Cross-Dataset Generalization

Partially manipulated (half-truth) speech, where a short synthesised segment is spliced into an otherwise genuine utterance, is a harder and more realistic forensic threat than the fully synthesised deepfakes that dominate the literature. We present CAFNet, a lightweight (576K-parameter, 2.24 MB) cross-attentive architecture that fuses MFCC, LFCC, and Chroma-STFT features to jointly classify audio as real, fully fake, or half-truth, and regress the temporal boundaries of the synthesised region, at approximately 14 ms CPU latency. A component ablation shows cross-attention fusion is CAFNet's most load-bearing component; a deeply supervised auxiliary classification head from earlier iterations is not, and removing it improves every in-domain metric under 3-seed replication with substantially lower variance. On MLADDC T2+T3 the model reaches 97.55%$\pm$0.69% ternary accuracy and 0.037 s boundary mean absolute error (MAE), to our knowledge, the first reported continuous splice- boundary localisation result on this benchmark. Zero-shot evaluation on two independent benchmarks shows transfer is capability- and corpus-dependent rather than uniform: on Half-Truth Audio Detection dataset (HAD), detection recall reaches 84.9% and ternary classification resolves half-truth correctly on half of true half-truth clips (50.4%), while on PartialSpoof, binary detection stays near chance (AUC 0.5544). We treat this asymmetry, not a single generalization verdict, as the finding. HAD localisation improves in absolute terms but degrades in relative terms, since in-domain localisation improved faster. An architectural change validated purely in-domain thus shifted the cross-corpus transfer profile, evidence that cross-corpus evaluation should accompany, not follow, in-domain architecture decisions.

It is often desirable to generalize medical imaging AI models trained with dense annotations to data acquired from different ultrasound scanners or clinical sites; however, retraining these models with new annotations is often difficult and costly. We examine this challenge in pediatric wrist fracture assessment using point-of-care ultrasound (POCUS), where fractures are common and can be effectively triaged via ultrasound. AI has shown radiologist-level performance for fracture detection, often aided by high-quality bony structure segmentation. However, due to significant domain shifts, models perform poorly on data from other centers or probes, and obtaining segmentation labels across devices is impractical due to manual annotation effort and data privacy concerns. To address this, we propose a target-informed self-supervised pretraining and model-ensemble strategy. Specifically, our approach combines masked image modeling (MIM) and contrastive learning to learn target-domain structural representations without labels, and introduces a confidence-aware infusion head to adaptively integrate predictions. The source dataset, collected with a Philips Lumify probe, contained dense labels, while the target dataset, acquired with a TeleMED portable probe, was unlabeled. The datasets were kept strictly separate throughout the entire process. Our method used labeled source data for supervised training and leveraged target-domain pretraining to improve generalization. On 318 images from 62 pediatric POCUS videos, this approach significantly improved cross-device performance, achieving over 6% Dice improvement on the target domain versus the baseline. These results demonstrate a label-efficient and privacy-preserving approach for cross-device-robust ultrasound AI, offering a framework that can be extended to multi-center studies or federated learning setups.

With the rise of tabular foundation models alongside traditional models still performing well on many tasks, choosing the right model for a tabular dataset remains difficult. We investigate whether dataset meta-features can explain performance gaps between model families on tabular prediction tasks. Using the TabArena benchmark results, we analyze dataset-level performance gaps and relate them to model-agnostic meta-features. After strict statistical tests with false discovery control, we find that (1) for neural network vs. tree gaps, no meta-feature survives false discovery control, (2) for non-foundation vs. foundation model gaps, one association is robust but does not generalize when tested in leave-one-dataset-out prediction, and (3) for TabICLv2 vs. TabPFN-2.6, one robust association also improves held-out prediction. Furthermore, we conduct a leave-one-dataset-out analysis and find that meta-feature predictors fail to improve meaningfully over a simple baseline. Overall, our results show the heterogeneity of tabular datasets and that global meta-feature approaches are not robust enough to offer explanations on the 51 TabArena datasets.

While spatial foundation models have demonstrated impressive performance on standard datasets, a critical question remains: are they truly all-round players capable of generalizing robustly across diverse downstream tasks, arbitrary viewpoints, shifting scene domains, varying input densities, and specific hardware constraints? Answering this overarching question requires a holistic assessment, yet current models are mainly evaluated on specific domains for which they were specifically designed or trained. Such evaluations are intrinsically limited by narrow paradigm coverage, limited scene domains, and arbitrary frame sampling, making it fundamentally difficult to assess their true generalization capabilities. To address this gap, we present SpatialBench, a cross-paradigm, domain-diverse benchmark for spatial foundation models with deterministic sampling. SpatialBench features unprecedented scale and rigorous deterministic design, comprising 19 datasets and 546 scenes across 5 diverse spatial domains. It comprehensively evaluates 41 models across 6 paradigms on 5 task suites under 4 different input density settings. Our extensive evaluation reveals that current models are not yet all-round players, and uncovers crucial insights for future advancement. Specifically, we demonstrate that full-context attention maximizes accuracy while bounded-memory strategies unlock long-sequence scalability. Moreover, our empirical evaluations in challenging embodied and egocentric tasks demonstrate that strict domain alignment and high data quality are far more critical to performance than simple dataset scaling. Furthermore, to address the largest data gap identified in our analysis, we go beyond evaluation by introducing a large-scale dataset, DA-Next-5M, and a strong baseline model, DA-Next, pushing the boundaries of spatial representation learning.

Pathology foundation models (PFMs) have advanced rapidly in recent years and support training classifiers for a range of histopathology tasks. However, their robustness across hospitals remains limited: performance often degrades when training a classifier on data from one hospital and evaluating it on another target hospital. We address this challenge by fine-tuning PFMs with a local maximum mean discrepancy (LMMD) objective that applies to two settings: domain adaptation, where unlabeled target-hospital data is available, and domain generalization, where target-hospital data is unavailable at all. Experiments at both the patch- and slide-level show consistent improvements across multiple PFMs and tasks.

AI for Science (AI4Science) workflows often treat the released dataset as a fixed interface to the underlying system. However, in domains relying on indirect observation, the learner observes a derivative representation produced by multi-stage measurement, reconstruction, and preprocessing pipelines. **We argue that these measurement-to-dataset pipelines are inference components: treating their outputs as given data'' freezes an observation model and obscures uncertainty over feasible pipeline choices.** We identify three failure modes arising from this frozen lens'': (C1) hidden hypothesis space, where the released dataset does not specify the pipeline configuration or its validity conditions; (C2) uncertified transportability, where a pipeline may be documented but its regime of validity is untested, so failures under distribution shift cannot be adjudicated; (C3) ungoverned multiplicity, where many defensible pipelines exist and dispersion is real but not propagated into uncertainty-aware evidence. We stress-test these claims with a large-scale neuroscience empirical audit, finding a survival rate of $\approx 0.0004\%$ under a cross-dataset stability criterion. We call on the AI4Science community to make pipelines computable inference objects via domain-specific Computable Observation Frameworks. This shift enables quantifying pipeline adequacy and stability, converting implicit implementation choices into auditable, reproducible, and cumulative scientific evidence.

Machine learning models for chronic kidney disease (CKD) risk prediction often post strong discrimination scores on internal test sets. Calibration and uncertainty quantification get far less attention, leaving clinicians without reliable information about whether the probability outputs are accurate. We trained five classifiers on the UCI CKD dataset (400 patients, 62.5% CKD prevalence): logistic regression, random forest, XGBoost, SVM with Platt scaling, and Gaussian naive Bayes. We evaluated each across calibration quality, conformal prediction coverage, and an eight-criterion deployment readiness framework. A distributional stress-test applied the best-calibrated variant of each model to the open-access MIMIC-IV demo cohort (97 patients, 23.7% CKD) to assess behaviour under prevalence shift and feature missingness. We measured calibration before and after Platt scaling and isotonic regression using Expected Calibration Error and Brier Score, and quantified uncertainty through split conformal prediction targeting 90% marginal coverage. All five models reached AUROC 1.00 on the UCI test set. Isotonic recalibration reduced internal ECE to 0.000-0.022. On MIMIC-IV, AUROC fell to 0.48-0.58, ECE rose to 0.68-0.76, and conformal coverage dropped from 0.80-0.98 to 0.21-0.25 against a 90% target. No model scored above 4 out of 16 on the deployment readiness checklist. Near-perfect internal performance did not transfer. Calibration stability and conformal coverage should be evaluated on external data before any clinical prediction model moves toward deployment.

Graph Anomaly Detection (GAD) is increasingly shifting to Generalist GAD (GGAD) for cross-domain "one-for-all" detection, but existing GGAD methods predominantly rely on the neighbor consistency principle, falling into the Node-to-Neighbor Consistency Paradigm for anomaly quantification. These methods suffer from complex training pipelines, heavy training data dependency, high computational costs, and unstable cross-domain generalization. To address these limitations, we propose NeighborDiv, a training-free generalist graph anomaly detection framework based on neighbor diversity. Departing from the dominant Node-to-Neighbor Consistency Paradigm, we shift the focus to the Neighbor-to-Neighbor Diversity Paradigm, and uncover that the internal structural dispersion of a node's neighbor set is a powerful, independently discriminative anomaly signal. We quantify neighbor diversity via the variance of inter-neighbor feature similarities, which captures how a node organizes its local graph environment, and operates independently of conventional node-to-neighbor consistency frameworks. Extensive experiments under two standard GGAD evaluation paradigms show NeighborDiv achieves state-of-the-art performance, with relative gains of 10.25% in average AUC and 17.78% in average AP over the second-best baseline under Single-Domain Independent Training (SDIT), and 6.89%/9.58% in AUC/AP under Unified Multi-Domain Training (UMDT), respectively. Notably, NeighborDiv yields zero performance volatility across all datasets, eliminating training-set dependency and establishing a lightweight and highly practical GGAD framework.

Deep learning models for echocardiography segmentation often struggle to generalise across institutions, scanners, and patient populations, where collecting large, consistently annotated datasets is infeasible. Data augmentation is inexpensive and widely used to improve the robustness of deep learning models; however, its role in enhancing cross-dataset generalisability in echocardiography remains insufficiently understood. This study presents a large-scale multi-dataset evaluation of 29 data augmentation techniques and their pairwise combinations for 2D left ventricular segmentation using a U-Net trained on Unity, CAMUS, and EchoNet Dynamic datasets. Each augmentation was explored under several hyperparameter settings and assessed through repeated runs using Dice and IoU in both in-domain and cross-dataset scenarios, with statistical significance quantified via independent t-tests. In-domain accuracy was near-saturated and insensitive to augmentation, whereas cross-dataset performance varied widely. Geometry-based augmentations including affine, shift-scale-rotate, flip, and perspective produced the largest and most consistent gains, while aggressive intensity- and artefact-based transforms often degraded transfer. Moreover, pairwise combinations outperformed individual augmentations mainly when the two transformations were complementary, particularly by improving some difficult domain-shift cases from poor to acceptable performance. These findings provide empirical guidance for designing augmentation policies that improve the robustness and transferability of echocardiography segmentation models.

A main barrier for the deployment of AI radiomic systems in clinical routine is their drop in performance under heterogeneous multicentre acquisition protocols. This work presents a performance-oriented framework for quantifying scan parameter sensitivity of radiomic AI models, while identifying clinically significant parameter regions associated with improved cross-dataset robustness. We formulate a mixed-effects framework for quantifying the influence that clinically relevant acquisition parameters have on models performance, while accounting for subject-level random effects. We have applied our framework to lung cancer diagnosis in CT scans using two independent multicentre datasets (a public database and own-collected data) and several SoA architectures. To evaluate across-database reproducibility, CT parameters have been adjusted using the data collected and tested on the public set. The optimal configuration selected is the current of the X-ray tube >= 200 mA, spiral pitch <= 1.5, slice thickness <= 1.25 mm, which balances diagnostic quality with low radiation dose. These configuration push metrics from 0.79+-0.04 sensitivity, 0.47+-0.10 specificity in low quality scans to 0.90+-0.10 sensitivity, 0.79 +- 0.13 specificity in high quality ones.

TokaMind is a multi-modal transformer (MMT) foundation model pre-trained on tokamak plasma diagnostics data from MAST, where it was shown to outperform CNN-based approaches on fusion benchmarks. We investigate whether its learned representations generalize to physically distinct but structurally analogous domains. Through systematic experimentation across four domains-industrial bearing degradation, NASA CMAPSS turbofan degradation, and two independent power grid PMU datasets-we identify four transfer-favoring characteristics that help explain where TokaMind's pretrained representations are most effective. Power grid synchrophasor data matches this target-domain profile most directly, while industrial degradation datasets demonstrate that TokaMind can still yield useful performance under partial alignment, especially when task design and feature construction expose physically meaningful degradation structure. On the GESL/PNNL 500-event benchmark with provider-aware evaluation, TokaMind achieves test $\text{F1} = 0.837 \pm 0.040$ (3seeds) for severe event classification. Our central finding, however, is not the aggregate score: classification difficulty is structurally determined by provider-level grid topology, not model capacity. In the single-window early-warning regime, TokaMind outperforms a CNN baseline (F10.889 vs.~0.878)--a reversal that disappears as more event windows are provided. Furthermore, Critical Slowing Down (CSD) indicators, used as a confidence gate rather than a classification label, improve F1 from 0.696 to 0.750 at 63% coverage-outperforming the CNN baseline (0.636) at any coverage level. These results establish the first cross-domain validation of TokaMind outside nuclear fusion and propose a transferability framework and revised evaluation protocol for multi-source PMU datasets.

The pursuit of autonomous driving has produced one of the richest sensor data collections in all of robotics. However, its scale and diversity remain largely untapped. Each dataset adopts different 2D and 3D modalities, such as cameras, lidar, ego states, annotations, traffic lights, and HD maps, with different rates and synchronization schemes. They come in fragmented formats requiring complex dependencies that cannot natively coexist in the same development environment. Further, major inconsistencies in annotation conventions prevent training or measuring generalization across multiple datasets. We present 123D, an open-source framework that unifies such multi-modal driving data through a single API. To handle synchronization, we store each modality as an independent timestamped event stream with no prescribed rate, enabling synchronous or asynchronous access across arbitrary datasets. Using 123D, we consolidate eight real-world driving datasets spanning 3,300 hours and 90,000 kilometers, together with a synthetic dataset with configurable collection scripts, and provide tools for data analysis and visualization. We conduct a systematic study comparing annotation statistics and assessing each dataset's pose and calibration accuracy. Further, we showcase two applications 123D enables: cross-dataset 3D object detection transfer and reinforcement learning for planning, and offer recommendations for future directions. Code and documentation are available at https://github.com/kesai-labs/py123d.

Accurate sleep stage classification across datasets remains challenging due to variability in EEG channel montages, sampling rates, recording environments, and subject populations. Although deep learning has shown considerable promise for automated sleep staging, most existing cross-dataset methods rely on one-dimensional EEG signal representations, whereas the use of two-dimensional spectrogram-based inputs within an unsupervised domain adaptation framework has remained largely unexplored. Here, we propose STDA-Net (Spectrogram-based Temporal Domain Adaptation Network), a framework that combines a convolutional neural network (CNN) for spectrogram-based feature extraction, a bidirectional long short-term memory (BiLSTM) module for temporal modeling of sleep dynamics, and a domain-adversarial neural network (DANN) for source-to-target feature alignment without requiring any labeled target-domain data during training. Experiments are conducted on three publicly available datasets Sleep-EDF, SHHS-1, and SHHS-2 under six cross-dataset transfer settings. Results show that the proposed framework achieves an average accuracy of 89.03% and an average macro F1-score of 87.64%, consistently outperforming existing 1D baseline methods in terms of balanced classification performance, with substantially lower variance across five independent runs, indicating improved stability and reproducibility. Overall, these findings demonstrate that 2D spectrogram-based representations, combined with temporal modeling and adversarial domain adaptation, provide a robust and competitive alternative to conventional 1D EEG inputs for cross-dataset sleep staging.

Mechanical equipment forms the critical backbone of modern industrial production, yet domain shift severely limits the generalization of deep learning based fault diagnosis models across different equipment and operating conditions.Inspired by the success of foundation models in achieving zero-shotgeneralization, we propose YOTOnet (You Only Train Once), a novel architecture specifically designed for cross-domain fault diagnosis in mechanical equipment.YOTOnet comprises three core components: (1) a physics-aware Invariant Feature Distiller that extracts domain-agnostic representations using multi-scale dilated convolutions and FFT-based time-frequency fusion,(2) Domain-Conditioned Sparse Experts (DC-MoE) that adaptively route inputs to specialized processors via learned gating without external meta-data, and (3) a dual-head classification system with auxiliary supervision.Extensive validation on five public bearing datasets (CWRU, MFPT, XJTU,OTTAWA, HUST) through 30 cross-dataset protocols demonstrates the superiority of YOTOnet compared with other state-of-the-art methods. Critically, we observe a clear scaling effect-average test F1 improves from 0.5339(1 training dataset) to 0.705 (4 datasets), with a clear gain when moving from 3 to 4 datasets. These findings provide empirical evidence that foundation model principles can enable robust, train-once deployment for industrial fault diagnosis.

In recent years, machine learning has made significant progress in clinical outcome prediction, demonstrating increasingly accurate results. However, the substantial resources required for hospitals to train these models, such as data collection, labeling, and computational power, limit the feasibility for smaller hospitals to develop their own models. An alternative approach involves transferring a machine learning model trained by a large hospital to smaller hospitals, allowing them to fine-tune the model on their specific patient data. However, these models are often trained and validated on data from a single hospital, raising concerns about their generalizability to new data. Our research shows that there are notable differences in measurement distributions and frequencies across various regions in the United States. To address this, we propose a benchmark that tests a machine learning model's ability to transfer from a source domain to different regions across the country. This benchmark assesses a model's capacity to learn meaningful information about each new domain while retaining key features from the original domain. Using this benchmark, we frame the transfer of a machine learning model from one region to another as a domain incremental learning problem. While the task of patient outcome prediction remains the same, the input data distribution varies, necessitating a model that can effectively manage these shifts. We evaluate two popular domain incremental learning methods: data replay, which stores examples from previous data sources for fine-tuning on the current source, and Elastic Weight Consolidation (EWC), a model parameter regularization method that maintains features important for both data sources.

Artificial intelligence (AI) is becoming a clinical tool for prostate pathology, but generalization across variations in sample preparation and preservation over prolonged time periods remains poorly understood. We evaluated GleasonAI, an end-to-end attention-based multiple instance learning model, on an independent validation cohort comprising 10,366 biopsy cores from 1,028 patients across 14 Swedish regions, using archival diagnostic specimens from the ProMort cohorts collected between 1998-2015. The model achieved an overall quadratic-weighted kappa of 0.86 for core-level ISUP grading, comparable to several experienced pathologists and consistent across geographic regions. Notably, performance remained stable across the 17-year collection period, demonstrating robustness to time-related variation in archival material, a property not consistently observed with foundation model-based approaches, with exploratory analysis demonstrating a significant prognostic gradient across AI-assigned grade groups for prostate cancer-specific mortality. These findings support the generalizability of the AI grading model and demonstrate the potential of pathology archives as a large-scale resource for AI development, validation, and retrospective prognostic research.

Accurate bandgap prediction is crucial for semiconductor applications, yet machine learning models trained on computational data often struggle to generalize to experimental bandgap measurements. Challenges related to data fidelity, domain generalization, and model interpretability remain insufficiently addressed in existing evaluation frameworks. To bridge this gap, we introduce RealMat-BaG, a benchmark for assessing model reliability under experimentally relevant conditions. We curate an open-access dataset of experimental bandgaps with aligned crystal structures and compare graph neural networks as well as classical machine learning baselines. Our framework evaluates performance across statistical and domain-based splits, examines transfer from DFT-computed to experimental bandgaps, and analyzes interpretability at both elemental-property and structural levels. Our results reveal the fundamental generalization limitations of current bandgap prediction models and establish a benchmark aligned with experimental measurements for developing more reliable learning strategies for materials discovery.

Developing robust and clinically reliable EEG biomarkers requires evaluation frameworks that explicitly address cross population generalization in multi site settings such as Parkinsons disease (PD) detection. Models trained under i.i.d. assumptions often capture population specific artifacts rather than disease relevant neural structure, leading to poor generalization across clinical cohorts. EEG further amplifies this challenge due to low signal to noise ratio and heterogeneous acquisition conditions. We propose a population aware evaluation framework to assess the robustness and clinical reliability of EEG biomarkers under distribution shift. Using an n gram expansion strategy, we enumerate all cross population train test configurations across five independent cohorts, resulting in 75 directional evaluations. A nested cross validation design with integrated channel selection ensures prospective biomarker identification without population leakage. Results show that cross population transfer is asymmetric and that both accuracy and biomarker stability improve with increasing training population diversity, achieving up to 94.1% accuracy on held out cohorts. A theoretical analysis based on mixture risk optimization and hypothesis space contraction explains these trends, showing that multi population training promotes population robust representations. This work establishes a principled framework for learning robust, generalizable, and clinically reliable EEG biomarkers for multi site biomedical applications.

Nowadays, visual data forgery detection plays an increasingly important role in social and economic security with the rapid development of generative models. Existing face forgery detectors still can't achieve satisfactory performance because of poor generalization ability across datasets. The key factor that led to this phenomenon is the lack of suitable metrics: the commonly used cross-dataset AUC metric fails to reveal an important issue where detection scores may shift significantly across data domains. To explicitly evaluate cross-domain score comparability, we propose Cross-AUC, an evaluation metric that can compute AUC across dataset pairs by contrasting real samples from one dataset with fake samples from another (and vice versa). It is interesting to find that evaluating representative detectors under the Cross-AUC metric reveals substantial performance drops, exposing an overlooked robustness problem. Besides, we also propose the novel framework Semantic Fine-grained Alignment and Mixture-of-Experts (SFAM), consisting of a patch-level image-text alignment module that enhances CLIP's sensitivity to manipulation artifacts, and the facial region mixture-of-experts module, which routes features from different facial regions to specialized experts for region-aware forgery analysis. Extensive qualitative and quantitative experiments on the public datasets prove that the proposed method achieves superior performance compared with the state-of-the-art methods with various suitable metrics.

New geospatial foundation models introduce a new model architecture and pretraining dataset, often sampled using different notions of data diversity. Performance differences are largely attributed to the model architecture or input modalities, while the role of the pretraining dataset is rarely studied. To address this research gap, we conducted a systematic study on how the geographic composition of pretraining data affects a model's downstream performance. We created global and per-continent pretraining datasets and evaluated them on global and per-continent downstream datasets. We found that the pretraining dataset from Europe outperformed global and continent-specific pretraining datasets on both global and local downstream evaluations. To investigate the factors influencing a pretraining dataset's downstream performance, we analysed 10 pretraining datasets using diversity across continents, biomes, landcover and spectral values. We found that only spectral diversity was strongly correlated with performance, while others were weakly correlated. This finding establishes a new dimension of diversity to be accounted for when creating a high-performing pretraining dataset. We open-sourced 7 new pretraining datasets, pretrained models, and our experimental framework at https://github.com/kerner-lab/pretrain-where.

Automatic Bloom's taxonomy classification of assessment questions can substantially reduce instructor workload, but labeling is subjective and teacher-dependent. Prior machine learning (ML) and deep learning (DL) approaches reported strong within-dataset results, yet were rarely evaluated in cross-dataset settings, leaving real-world generalizability unclear; meanwhile, LLM effectiveness for Bloom question classification has not been systematically studied. We evaluated the cross-dataset generalization of existing ML/DL methods and assessed LLMs with multiple prompting strategies on five datasets; the best prompting strategy combined in-context examples with course-specific action verbs. Supervised ML/DL models degraded substantially on unseen datasets, whereas LLMs were more stable, suggesting a robust alternative across diverse educational contexts. Based on the best prompting strategy, we also presented a lightweight UI that supports instructors in automatically classifying large question banks; a usability study indicated low workload and high usability.

Human perception generalizes well across different domains, but most vision models struggle beyond their training data. This gap motivates multi-dataset learning, where a single model is trained on diverse datasets to improve robustness under domain shifts. However, unified training remains challenging due to inconsistencies in data distributions and label semantics. Mixture-of-Experts (MoE) models provide a scalable solution by routing inputs to specialized subnetworks (experts). Yet, existing MoEs often fail to specialize effectively, as their load-balancing mechanisms enforce uniform input distribution across experts. This fairness conflicts with domain-aware routing, causing experts to learn redundant representations, and reducing performance especially on rare or out-of-distribution domains. We propose GEM (Global Expert Mapping), a planner-compiler framework that replaces the learned router with a global scheduler. Our planner, based on linear programming relaxation, computes a fractional assignment of datasets to experts, while the compiler applies hierarchical rounding to convert this soft plan into a deterministic, capacity-aware mapping. Unlike prior MoEs, GEM avoids balancing loss, resolves the conflict between fairness and specialization, and produces interpretable routing. Experiments show that GEM-DINO achieves state-of-the-art performance on the UODB benchmark, with notable gains on underrepresented datasets and solves task interference in few-shot adaptation scenarios.

Pre-training is a general method that is used in a range of deep learning tasks. By first training a model on one task, and then further training on the downstream task used for final evaluation, the model is forced to learn a more general understanding of the input data. While pre-training has been applied to 3D Human Pose Estimation (HPE) previously, the scope of datasets used is typically very limited to some strong benchmarks, like Human3.6M. Therefore, in this project, we expand the scope of an existing 3D HPE scheme to be compatible with additional 2D and 3D HPE datasets, like Occlusion Person. We perform an extensive study on how aspects of 2D pre-training, such as model size, affect downstream performance, and to what extent pre-training can help the model generalize to different datasets. Experimental results show that 2D pre-training consistently outperforms training on 3D data alone, particularly in terms of computational efficiency. Finally, using MPII and Human3.6M, we are able to obtain an MPJPE score of under 64.5mm.

Diabetic retinopathy (DR) is a leading cause of vision impairment worldwide, and automated grading systems play a crucial role in large-scale screening programs. However, deep learning models often exhibit degraded performance when deployed across datasets acquired under different imaging conditions. This study presents a robust dual-resolution deep learning framework for DR grading that integrates attention-based feature fusion with ordinal regression to improve cross-dataset generalization. The proposed method employs two parallel EfficientNet backbones operating at different spatial resolutions to capture complementary retinal features. A learnable attention mechanism adaptively fuses multi-resolution representations, while an ordinal regression formulation based on the cumulative link model (CORAL) explicitly accounts for the ordered nature of DR severity levels. To mitigate domain discrepancies between datasets, a preprocessing strategy combining circular cropping, contrast enhancement, and histogram matching is applied. The model was trained on the APTOS 2019 dataset and evaluated on both an internal validation split and an external Messidor-2 test set. Experimental results demonstrate strong grading performance, achieving a quadratic weighted kappa (QWK) of 0.88 on the APTOS validation set and 0.68 on the unseen Messidor-2 dataset, indicating improved robustness for cross-dataset DR grading applications.

External validation is widely regarded as the gold standard for prognostic model evaluation. In this study, we challenge the assumption that successful external calibration guarantees model generalizability and propose two complementary strategies to improve transportability of prognostic models across cohorts. Using six real-world surgical cohorts from tertiary academic centers, we tested whether successful external calibration depends largely on similarity in covariates and outcomes between training and validation cohorts, quantified using Kullback-Leibler (KL) divergence, with calibration assessed by the Integrated Calibration Index (ICI). From the model-developer's perspective, we trained the "best-on-average" prognostic model by tuning toward a meta-analysis-derived covariate and outcome distribution as an approximation of the broader target population. From the end-user perspective, we proposed a simple measure for cohort outcome similarity to identify, among published models, the one most suitable for a given target cohort in terms of both calibration and clinical utility. External calibration worsened as distributional mismatch increased. Higher KL divergence was associated with higher ICI in both surgery-alone (Spearman $ρ=0.614$, $p=0.004$) and surgery + adjuvant chemotherapy cohorts (Spearman $ρ=0.738$, $p<0.001$). Meta-analysis-informed weighting improved calibration in most settings without materially affecting discrimination, with the clearest benefit when evaluated on the aggregated external population ($p=0.037$). Models developed in more similar cohorts achieved lower ICI in surgery-alone (Spearman $ρ=0.803$, $p<0.001$) and surgery + adjuvant chemotherapy cohorts (Spearman $ρ=0.737$, $p<0.001$), and provided greater clinical utility on DCA.

Do multilingual embedding models encode a language-general representation of proficiency? We investigate this by training linear and non-linear probes on hidden-state activations from seven embedding models (0.3-8B) to predict CEFR proficiency levels from learner texts across nine corpora and seven languages. We compare five probing architectures against a baseline trained on surface-level text features. Under in-distribution evaluation, probes achieve strong performance (Quadratic Weighted Kappa $\approx0.7$), substantially outperforming the surface baseline, with middle layers consistently yielding the best predictions. However, in cross-corpus evaluation performance collapses across all probe types and model sizes. Residual analysis reveals that out-of-distribution probes converge towards predicting uniformly distributed labels, indicating that the learned mappings capture corpus-specific distributional properties (topic, language, task type, rating methodology) rather than an abstract, transferable proficiency dimension. These results suggest that current multilingual embeddings do not straightforwardly encode language-general proficiency, with implications for representation-based approaches to proficiency-adaptive language technology.

Immune checkpoint inhibitors (ICIs) have transformed cancer therapy; yet substantial proportion of patients exhibit intrinsic or acquired resistance, making accurate pre-treatment response prediction a critical unmet need. Transcriptomics-based biomarkers derived from bulk and single-cell RNA sequencing (scRNA-seq) offer a promising avenue for capturing tumour-immune interactions, yet the cross-cohort generalisability of existing prediction models remains unclear.We systematically benchmark nine state-of-the-art transcriptomic ICI response predictors, five bulk RNA-seq-based models (COMPASS, IRNet, NetBio, IKCScore, and TNBC-ICI) and four scRNA-seq-based models (PRECISE, DeepGeneX, Tres and scCURE), using publicly available independent datasets unseen during model development. Overall, predictive performance was modest: bulk RNA-seq models performed at or near chance level across most cohorts, while scRNA-seq models showed only marginal improvements. Pathway-level analyses revealed sparse and inconsistent biomarker signals across models. Although scRNA-seq-based predictors converged on immune-related programs such as allograft rejection, bulk RNA-seq-based models exhibited little reproducible overlap. PRECISE and NetBio identified the most coherent immune-related themes, whereas IRNet predominantly captured metabolic pathways weakly aligned with ICI biology. Together, these findings demonstrate the limited cross-cohort robustness and biological consistency of current transcriptomic ICI prediction models, underscoring the need for improved domain adaptation, standardised preprocessing, and biologically grounded model design.

Subject-exclusive cross-validation is the standard evaluation protocol for facial Action Unit (AU) detection, yet reported improvements are often small. We show that cross-validation itself introduces measurable stochastic variance. On BP4D+, repeated 3-fold subject-exclusive splits produce an empirical noise floor of $\pm 0.065$ in average F1, with substantially larger variation for low-prevalence AUs. Operating-point metrics such as F1 fluctuate more than threshold-independent measures such as AUC, and model ranking can change under different fold assignments. We further evaluate cross-dataset robustness using a Leave-One-Dataset-Out (LODO) protocol across five AU datasets. LODO removes partition randomness and exposes domain-level instability that is not visible under single-dataset cross-validation. Together, these results suggest that gains often reported in cross-fold validation may fall within protocol variance. Leave-one-dataset-out cross-validation yields more stable and interpretable findings

Large language models often achieve strong benchmark gains without corresponding improvements in broader capability. We hypothesize that this discrepancy arises from differences in training regimes induced by data distribution. To investigate this, we design controlled data interventions that isolate distributional effects under fixed training settings. We find that benchmark-aligned data improves narrow evaluation metrics while limiting broader representational development, whereas coverage-expanding data leads to more distributed parameter adaptation and better generalization. We further introduce parameter-space diagnostics based on spectral and rank analyses, which reveal distinct structural signatures of these regimes. Similar patterns are observed across diverse open-source model families, including multimodal models as a key case study, suggesting that these effects extend beyond controlled settings. A case study on prompt repetition shows that not all data artifacts induce regime shifts. These results indicate that benchmark performance alone is insufficient to characterize model capability, and highlight the importance of data distribution in shaping learning dynamics.

Datasets used in immunotherapy response prediction are typically small in size, as well as diverse in cancer type, drug administered, and sequencer used. Models often drop in performance when tested on patient cohorts that are not included in the training process. Recent work has shown that transformer-based models along with self-supervised learning show better generalisation performance than threshold-based biomarkers, but is still suboptimal. We present BioCOMPASS, an extension of a transformer-based model called COMPASS, that integrates biomarkers and treatment information to further improve its generalisability. Instead of feeding biomarker data as input, we built loss components to align them with the model's intermediate representations. We found that components such as treatment gating and pathway consistency loss improved generalisability when evaluated with Leave-one-cohort-out, Leave-one-cancer-type-out and Leave-one-treatment-out strategies. Results show that building components that exploit biomarker and treatment information can help in generalisability of immunotherapy response prediction. Careful curation of additional components that leverage complementary clinical information and domain knowledge represents a promising direction for future research.