross four classical backends -- nearest-neighbor cosine distance, Mahalanobis distance, locally density-normalized kNN, and PCA-subspace reconstruction residual -- with three temporal poolings (mean, GeM, max). Switching the backend moves target-domain AUC by 13.8 points on average (up to 53.8), whereas switching the pooling moves it by only 3.2 points: in this training-free regime, the backend, not the pooling, dominates domain-shift robustness. No backend wins everywhere, but the machine-dependent pattern reproduces on the DCASE 2025 development data (fan, bearing). Exploiting this, we propose a label-free score fusion that z-normalizes each backend with its training-bank self-scores and takes the minimum; it reaches a harmonic-mean target AUC of 63.3% versus 64.4% for the per-machine oracle, surpassing every fixed single backend while preserving source-domain accuracy. We also report a negative result: selecting a backend by source-domain pseudo-validation with proxy outliers fails, because all backends saturate on the proxy task.

Recent advances in generative AI, such as diffusion models and face-swapping tools, have enabled the creation of highly realistic deepfakes, leading to real-world harms including financial fraud and non-consensual explicit content. In response, deepfake detection has become an active research area, with recent methods increasingly focusing on improving generalization to unseen manipulations. This is typically evaluated using the Area Under the ROC Curve (AUC) measured separately across multiple datasets. However, such an evaluation fails to reflect real-world scenarios where detectors face a mixture of data sources and varying artifact types. To address this limitation, we introduce a novel metric, Cross-dataset AUC (Cross-AUC) that averages per-domain AUCs with a measure of prediction polarization for taking into account the robustness to domain shift. The polarization extent is quantified by the Wasserstein Distance between class score distributions. Cross-AUC not only assesses the generalization capabilities of deepfake detectors under domain shifts more realistically, but it is also interpretable as it better explains the reason behind a drop in performance. Experiments performed on seven benchmark datasets demonstrate its practical relevance.

When a bacterial sample is exposed to several antibiotics, not every applied drug necessarily acts: if the organism is resistant to one of them, that drug leaves no morphological trace. The clinically meaningful quantity is therefore not which antibiotics were applied, but which ones were active. We show that these two are sharply decoupled in real E. coli microscopy - naively assuming the applied combination equals the active one is correct only about 37% of the time - yet existing computational tools are ill-suited to recovering the active set. Forward perturbation models such as scGen, CPA, and IMPA are designed to predict appearance from treatment, not the reverse, and inverting them degrades sharply; discriminative image classifiers tend to memorise strain- and batch-specific texture and fail to transfer across experimental replicates. We introduce AURA, which reframes the task as constrained, energy-based inverse attribution. Its central inductive bias is that the active set must be a subset of the applied set; this collapses the candidate space and lets AURA infer the active subset of applied antibiotics by decomposing residual morphology into antibiotic response atoms and selecting the subset with the lowest reconstruction energy, using no strain label at test time. AURA-E adds evidence-aware abstention, withholding a prediction when candidate explanations remain near-equally plausible. On cross-replicate transfer in an E. coli cytological profiling dataset, AURA recovers the active antibiotic combination with 95.47% exact-match accuracy.

Vision-Language-Action (VLA) models enable end-to-end robot control and have garnered widespread attention. However, the memorization of training data inherent to VLA, coupled with the high cost of robotic data acquisition, raises serious concerns regarding data privacy leakage and intellectual property infringement. Membership inference attacks (MIAs) aim to determine whether a given sample belongs to the training set. While representing a significant privacy threat, this attack remains underexplored in the context of VLA models. To bridge this gap, we propose VLALeaks, which is based on attention discrepancies in VLA models. We reveal, for the first time, the privacy vulnerabilities of VLA models. Specifically, it comprises a two-stage process: (1) membership feature extraction, and (2) attack model construction. Experimental results across multiple VLA benchmarks demonstrate that VLALeaks readily reveals membership information and achieves optimal attack AUC and TPR@1%FPR, highlighting the privacy vulnerabilities in current VLA model deployments. Our work is the first systematic study of MIAs on VLA models, aiming to provide insights for secure and trustworthy VLA models.

Deployed knowledge-tracing models are typically frozen after training, yet systematic per-item logit bias arises, from limited per-item expressivity in backbone architectures and from post-deployment shifts in item properties, degrading prediction quality. Global post-hoc calibrators such as Platt scaling, temperature scaling, and isotonic regression improve probability estimates but leave discriminative ability, as measured by AUC, unchanged. This AUC invariance is a structural consequence of monotone score-only transforms; recovering the stranded discrimination requires conditioning on item identity. We propose SLC (State-space Logit Correction), which converts binary observations to Gaussian pseudo-observations via Laplace/IRLS, applies empirical-Bayes shrinkage through a Kalman smoother, and fits an offset-Platt link. The state-space formulation also yields a detectability bound that characterizes the Bernoulli information floor, explaining why temporal tracking provides no benefit at current data densities. Across four datasets, five backbones, and three seeds, SLC improves AUC on all four datasets and NLL on three, with the advantage concentrating on sparse items. Cross-domain controls suggest that the same phenomenon can arise beyond education when the deployed backbone leaves entity-level bias.

Visualization-based malware detection maps raw binary bytes to grayscale images and applies learned visual classifiers, providing an evasion-resistant and disassembly-free alternative to conventional analysis pipelines. However, executable packing remains a critical failure mode: packed binaries produce high-entropy images that obscure the structural patterns these models rely on. Because packing is also prevalent in benign software (e.g., for compression or copy protection), packing state alone is not a reliable indicator of maliciousness, and existing approaches do not address this challenge within a unified supervised framework. We present ViPER, a Vision-based Packing-Aware Encoder for Robust malware detection. ViPER builds on a LoRA-adapted ViT-B/14 backbone with a dual-head architecture that jointly learns malware classification and packing detection. A packing-aware gating mechanism conditions malware predictions on the inferred packing state, enabling distinct decision boundaries for packed and unpacked inputs. To address packing label skew during training, we employ frequency-weighted losses with stratified sampling over joint class-packing strata. Evaluated on 200,000 Windows PE byteplot images, ViPER achieves a balanced accuracy of 0.8521, ROC-AUC of 0.9260, and AUPR of 0.9279, outperforming representative state-of-the-art baselines across all primary metrics, while attaining a packing detection AUC of 0.9949.

AI governance for medical imaging is formalizing: the 2026 ACR-SIIM Practice Parameter recommends local acceptance testing and ongoing drift monitoring, and the ACR Assess-AI registry monitors AI outputs using DICOM metadata for context. We argue that a necessary, currently unmonitored layer sits beneath output metrics: whether incoming studies remain within the acquisition envelope a model was validated on. Using a LUNA16-trained MONAI RetinaNet lung-nodule detector, we test whether acquisition state behaves as a structured, measurable variable. On real paired CT differing only in reconstruction kernel (NLST B30f vs B80f), kernel alone shifted AI-measured diameter and flipped a Fleischner size category in 5.2% (8 of 155) of nodules at fixed patient and acquisition, while detection confidence was unchanged (Wilcoxon p=0.22). Under controlled LIDC-IDRI perturbations the effects dissociated by axis: the noise axis degraded detection confidence (p=5.9e-32, concentrated in nodules under 6 mm) but not measurement, while the frequency/kernel axis corrupted measurement (p=8.6e-13) but not detection. A 4-feature pixel fingerprint recovered reconstruction identity (patient-level AUC about 0.95 on real CT, 0.995 on a QIBA phantom) where the ConvolutionKernel DICOM tag was uninformative (identical labels across reconstructions). The kernel axis transported across four manufacturers (leave-one-vendor-out AUC 0.94-0.98, matching the within-vendor ceiling). Acquisition state thus maps to distinct AI failure modes, frequency content to measurement reliability and noise to detection sensitivity, and is not recoverable from metadata. Acquisition-aware, input-side validation is the missing layer for the acceptance-testing and drift-monitoring requirements now entering imaging-AI accreditation.

We study whether demonstration-curation metrics that detect defective training episodes also improve the downstream behavior-cloning policy that trains on the curated data. On a contact-rich LIBERO pick-and-place benchmark with a controlled structural defect (early gripper release during the carry phase), we find that the two quantities are sharply decoupled. The metric with the highest defect-detection AUROC (0.804) produces the worst curated policy (13.3% task success), while a metric with a substantially lower AUROC (0.638) produces a policy that nearly matches the oracle trained on ground-truth clean data (90.0% vs. 93.3%). We further show that five of the seven metrics we evaluate exploit episode length as a trivial proxy for the defect label, a confound that inflates reported AUROCs to near-perfect values and disappears once episode length is controlled. Across all conditions, the contaminated baseline succeeds on only 3.3% of rollouts, and the two best curation methods close this to within 3 percentage points of the 93.3% oracle ceiling. Our results argue that curation methods should be evaluated by the policy they produce, not the defects they flag, and that any curation benchmark must control for episode length before reporting detection accuracy. We release the testbed, all metric implementations, and the evaluation pipeline.

We ask whether structural properties of intermediate grid states predict whether a symbolic ARC-AGI solver will succeed, framed as a test of conditional mutual information I(X;Y|task) > 0. Across 44,800 runs spanning two architecturally distinct solvers (beam search and Stochastic DFS), 400 ARC tasks, 28 configurations per solver, and both training and evaluation splits, hand-crafted grid descriptors measured at 50% trajectory completion discriminate successful from failed runs within the same task (mean within-task best-feature AUC = 0.885, p < 0.001 under within-task label permutation). Most predictive content lies along a single grid-complexity axis. The result generalizes across solver architectures: a feature selected on one solver predicts success on the other with AUC 0.747-0.762 in all four transfer directions (p < 0.001, leakage controlled). On a pre-registered held-out set of 41 reliable tasks, the frozen feature n_components_final achieves AUC = 0.765 (95% CI [0.717, 0.810], p < 0.001), robust under task-clustered bootstrap resampling and cross-solver task collapsing. The signal is not explained by solver capacity (configuration-residualized AUC = 0.927 and 0.896 for beam search and SDFS, p < 0.001) and is only weakly coupled to score trajectories (R^2 approximately 0). Early stopping at 50% completion reduces beam-search compute by 33.6% while retaining 98.9% of solves; degenerate-trajectory detection reduces SDFS compute by 65.3% with no solve loss. Finally, on 229 of 400 evaluation tasks the DSL primitive library produces no valid transition from the input grid. This 0-step collapse is invariant to search budget and universally failed by beam search, indicating a DSL coverage limitation rather than a search-budget effect.