Daily Papers

Computer-Aided Design is pivotal in modern manufacturing, yet existing automated methods predominantly rely on open-loop, one-shot generation, creating a mismatch with iterative real-world practices. In this paper, we present IterCAD, a unified multimodal agent framework for closed-loop, interactive CAD generation and editing. We formulate the task as a multi-turn interaction between a multimodal agent and an executable CAD sandbox, covering three tasks: Drawing-to-Code, Text-to-Code, and Interactive Editing. To support this, we develop a data synthesis pipeline incorporating advanced industrial manufacturing features to generate standard-compliant multi-view engineering drawings, complex code-editing tasks, and high-fidelity interaction trajectories. We optimize the agent via progressive SFT followed by geometry-aware reinforcement learning with viable-prefix masking to enhance code executability and geometric fidelity. Finally, we introduce the IterCAD-Bench evaluation suite and propose the Chamfer Distance Tolerance-Recall (CD-TR) curve alongside its AUC-TR metric, establishing a survivor-bias-free standard that unifies code validity and geometric precision. Extensive experiments demonstrate that IterCAD achieves highly competitive performance across multiple benchmarks, significantly outperforming existing approaches in both code executability and geometric precision, while exhibiting superior capabilities in closed-loop iterative refinement.

Bone density prediction via CT scans to estimate T-scores is crucial, providing a more precise assessment of bone health compared to traditional methods like X-ray bone density tests, which lack spatial resolution and the ability to detect localized changes. However, CT-based prediction faces two major challenges: the high computational complexity of transformer-based architectures, which limits their deployment in portable and clinical settings, and the imbalanced, long-tailed distribution of real-world hospital data that skews predictions. To address these issues, we introduce MedConv, a convolutional model for bone density prediction that outperforms transformer models with lower computational demands. We also adapt Bal-CE loss and post-hoc logit adjustment to improve class balance. Extensive experiments on our AustinSpine dataset shows that our approach achieves up to 21% improvement in accuracy and 20% in ROC AUC over previous state-of-the-art methods.

All prior membership inference attacks for fine-tuned language models use hand-crafted heuristics (e.g., loss thresholding, Min-K\%, reference calibration), each bounded by the designer's intuition. We introduce the first transferable learned attack, enabled by the observation that fine-tuning any model on any corpus yields unlimited labeled data, since membership is known by construction. This removes the shadow model bottleneck and brings membership inference into the deep learning era: learning what matters rather than designing it, with generalization through training diversity and scale. We discover that fine-tuning language models produces an invariant signature of memorization detectable across architectural families and data domains. We train a membership inference classifier exclusively on transformer-based models. It transfers zero-shot to Mamba (state-space), RWKV-4 (linear attention), and RecurrentGemma (gated recurrence), achieving 0.963, 0.972, and 0.936 AUC respectively. Each evaluation combines an architecture and dataset never seen during training, yet all three exceed performance on held-out transformers (0.908 AUC). These four families share no computational mechanisms, their only commonality is gradient descent on cross-entropy loss. Even simple likelihood-based methods exhibit strong transfer, confirming the signature exists independently of the detection method. Our method, Learned Transfer MIA (LT-MIA), captures this signal most effectively by reframing membership inference as sequence classification over per-token distributional statistics. On transformers, LT-MIA achieves 2.8times higher TPR at 0.1\% FPR than the strongest baseline. The method also transfers to code (0.865 AUC) despite training only on natural language texts. Code and trained classifier available at https://github.com/JetBrains-Research/learned-mia.

Cardiotoxicity induced by the breast cancer treatments (i.e., chemotherapy, targeted therapy and radiation therapy) is a significant problem for breast cancer patients. The cardiotoxicity risk for breast cancer patients receiving different treatments remains unclear. We developed and evaluated risk predictive models for cardiotoxicity in breast cancer patients using EHR data. The AUC scores to predict the CHF, CAD, CM and MI are 0.846, 0.857, 0.858 and 0.804 respectively. After adjusting for baseline differences in cardiovascular health, patients who received chemotherapy or targeted therapy appeared to have higher risk of cardiotoxicity than patients who received radiation therapy. Due to differences in baseline cardiac health across the different breast cancer treatment groups, caution is recommended in interpreting the cardiotoxic effect of these treatments.

Dual-encoder retrievers depend on the principle that relevant documents should score higher than irrelevant ones for a given query. Yet the dominant Noise Contrastive Estimation (NCE) objective, which underpins Contrastive Loss, optimizes a softened ranking surrogate that we rigorously prove is fundamentally oblivious to score separation quality and unrelated to AUC. This mismatch leads to poor calibration and suboptimal performance in downstream tasks like retrieval-augmented generation (RAG). To address this fundamental limitation, we introduce the MW loss, a new training objective that maximizes the Mann-Whitney U statistic, which is mathematically equivalent to the Area under the ROC Curve (AUC). MW loss encourages each positive-negative pair to be correctly ranked by minimizing binary cross entropy over score differences. We provide theoretical guarantees that MW loss directly upper-bounds the AoC, better aligning optimization with retrieval goals. We further promote ROC curves and AUC as natural threshold free diagnostics for evaluating retriever calibration and ranking quality. Empirically, retrievers trained with MW loss consistently outperform contrastive counterparts in AUC and standard retrieval metrics. Our experiments show that MW loss is an empirically superior alternative to Contrastive Loss, yielding better-calibrated and more discriminative retrievers for high-stakes applications like RAG.

The standard basis of transformer hidden states is a training-free, architecture-general feature basis for detecting concepts and, in language models, steering them; with no learned dictionary. Individual dimensions act as binary registers read one at a time: their signs (+/-1) encode content, their magnitudes strength. A feature is just a subset of dimensions with a consistent sign pattern, read by counting sign agreements. We validate this Bag of Dims (BoD) framework across seven models spanning language, vision, and audio; reading dimensions one at a time loses nothing, as a full-capacity MLP adds zero AUC over per-dim reading. The same per-dimension signs appear in every modality, so they reflect transformer training itself, not the language objective. Sign alone carries predictive content: setting all magnitudes to unity preserves 60-93% top-5 next-token accuracy through the LM head. From a single-token cache (one forward pass per token, no labels) we detect 175 categories at AUC 0.97-0.99 by counting sign agreements, and from random seeds alone discovery scales to 1500 features per model. A trained probe adds only +0.018 AUC and converges to axis-aligned weights: the rotation dictionaries learn buys little. Signs are causally operative: they survive the attention projections, and flipping a concept's sign pattern in the live forward pass suppresses it. Reading and steering are separate roles in the same basis: a concept's reader dimensions are not its writer dimensions. The writer target is just as cheap, the sign of the summed unembedding rows over a few seeds, no training. Injected through the attention output pathway under closed-loop control, it steers concepts into fluent text on four language models (62-92% of twelve concepts). The signs were in the standard basis all along; the open problem is no longer finding the right rotation but cataloging what each dimension encodes.

Transformer-based visual trackers have demonstrated significant progress owing to their superior modeling capabilities. However, existing trackers are hampered by low speed, limiting their applicability on devices with limited computational power. To alleviate this problem, we propose HiT, a new family of efficient tracking models that can run at high speed on different devices while retaining high performance. The central idea of HiT is the Bridge Module, which bridges the gap between modern lightweight transformers and the tracking framework. The Bridge Module incorporates the high-level information of deep features into the shallow large-resolution features. In this way, it produces better features for the tracking head. We also propose a novel dual-image position encoding technique that simultaneously encodes the position information of both the search region and template images. The HiT model achieves promising speed with competitive performance. For instance, it runs at 61 frames per second (fps) on the Nvidia Jetson AGX edge device. Furthermore, HiT attains 64.6% AUC on the LaSOT benchmark, surpassing all previous efficient trackers.

Multilingual watermarking aims to make large language model (LLM) outputs traceable across languages, yet current methods still fall short. Despite claims of cross-lingual robustness, they are evaluated only on high-resource languages. We show that existing multilingual watermarking methods are not truly multilingual: they fail to remain robust under translation attacks in medium- and low-resource languages. We trace this failure to semantic clustering, which fails when the tokenizer vocabulary contains too few full-word tokens for a given language. To address this, we introduce STEAM, a back-translation-based detection method that restores watermark strength lost through translation. STEAM is compatible with any watermarking method, robust across different tokenizers and languages, non-invasive, and easily extendable to new languages. With average gains of +0.19 AUC and +40%p TPR@1% on 17 languages, STEAM provides a simple and robust path toward fairer watermarking across diverse languages.

Multiparametric magnetic resonance imaging (mpMRI) has demonstrated promising results in prostate cancer (PCa) detection using deep convolutional neural networks (CNNs). Recently, transformers have achieved competitive performance compared to CNNs in computer vision. Large-scale transformers need abundant annotated data for training, which are difficult to obtain in medical imaging. Self-supervised learning can effectively leverage unlabeled data to extract useful semantic representations without annotation and its associated costs. This can improve model performance on downstream tasks with limited labelled data and increase generalizability. We introduce a novel end-to-end Cross-Shaped windows (CSwin) transformer UNet model, CSwin UNet, to detect clinically significant prostate cancer (csPCa) in prostate bi-parametric MR imaging (bpMRI) and demonstrate the effectiveness of our proposed self-supervised pre-training framework. Using a large prostate bpMRI dataset with 1500 patients, we first pre-train CSwin transformer using multi-task self-supervised learning to improve data-efficiency and network generalizability. We then finetuned using lesion annotations to perform csPCa detection. Five-fold cross validation shows that self-supervised CSwin UNet achieves 0.888 AUC and 0.545 Average Precision (AP), significantly outperforming four state-of-the-art models (Swin UNETR, DynUNet, Attention UNet, UNet). Using a separate bpMRI dataset with 158 patients, we evaluated our model robustness to external hold-out data. Self-supervised CSwin UNet achieves 0.79 AUC and 0.45 AP, still outperforming all other comparable methods and demonstrating generalization to a dataset shift.

Deepfakes are the result of digital manipulation to forge realistic yet fake imagery. With the astonishing advances in deep generative models, fake images or videos are nowadays obtained using variational autoencoders (VAEs) or Generative Adversarial Networks (GANs). These technologies are becoming more accessible and accurate, resulting in fake videos that are very difficult to be detected. Traditionally, Convolutional Neural Networks (CNNs) have been used to perform video deepfake detection, with the best results obtained using methods based on EfficientNet B7. In this study, we focus on video deep fake detection on faces, given that most methods are becoming extremely accurate in the generation of realistic human faces. Specifically, we combine various types of Vision Transformers with a convolutional EfficientNet B0 used as a feature extractor, obtaining comparable results with some very recent methods that use Vision Transformers. Differently from the state-of-the-art approaches, we use neither distillation nor ensemble methods. Furthermore, we present a straightforward inference procedure based on a simple voting scheme for handling multiple faces in the same video shot. The best model achieved an AUC of 0.951 and an F1 score of 88.0%, very close to the state-of-the-art on the DeepFake Detection Challenge (DFDC).

International Classification of Disease (ICD) coding procedure which refers to tagging medical notes with diagnosis codes has been shown to be effective and crucial to the billing system in medical sector. Currently, ICD codes are assigned to a clinical note manually which is likely to cause many errors. Moreover, training skilled coders also requires time and human resources. Therefore, automating the ICD code determination process is an important task. With the advancement of artificial intelligence theory and computational hardware, machine learning approach has emerged as a suitable solution to automate this process. In this project, we apply a transformer-based architecture to capture the interdependence among the tokens of a document and then use a code-wise attention mechanism to learn code-specific representations of the entire document. Finally, they are fed to separate dense layers for corresponding code prediction. Furthermore, to handle the imbalance in the code frequency of clinical datasets, we employ a label distribution aware margin (LDAM) loss function. The experimental results on the MIMIC-III dataset show that our proposed model outperforms other baselines by a significant margin. In particular, our best setting achieves a micro-AUC score of 0.923 compared to 0.868 of bidirectional recurrent neural networks. We also show that by using the code-wise attention mechanism, the model can provide more insights about its prediction, and thus it can support clinicians to make reliable decisions. Our code is available online (https://github.com/biplob1ly/TransICD)

Modern neural networks rely on generic activation functions (ReLU, GELU, SiLU) that ignore the mathematical structure inherent in scientific data. We propose Neuro-Symbolic Activation Discovery, a framework that uses Genetic Programming to extract interpretable mathematical formulas from data and inject them as custom activation functions. Our key contribution is the discovery of a Geometric Transfer phenomenon: activation functions learned from particle physics data successfully generalize to ecological classification, outperforming standard activations (ReLU, GELU, SiLU) in both accuracy and parameter efficiency. On the Forest Cover dataset, our Hybrid Transfer model achieves 82.4% accuracy with only 5,825 parameters, compared to 83.4% accuracy requiring 31,801 parameters for a conventional heavy network -- a 5.5x parameter reduction with only 1% accuracy loss. We introduce a Parameter Efficiency Score (E_{param} = AUC / log_{10}(Params)) and demonstrate that lightweight hybrid architectures consistently achieve 18-21% higher efficiency than over-parameterized baselines. Crucially, we establish boundary conditions: while Physics to Ecology transfer succeeds (both involve continuous Euclidean measurements), Physics to Text transfer fails (discrete word frequencies require different mathematical structures). Our work opens pathways toward domain-specific activation libraries for efficient scientific machine learning.

We propose a machine learning-based extension of the classical binomial option pricing model that incorporates key market microstructure effects. Traditional models assume frictionless markets, overlooking empirical features such as bid-ask spreads, discrete price movements, and serial return correlations. Our framework augments the binomial tree with path-dependent transition probabilities estimated via Random Forest classifiers trained on high-frequency market data. This approach preserves no-arbitrage conditions while embedding real-world trading dynamics into the pricing model. Using 46,655 minute-level observations of SPY from January to June 2025, we achieve an AUC of 88.25% in forecasting one-step price movements. Order flow imbalance is identified as the most influential predictor, contributing 43.2% to feature importance. After resolving time-scaling inconsistencies in tree construction, our model yields option prices that deviate by 13.79% from Black-Scholes benchmarks, highlighting the impact of microstructure on fair value estimation. While computational limitations restrict the model to short-term derivatives, our results offer a robust, data-driven alternative to classical pricing methods grounded in empirical market behavior.

Large Language Models (LLMs) and multi-agent systems have shown impressive capabilities in natural language tasks but face challenges in clinical trial applications, primarily due to limited access to external knowledge. Recognizing the potential of advanced clinical trial tools that aggregate and predict based on the latest medical data, we propose an integrated solution to enhance their accessibility and utility. We introduce Clinical Agent System (ClinicalAgent), a clinical multi-agent system designed for clinical trial tasks, leveraging GPT-4, multi-agent architectures, LEAST-TO-MOST, and ReAct reasoning technology. This integration not only boosts LLM performance in clinical contexts but also introduces novel functionalities. The proposed method achieves competitive predictive performance in clinical trial outcome prediction (0.7908 PR-AUC), obtaining a 0.3326 improvement over the standard prompt Method. Publicly available code can be found at https://anonymous.4open.science/r/ClinicalAgent-6671.

Many applied problems contain signal that becomes clear only after combining multiple raw measurements. Ratios and rates are common examples. In gradient boosted trees, this combination is not an explicit operation: the model must synthesize it through coordinated splits on the component features. We study whether intra-tree column subsampling in XGBoost makes that synthesis harder. We use two synthetic data generating processes with cancellation-style structure. In both, two primitive features share a strong nuisance factor, while the target depends on a smaller differential factor. A log ratio cancels the nuisance and isolates the signal. We vary colsample_bylevel and colsample_bynode over s in {0.4, 0.6, 0.8, 0.9}, emphasizing mild subsampling (s >= 0.8). A control feature set includes the engineered ratio, removing the need for synthesis. Across both processes, intra-tree column subsampling reduces test PR-AUC in the primitives-only setting. In the main process the relative decrease reaches 54 percent when both parameters are set to 0.4. The effect largely disappears when the engineered ratio is present. A path-based co-usage metric drops in the same cells where performance deteriorates. Practically, if ratio-like structure is plausible, either avoid intra-tree subsampling or include the intended ratio features.

Driver distraction remains a leading cause of traffic accidents, posing a critical threat to road safety globally. As intelligent transportation systems evolve, accurate and real-time identification of driver distraction has become essential. However, existing methods struggle to capture both global contextual and fine-grained local features while contending with noisy labels in training datasets. To address these challenges, we propose DSDFormer, a novel framework that integrates the strengths of Transformer and Mamba architectures through a Dual State Domain Attention (DSDA) mechanism, enabling a balance between long-range dependencies and detailed feature extraction for robust driver behavior recognition. Additionally, we introduce Temporal Reasoning Confident Learning (TRCL), an unsupervised approach that refines noisy labels by leveraging spatiotemporal correlations in video sequences. Our model achieves state-of-the-art performance on the AUC-V1, AUC-V2, and 100-Driver datasets and demonstrates real-time processing efficiency on the NVIDIA Jetson AGX Orin platform. Extensive experimental results confirm that DSDFormer and TRCL significantly improve both the accuracy and robustness of driver distraction detection, offering a scalable solution to enhance road safety.

Transfer learning is a popular method for tuning pretrained (upstream) models for different downstream tasks using limited data and computational resources. We study how an adversary with control over an upstream model used in transfer learning can conduct property inference attacks on a victim's tuned downstream model. For example, to infer the presence of images of a specific individual in the downstream training set. We demonstrate attacks in which an adversary can manipulate the upstream model to conduct highly effective and specific property inference attacks (AUC score > 0.9), without incurring significant performance loss on the main task. The main idea of the manipulation is to make the upstream model generate activations (intermediate features) with different distributions for samples with and without a target property, thus enabling the adversary to distinguish easily between downstream models trained with and without training examples that have the target property. Our code is available at https://github.com/yulongt23/Transfer-Inference.

Labeling social-media data for custom dimensions of toxicity and social bias is challenging and labor-intensive. Existing transfer and active learning approaches meant to reduce annotation effort require fine-tuning, which suffers from over-fitting to noise and can cause domain shift with small sample sizes. In this work, we propose a novel Active Transfer Few-shot Instructions (ATF) approach which requires no fine-tuning. ATF leverages the internal linguistic knowledge of pre-trained language models (PLMs) to facilitate the transfer of information from existing pre-labeled datasets (source-domain task) with minimum labeling effort on unlabeled target data (target-domain task). Our strategy can yield positive transfer achieving a mean AUC gain of 10.5% compared to no transfer with a large 22b parameter PLM. We further show that annotation of just a few target-domain samples via active learning can be beneficial for transfer, but the impact diminishes with more annotation effort (26% drop in gain between 100 and 2000 annotated examples). Finally, we find that not all transfer scenarios yield a positive gain, which seems related to the PLMs initial performance on the target-domain task.

Some recent studies have described deep convolutional neural networks to diagnose breast cancer in mammograms with similar or even superior performance to that of human experts. One of the best techniques does two transfer learnings: the first uses a model trained on natural images to create a "patch classifier" that categorizes small subimages; the second uses the patch classifier to scan the whole mammogram and create the "single-view whole-image classifier". We propose to make a third transfer learning to obtain a "two-view classifier" to use the two mammographic views: bilateral craniocaudal and mediolateral oblique. We use EfficientNet as the basis of our model. We "end-to-end" train the entire system using CBIS-DDSM dataset. To ensure statistical robustness, we test our system twice using: (a) 5-fold cross validation; and (b) the original training/test division of the dataset. Our technique reached an AUC of 0.9344 using 5-fold cross validation (accuracy, sensitivity and specificity are 85.13% at the equal error rate point of ROC). Using the original dataset division, our technique achieved an AUC of 0.8483, as far as we know the highest reported AUC for this problem, although the subtle differences in the testing conditions of each work do not allow for an accurate comparison. The inference code and model are available at https://github.com/dpetrini/two-views-classifier

Selecting LLM-generated code candidates using LLM-generated tests is challenging because the tests themselves may be incorrect. Existing methods either treat all tests equally or rely on ad-hoc heuristics to filter unreliable tests. Yet determining test correctness requires knowing which codes are correct, creating a circular dependency. Our key insight is that we need not determine test correctness at all: test votes should rank, not merely count. What matters is not how many codes pass a test, but whether the test can distinguish correct from incorrect code. We break the circular dependency via leave-one-out evaluation: hold out one test, rank codes by their aggregate scores on all remaining tests, and measure whether the held-out test's pass/fail pattern agrees with this ranking. We formalize this agreement as the leave-one-out AUC~(LOO-AUC) and prove that the expected LOO-AUC is proportional to each test's ability to separate correct code from incorrect code. Building on this, we propose ACES~(AUC ConsistEncy Scoring) with two complementary variants: ACES-C provides closed-form weights that provably approximate the oracle in expectation under a mild assumption on average test quality; ACES-O drops this assumption and iteratively optimizes a differentiable LOO-AUC objective. Both operate solely on the binary pass matrix with negligible overhead, and achieve state-of-the-art Pass@k on multiple code generation benchmarks.

Two-hop QA retrieval splits queries into two regimes determined by whether the hop-2 entity is explicitly named in the question (Q-dominant) or only in the bridge passage (B-dominant). We formalize this split with three theorems: (T1) per-query AUC is a monotone function of the cosine separation margin, with R^2 >= 0.90 for six of eight type-encoder pairs; (T2) regime is characterized by two surface-text predicates, with P1 decisive for routing and P2 qualifying the B-dominant case, holding across three encoders and three datasets; and (T3) bridge advantage requires the relation-bearing sentence, not entity name alone, with removal causing an 8.6-14.1 pp performance drop (p < 0.001). Building on this theory, we propose RegimeRouter, a lightweight binary router that selects between question-only and question-plus-relation-sentence retrieval using five text features derived directly from the predicate definitions. Trained on 2WikiMultiHopQA (n = 881, 5-fold cross-fitted) and applied zero-shot to MuSiQue and HotpotQA, RegimeRouter achieves +5.6 pp (p < 0.001), +5.3 pp (p = 0.002), and +1.1 pp (non-significant, no-regret) R@5 improvement, respectively, with artifact-driven.

Generative diffusion models are increasingly used for medical imaging data augmentation, but text prompting cannot produce causal training data. Re-prompting rerolls the entire generation trajectory, altering anatomy, texture, and background. Inversion-based editing methods introduce reconstruction error that causes structural drift. We propose MedSteer, a training-free activation-steering framework for endoscopic synthesis. MedSteer identifies a pathology vector for each contrastive prompt pair in the cross-attention layers of a diffusion transformer. At inference time, it steers image activations along this vector, generating counterfactual pairs from scratch where the only difference is the steered concept. All other structure is preserved by construction. We evaluate MedSteer across three experiments on Kvasir v3 and HyperKvasir. On counterfactual generation across three clinical concept pairs, MedSteer achieves flip rates of 0.800, 0.925, and 0.950, outperforming the best inversion-based baseline in both concept flip rate and structural preservation. On dye disentanglement, MedSteer achieves 75% dye removal against 20% (PnP) and 10% (h-Edit). On downstream polyp detection, augmenting with MedSteer counterfactual pairs achieves ViT AUC of 0.9755 versus 0.9083 for quantity-matched re-prompting, confirming that counterfactual structure drives the gain. Code is at link https://github.com/phamtrongthang123/medsteer

This study presents the first multi-platform sentiment analysis of public opinion on the 15-minute city concept across Twitter, Reddit, and news media. Using compressed transformer models and Llama-3-8B for annotation, we classify sentiment across heterogeneous text domains. Our pipeline handles long-form and short-form text, supports consistent annotation, and enables reproducible evaluation. We benchmark five models (DistilRoBERTa, DistilBERT, MiniLM, ELECTRA, TinyBERT) using stratified 5-fold cross-validation, reporting F1-score, AUC, and training time. DistilRoBERTa achieved the highest F1 (0.8292), TinyBERT the best efficiency, and MiniLM the best cross-platform consistency. Results show News data yields inflated performance due to class imbalance, Reddit suffers from summarization loss, and Twitter offers moderate challenge. Compressed models perform competitively, challenging assumptions that larger models are necessary. We identify platform-specific trade-offs and propose directions for scalable, real-world sentiment classification in urban planning discourse.

We present a Multimodal Backdoor Defense technique TIJO (Trigger Inversion using Joint Optimization). Recent work arXiv:2112.07668 has demonstrated successful backdoor attacks on multimodal models for the Visual Question Answering task. Their dual-key backdoor trigger is split across two modalities (image and text), such that the backdoor is activated if and only if the trigger is present in both modalities. We propose TIJO that defends against dual-key attacks through a joint optimization that reverse-engineers the trigger in both the image and text modalities. This joint optimization is challenging in multimodal models due to the disconnected nature of the visual pipeline which consists of an offline feature extractor, whose output is then fused with the text using a fusion module. The key insight enabling the joint optimization in TIJO is that the trigger inversion needs to be carried out in the object detection box feature space as opposed to the pixel space. We demonstrate the effectiveness of our method on the TrojVQA benchmark, where TIJO improves upon the state-of-the-art unimodal methods from an AUC of 0.6 to 0.92 on multimodal dual-key backdoors. Furthermore, our method also improves upon the unimodal baselines on unimodal backdoors. We present ablation studies and qualitative results to provide insights into our algorithm such as the critical importance of overlaying the inverted feature triggers on all visual features during trigger inversion. The prototype implementation of TIJO is available at https://github.com/SRI-CSL/TIJO.

We introduce TravelFraudBench (TFG), a configurable benchmark for evaluating graph neural networks (GNNs) on fraud ring detection in travel platform graphs. Existing benchmarks--YelpChi, Amazon-Fraud, Elliptic, PaySim--cover single node types or domain-generic patterns with no mechanism to evaluate across structurally distinct fraud ring topologies. TFG simulates three travel-specific ring types--ticketing fraud (star topology with shared device/IP clusters), ghost hotel schemes (reviewer x hotel bipartite cliques), and account takeover rings (loyalty transfer chains)--in a heterogeneous graph with 9 node types and 12 edge types. Ring size, count, fraud rate, scale (500 to 200,000 nodes), and composition are fully configurable. We evaluate six methods--MLP, GraphSAGE, RGCN-proj, HAN, RGCN, and PC-GNN--under a ring-based split where each ring appears entirely in one partition, eliminating transductive label leakage. GraphSAGE achieves AUC=0.992 and RGCN-proj AUC=0.987, outperforming the MLP baseline (AUC=0.938) by 5.5 and 5.0 pp, confirming graph structure adds substantial discriminative power. HAN (AUC=0.935) is a negative result, matching the MLP baseline. On the ring recovery task (>=80% of ring members flagged simultaneously), GraphSAGE achieves 100% recovery across all ring types; MLP recovers only 17-88%. The edge-type ablation shows device and IP co-occurrence are the primary signals: removing uses_device drops AUC by 5.2 pp. TFG is released as an open-source Python package (MIT license) with PyG, DGL, and NetworkX exporters and pre-generated datasets at https://huggingface.co/datasets/bsajja7/travel-fraud-graphs, with Croissant metadata including Responsible AI fields.

This work characterizes large language models' chain-of-thought generation as a structured trajectory through representation space. We show that mathematical reasoning traverses functionally ordered, step-specific subspaces that become increasingly separable with layer depth. This structure already exists in base models, while reasoning training primarily accelerates convergence toward termination-related subspaces rather than introducing new representational organization. While early reasoning steps follow similar trajectories, correct and incorrect solutions diverge systematically at late stages. This late-stage divergence enables mid-reasoning prediction of final-answer correctness with ROC-AUC up to 0.87. Furthermore, we introduce trajectory-based steering, an inference-time intervention framework that enables reasoning correction and length control based on derived ideal trajectories. Together, these results establish reasoning trajectories as a geometric lens for interpreting, predicting, and controlling LLM reasoning behavior.

While machine learning shows promise for maternal health risk prediction, clinical adoption in resource-constrained settings faces a critical barrier: lack of explainability and trust. This study presents a hybrid explainable AI (XAI) framework combining ante-hoc fuzzy logic with post-hoc SHAP explanations, validated through systematic clinician feedback. We developed a fuzzy-XGBoost model on 1,014 maternal health records, achieving 88.67% accuracy (ROC-AUC: 0.9703). A validation study with 14 healthcare professionals in Bangladesh revealed strong preference for hybrid explanations (71.4% across three clinical cases) with 54.8% expressing trust for clinical use. SHAP analysis identified healthcare access as the primary predictor, with the engineered fuzzy risk score ranking third, validating clinical knowledge integration (r=0.298). Clinicians valued integrated clinical parameters but identified critical gaps: obstetric history, gestational age, and connectivity barriers. This work demonstrates that combining interpretable fuzzy rules with feature importance explanations enhances both utility and trust, providing practical insights for XAI deployment in maternal healthcare.

Developing a face anti-spoofing model that meets the security requirements of clients worldwide is challenging due to the domain gap between training datasets and diverse end-user test data. Moreover, for security and privacy reasons, it is undesirable for clients to share a large amount of their face data with service providers. In this work, we introduce a novel method in which the face anti-spoofing model can be adapted by the client itself to a target domain at test time using only a small sample of data while keeping model parameters and training data inaccessible to the client. Specifically, we develop a prototype-based base model and an optimal transport-guided adaptor that enables adaptation in either a lightweight training or training-free fashion, without updating base model's parameters. Furthermore, we propose geodesic mixup, an optimal transport-based synthesis method that generates augmented training data along the geodesic path between source prototypes and target data distribution. This allows training a lightweight classifier to effectively adapt to target-specific characteristics while retaining essential knowledge learned from the source domain. In cross-domain and cross-attack settings, compared with recent methods, our method achieves average relative improvements of 19.17% in HTER and 8.58% in AUC, respectively.

WHO has declared that more than 2.2 billion people worldwide are suffering from visual disorders, such as media haze, glaucoma, and drusen. At least 1 billion of these cases could have been either prevented or successfully treated, yet they remain unaddressed due to poverty, a lack of specialists, inaccurate ocular fundus diagnoses by ophthalmologists, or the presence of a rare disease. To address this, the research has developed the Hybrid Trio-Network Model Algorithm for accurately diagnosing 12 distinct common and rare eye diseases. This algorithm utilized the RFMiD dataset of 3,200 fundus images and the Binary Relevance Method to detect diseases separately, ensuring expandability and avoiding incorrect correlations. Each detector, incorporating finely tuned hyperparameters to optimize performance, consisted of three feature components: A classical transfer learning CNN model, a two-stage CNN model, and a Siamese Network. The diagnosis was made using features extracted through this Trio-Model with Ensembled Machine Learning algorithms. The proposed model achieved an average accuracy of 97% and an AUC score of 0.96. Compared to past benchmark studies, an increase of over 10% in the F1-score was observed for most diseases. Furthermore, using the Siamese Network, the model successfully made predictions in diseases like optic disc pallor, which past studies failed to predict due to low confidence. This diagnostic tool presents a stable, adaptive, cost-effective, efficient, accessible, and fast solution for globalizing early detection of both common and rare diseases.

While leveraging additional training data is well established to improve adversarial robustness, it incurs the unavoidable cost of data collection and the heavy computation to train models. To mitigate the costs, we propose Guided Adversarial Training (GAT), a novel adversarial training technique that exploits auxiliary tasks under a limited set of training data. Our approach extends single-task models into multi-task models during the min-max optimization of adversarial training, and drives the loss optimization with a regularization of the gradient curvature across multiple tasks. GAT leverages two types of auxiliary tasks: self-supervised tasks, where the labels are generated automatically, and domain-knowledge tasks, where human experts provide additional labels. Experimentally, GAT increases the robust AUC of CheXpert medical imaging dataset from 50% to 83% and On CIFAR-10, GAT outperforms eight state-of-the-art adversarial training and achieves 56.21% robust accuracy with Resnet-50. Overall, we demonstrate that guided multi-task learning is an actionable and promising avenue to push further the boundaries of model robustness.

The research community has long studied computer-assisted pronunciation training (CAPT) methods in non-native speech. Researchers focused on studying various model architectures, such as Bayesian networks and deep learning methods, as well as on the analysis of different representations of the speech signal. Despite significant progress in recent years, existing CAPT methods are not able to detect pronunciation errors with high accuracy (only 60\% precision at 40\%-80\% recall). One of the key problems is the low availability of mispronounced speech that is needed for the reliable training of pronunciation error detection models. If we had a generative model that could mimic non-native speech and produce any amount of training data, then the task of detecting pronunciation errors would be much easier. We present three innovative techniques based on phoneme-to-phoneme (P2P), text-to-speech (T2S), and speech-to-speech (S2S) conversion to generate correctly pronounced and mispronounced synthetic speech. We show that these techniques not only improve the accuracy of three machine learning models for detecting pronunciation errors but also help establish a new state-of-the-art in the field. Earlier studies have used simple speech generation techniques such as P2P conversion, but only as an additional mechanism to improve the accuracy of pronunciation error detection. We, on the other hand, consider speech generation to be the first-class method of detecting pronunciation errors. The effectiveness of these techniques is assessed in the tasks of detecting pronunciation and lexical stress errors. Non-native English speech corpora of German, Italian, and Polish speakers are used in the evaluations. The best proposed S2S technique improves the accuracy of detecting pronunciation errors in AUC metric by 41\% from 0.528 to 0.749 compared to the state-of-the-art approach.

The rapid advancement of deep learning models that can generate and synthesis hyper-realistic videos known as Deepfakes and their ease of access to the general public have raised concern from all concerned bodies to their possible malicious intent use. Deep learning techniques can now generate faces, swap faces between two subjects in a video, alter facial expressions, change gender, and alter facial features, to list a few. These powerful video manipulation methods have potential use in many fields. However, they also pose a looming threat to everyone if used for harmful purposes such as identity theft, phishing, and scam. In this work, we propose a Convolutional Vision Transformer for the detection of Deepfakes. The Convolutional Vision Transformer has two components: Convolutional Neural Network (CNN) and Vision Transformer (ViT). The CNN extracts learnable features while the ViT takes in the learned features as input and categorizes them using an attention mechanism. We trained our model on the DeepFake Detection Challenge Dataset (DFDC) and have achieved 91.5 percent accuracy, an AUC value of 0.91, and a loss value of 0.32. Our contribution is that we have added a CNN module to the ViT architecture and have achieved a competitive result on the DFDC dataset.

Natural-language autoencoders score explanations of hidden activations by reconstruction: an explanation is deemed faithful if the activation can be regenerated from it. The test is structurally insensitive to individual false claims: if flipping a claim does not change the reconstruction, the claim is never penalized. We show the test is passed in two ways, neither faithful. On a released Qwen-2.5-7B verbalizer, explanations reconstruct well above chance while ~2% of specific claims are reconstruction-dependent, so the score tracks gist, not specific facts. Under exact synthetic ground truth, the standard recipe develops co-adapted private codes (false wording the reconstruction depends on) in 5/5 runs, and fixes that leave the target model unchanged do not help. We contribute two audit protocols, the grounded-vs-true cross and the evaluator swap, and RECAP (Readable Encodings via Co-trained Auxiliary Predictors): linear heads trained alongside the target model to keep designated content decodable. On RECAP-trained sandbox models, fresh verbalizers state the designated content truly and the codes vanish, at a +0.001-nat cost. This replicates on a pretrained Pythia-160M: the content becomes reliably probe-decodable, though a fresh verbalizer conveys it only in part (truth 0.44-0.46 vs a near-zero control). For interpretability, high reconstruction does not certify individual claims. For AI safety, RECAP makes designated internal content independently checkable against probes rather than asserted by prose a model can game: an independent probe scores the verbalizer's true claims above its false ones (AUC 0.96, vs 0.82 without RECAP). Against an adversary that edits an explanation to maximize the reconstruction score while lying (suppressing ~87% of its lie penalty), the RECAP probe still flags the lies (AUC 0.95) while the control probe collapses to chance (0.51).

Recent advances in music foundation models have improved audio representation learning, yet their effectiveness across diverse musical traditions remains limited. We introduce CultureMERT-95M, a multi-culturally adapted foundation model developed to enhance cross-cultural music representation learning and understanding. To achieve this, we propose a two-stage continual pre-training strategy that integrates learning rate re-warming and re-decaying, enabling stable adaptation even with limited computational resources. Training on a 650-hour multi-cultural data mix, comprising Greek, Turkish, and Indian music traditions, results in an average improvement of 4.9% in ROC-AUC and AP across diverse non-Western music auto-tagging tasks, surpassing prior state-of-the-art, with minimal forgetting on Western-centric benchmarks. We further investigate task arithmetic, an alternative approach to multi-cultural adaptation that merges single-culture adapted models in the weight space. Task arithmetic performs on par with our multi-culturally trained model on non-Western auto-tagging tasks and shows no regression on Western datasets. Cross-cultural evaluation reveals that single-culture models transfer with varying effectiveness across musical traditions, whereas the multi-culturally adapted model achieves the best overall performance. To support research on world music representation learning, we publicly release CultureMERT-95M and CultureMERT-TA-95M, fostering the development of more culturally aware music foundation models.

Microscopic assessment of histopathology images is vital for accurate cancer diagnosis and treatment. Whole Slide Image (WSI) classification and captioning have become crucial tasks in computer-aided pathology. However, microscopic WSI face challenges such as redundant patches and unknown patch positions due to subjective pathologist captures. Moreover, generating automatic pathology captions remains a significant challenge. To address these issues, we introduce a novel GNN-ViTCap framework for classification and caption generation from histopathological microscopic images. First, a visual feature extractor generates patch embeddings. Redundant patches are then removed by dynamically clustering these embeddings using deep embedded clustering and selecting representative patches via a scalar dot attention mechanism. We build a graph by connecting each node to its nearest neighbors in the similarity matrix and apply a graph neural network to capture both local and global context. The aggregated image embeddings are projected into the language model's input space through a linear layer and combined with caption tokens to fine-tune a large language model. We validate our method on the BreakHis and PatchGastric datasets. GNN-ViTCap achieves an F1 score of 0.934 and an AUC of 0.963 for classification, along with a BLEU-4 score of 0.811 and a METEOR score of 0.569 for captioning. Experimental results demonstrate that GNN-ViTCap outperforms state of the art approaches, offering a reliable and efficient solution for microscopy based patient diagnosis.

Language-image pre-training has demonstrated strong performance in 2D medical imaging, but its success in 3D modalities such as CT and MRI remains limited due to the high computational demands of volumetric data, which pose a significant barrier to training on large-scale, uncurated clinical studies. In this study, we introduce Hierarchical attention for Language-Image Pre-training (HLIP), a scalable pre-training framework for 3D medical imaging. HLIP adopts a lightweight hierarchical attention mechanism inspired by the natural hierarchy of radiology data: slice, scan, and study. This mechanism exhibits strong generalizability, e.g., +4.3% macro AUC on the Rad-ChestCT benchmark when pre-trained on CT-RATE. Moreover, the computational efficiency of HLIP enables direct training on uncurated datasets. Trained on 220K patients with 3.13 million scans for brain MRI and 240K patients with 1.44 million scans for head CT, HLIP achieves state-of-the-art performance, e.g., +32.4% balanced ACC on the proposed publicly available brain MRI benchmark Pub-Brain-5; +1.4% and +6.9% macro AUC on head CT benchmarks RSNA and CQ500, respectively. These results demonstrate that, with HLIP, directly pre-training on uncurated clinical datasets is a scalable and effective direction for language-image pre-training in 3D medical imaging. The code is available at https://github.com/Zch0414/hlip

We propose GE2E-KWS -- a generalized end-to-end training and evaluation framework for customized keyword spotting. Specifically, enrollment utterances are separated and grouped by keywords from the training batch and their embedding centroids are compared to all other test utterance embeddings to compute the loss. This simulates runtime enrollment and verification stages, and improves convergence stability and training speed by optimizing matrix operations compared to SOTA triplet loss approaches. To benchmark different models reliably, we propose an evaluation process that mimics the production environment and compute metrics that directly measure keyword matching accuracy. Trained with GE2E loss, our 419KB quantized conformer model beats a 7.5GB ASR encoder by 23.6% relative AUC, and beats a same size triplet loss model by 60.7% AUC. Our KWS models are natively streamable with low memory footprints, and designed to continuously run on-device with no retraining needed for new keywords (zero-shot).

Existing satellite video tracking methods often struggle with generalization, requiring scenario-specific training to achieve satisfactory performance, and are prone to track loss in the presence of occlusion. To address these challenges, we propose SatSAM2, a zero-shot satellite video tracker built on SAM2, designed to adapt foundation models to the remote sensing domain. SatSAM2 introduces two core modules: a Kalman Filter-based Constrained Motion Module (KFCMM) to exploit temporal motion cues and suppress drift, and a Motion-Constrained State Machine (MCSM) to regulate tracking states based on motion dynamics and reliability. To support large-scale evaluation, we propose MatrixCity Video Object Tracking (MVOT), a synthetic benchmark containing 1,500+ sequences and 157K annotated frames with diverse viewpoints, illumination, and occlusion conditions. Extensive experiments on two satellite tracking benchmarks and MVOT show that SatSAM2 outperforms both traditional and foundation model-based trackers, including SAM2 and its variants. Notably, on the OOTB dataset, SatSAM2 achieves a 5.84% AUC improvement over state-of-the-art methods. Our code and dataset will be publicly released to encourage further research.

Foundational feed-forward visual geometry models enable accurate and efficient camera pose estimation and scene reconstruction by learning strong scene priors from massive RGB datasets. However, their effectiveness drops when applied to mixed sensing modalities, such as RGB-thermal (RGB-T) images. We observe that while a visual geometry grounded transformer pretrained on RGB data generalizes well to thermal-only reconstruction, it struggles to align RGB and thermal modalities when processed jointly. To address this, we propose SEAR, a simple yet efficient fine-tuning strategy that adapts a pretrained geometry transformer to multimodal RGB-T inputs. Despite being trained on a relatively small RGB-T dataset, our approach significantly outperforms state-of-the-art methods for 3D reconstruction and camera pose estimation, achieving significant improvements over all metrics (e.g., over 29\% in AUC@30) and delivering higher detail and consistency between modalities with negligible overhead in inference time compared to the original pretrained model. Notably, SEAR enables reliable multimodal pose estimation and reconstruction even under challenging conditions, such as low lighting and dense smoke. We validate our architecture through extensive ablation studies, demonstrating how the model aligns both modalities. Additionally, we introduce a new dataset featuring RGB and thermal sequences captured at different times, viewpoints, and illumination conditions, providing a robust benchmark for future work in multimodal 3D scene reconstruction. Code and models are publicly available at https://www.github.com/Schindler-EPFL-Lab/SEAR.

Artificial intelligence is revealing what medicine never intended to encode. Deep vision models, trained on chest X-rays, can now detect not only disease but also invisible traces of social inequality. In this study, we show that state-of-the-art architectures (DenseNet121, SwinV2-B, MedMamba) can predict a patient's health insurance type, a strong proxy for socioeconomic status, from normal chest X-rays with significant accuracy (AUC around 0.70 on MIMIC-CXR-JPG, 0.68 on CheXpert). The signal was unlikely contributed by demographic features by our machine learning study combining age, race, and sex labels to predict health insurance types; it also remains detectable when the model is trained exclusively on a single racial group. Patch-based occlusion reveals that the signal is diffuse rather than localized, embedded in the upper and mid-thoracic regions. This suggests that deep networks may be internalizing subtle traces of clinical environments, equipment differences, or care pathways; learning socioeconomic segregation itself. These findings challenge the assumption that medical images are neutral biological data. By uncovering how models perceive and exploit these hidden social signatures, this work reframes fairness in medical AI: the goal is no longer only to balance datasets or adjust thresholds, but to interrogate and disentangle the social fingerprints embedded in clinical data itself.

The development of larger models for medical image analysis has led to increased performance. However, it also affected our ability to explain and validate model decisions. Models can use non-relevant parts of images, also called spurious correlations or shortcuts, to obtain high performance on benchmark datasets but fail in real-world scenarios. In this work, we challenge the capacity of convolutional neural networks (CNN) to classify chest X-rays and eye fundus images while masking out clinically relevant parts of the image. We show that all models trained on the PadChest dataset, irrespective of the masking strategy, are able to obtain an Area Under the Curve (AUC) above random. Moreover, the models trained on full images obtain good performance on images without the region of interest (ROI), even superior to the one obtained on images only containing the ROI. We also reveal a possible spurious correlation in the Chaksu dataset while the performances are more aligned with the expectation of an unbiased model. We go beyond the performance analysis with the usage of the explainability method SHAP and the analysis of embeddings. We asked a radiology resident to interpret chest X-rays under different masking to complement our findings with clinical knowledge. Our code is available at https://github.com/TheoSourget/MMC_Masking and https://github.com/TheoSourget/MMC_Masking_EyeFundus

The proliferation of Large Language Models (LLMs) with varying capabilities and costs has created a need for efficient model selection in AI systems. LLM routers address this need by dynamically choosing the most suitable model for a given query based on task requirements and budget constraints. However, existing routers face challenges in scalability and real-time adaptation, particularly in high-volume online environments. We present Eagle, a novel LLM routing approach that combines global and local ELO ranking modules to overcome these limitations. By evaluating both general and specialized LLM abilities, Eagle provides a scalable, training-free solution that enhances model selection quality while reducing computational overhead. Our experiments across multiple datasets show Eagle consistently outperforms baseline methods, with improvements of up to 23.52 percent in Area Under Curve (AUC) scores. Moreover, Eagle demonstrates remarkable efficiency, requiring only 1/20 of baseline methods' time for initialization and 100 to 200 times faster incremental updates in online scenarios, making it well-suited for dynamic, high-volume online serving environments.

Accurate diagnosis of heart arrhythmias requires the interpretation of electrocardiograms (ECG), which capture the electrical activity of the heart. Automating this process through machine learning is challenging due to the need for large annotated datasets, which are difficult and costly to collect. To address this issue, transfer learning is often employed, where models are pre-trained on large datasets and fine-tuned for specific ECG classification tasks with limited labeled data. Self-supervised learning has become a widely adopted pre-training method, enabling models to learn meaningful representations from unlabeled datasets. In this work, we explore the joint-embedding predictive architecture (JEPA) for self-supervised learning from ECG data. Unlike invariance-based methods, JEPA does not rely on hand-crafted data augmentations, and unlike generative methods, it predicts latent features rather than reconstructing input data. We create a large unsupervised pre-training dataset by combining ten public ECG databases, amounting to over one million records. We pre-train Vision Transformers using JEPA on this dataset and fine-tune them on various PTB-XL benchmarks. Our results show that JEPA outperforms existing invariance-based and generative approaches, achieving an AUC of 0.945 on the PTB-XL all statements task. JEPA consistently learns the highest quality representations, as demonstrated in linear evaluations, and proves advantageous for pre-training even in the absence of additional data.

Whole Slide Image (WSI) analysis is a powerful method to facilitate the diagnosis of cancer in tissue samples. Automating this diagnosis poses various issues, most notably caused by the immense image resolution and limited annotations. WSIs commonly exhibit resolutions of 100Kx100K pixels. Annotating cancerous areas in WSIs on the pixel level is prohibitively labor-intensive and requires a high level of expert knowledge. Multiple instance learning (MIL) alleviates the need for expensive pixel-level annotations. In MIL, learning is performed on slide-level labels, in which a pathologist provides information about whether a slide includes cancerous tissue. Here, we propose Self-ViT-MIL, a novel approach for classifying and localizing cancerous areas based on slide-level annotations, eliminating the need for pixel-wise annotated training data. Self-ViT- MIL is pre-trained in a self-supervised setting to learn rich feature representation without relying on any labels. The recent Vision Transformer (ViT) architecture builds the feature extractor of Self-ViT-MIL. For localizing cancerous regions, a MIL aggregator with global attention is utilized. To the best of our knowledge, Self-ViT- MIL is the first approach to introduce self-supervised ViTs in MIL-based WSI analysis tasks. We showcase the effectiveness of our approach on the common Camelyon16 dataset. Self-ViT-MIL surpasses existing state-of-the-art MIL-based approaches in terms of accuracy and area under the curve (AUC).

Tabular data (or tables) are the most widely used data format in machine learning (ML). However, ML models often assume the table structure keeps fixed in training and testing. Before ML modeling, heavy data cleaning is required to merge disparate tables with different columns. This preprocessing often incurs significant data waste (e.g., removing unmatched columns and samples). How to learn ML models from multiple tables with partially overlapping columns? How to incrementally update ML models as more columns become available over time? Can we leverage model pretraining on multiple distinct tables? How to train an ML model which can predict on an unseen table? To answer all those questions, we propose to relax fixed table structures by introducing a Transferable Tabular Transformer (TransTab) for tables. The goal of TransTab is to convert each sample (a row in the table) to a generalizable embedding vector, and then apply stacked transformers for feature encoding. One methodology insight is combining column description and table cells as the raw input to a gated transformer model. The other insight is to introduce supervised and self-supervised pretraining to improve model performance. We compare TransTab with multiple baseline methods on diverse benchmark datasets and five oncology clinical trial datasets. Overall, TransTab ranks 1.00, 1.00, 1.78 out of 12 methods in supervised learning, feature incremental learning, and transfer learning scenarios, respectively; and the proposed pretraining leads to 2.3% AUC lift on average over the supervised learning.

Deep AUC Maximization (DAM) is a new paradigm for learning a deep neural network by maximizing the AUC score of the model on a dataset. Most previous works of AUC maximization focus on the perspective of optimization by designing efficient stochastic algorithms, and studies on generalization performance of large-scale DAM on difficult tasks are missing. In this work, we aim to make DAM more practical for interesting real-world applications (e.g., medical image classification). First, we propose a new margin-based min-max surrogate loss function for the AUC score (named as AUC min-max-margin loss or simply AUC margin loss for short). It is more robust than the commonly used AUC square loss, while enjoying the same advantage in terms of large-scale stochastic optimization. Second, we conduct extensive empirical studies of our DAM method on four difficult medical image classification tasks, namely (i) classification of chest x-ray images for identifying many threatening diseases, (ii) classification of images of skin lesions for identifying melanoma, (iii) classification of mammogram for breast cancer screening, and (iv) classification of microscopic images for identifying tumor tissue. Our studies demonstrate that the proposed DAM method improves the performance of optimizing cross-entropy loss by a large margin, and also achieves better performance than optimizing the existing AUC square loss on these medical image classification tasks. Specifically, our DAM method has achieved the 1st place on Stanford CheXpert competition on Aug. 31, 2020. To the best of our knowledge, this is the first work that makes DAM succeed on large-scale medical image datasets. We also conduct extensive ablation studies to demonstrate the advantages of the new AUC margin loss over the AUC square loss on benchmark datasets. The proposed method is implemented in our open-sourced library LibAUC (www.libauc.org).

Knowledge tracing (KT) refers to the problem of predicting future learner performance given their past performance in educational applications. Recent developments in KT using flexible deep neural network-based models excel at this task. However, these models often offer limited interpretability, thus making them insufficient for personalized learning, which requires using interpretable feedback and actionable recommendations to help learners achieve better learning outcomes. In this paper, we propose attentive knowledge tracing (AKT), which couples flexible attention-based neural network models with a series of novel, interpretable model components inspired by cognitive and psychometric models. AKT uses a novel monotonic attention mechanism that relates a learner's future responses to assessment questions to their past responses; attention weights are computed using exponential decay and a context-aware relative distance measure, in addition to the similarity between questions. Moreover, we use the Rasch model to regularize the concept and question embeddings; these embeddings are able to capture individual differences among questions on the same concept without using an excessive number of parameters. We conduct experiments on several real-world benchmark datasets and show that AKT outperforms existing KT methods (by up to 6% in AUC in some cases) on predicting future learner responses. We also conduct several case studies and show that AKT exhibits excellent interpretability and thus has potential for automated feedback and personalization in real-world educational settings.

This research introduces a novel AI techniques as Mixture-of-Experts Transformers with Group Relative Policy Optimization (GRPO) for voice health care applications on voice pathology detection. With the architectural innovations, we adopt advanced training paradigms inspired by reinforcement learning, namely Proximal Policy Optimization (PPO) and Group-wise Regularized Policy Optimization (GRPO), to enhance model stability and performance. Experiments conducted on a synthetically generated voice pathology dataset demonstrate that our proposed models significantly improve diagnostic accuracy, F1 score, and ROC-AUC compared to conventional approaches. These findings underscore the potential of integrating transformer architectures with novel training strategies to advance automated voice pathology detection and ultimately contribute to more effective healthcare delivery. The code we used to train and evaluate our models is available at https://github.com/enkhtogtokh/voicegrpo

How can we verify whether copyrighted content was used to train a large vision-language model (VLM) without direct access to its training data? Motivated by the hypothesis that a VLM is able to recognize images from its training corpus, we propose DIS-CO, a novel approach to infer the inclusion of copyrighted content during the model's development. By repeatedly querying a VLM with specific frames from targeted copyrighted material, DIS-CO extracts the content's identity through free-form text completions. To assess its effectiveness, we introduce MovieTection, a benchmark comprising 14,000 frames paired with detailed captions, drawn from films released both before and after a model's training cutoff. Our results show that DIS-CO significantly improves detection performance, nearly doubling the average AUC of the best prior method on models with logits available. Our findings also highlight a broader concern: all tested models appear to have been exposed to some extent to copyrighted content. Our code and data are available at https://github.com/avduarte333/DIS-CO

How can we detect if copyrighted content was used in the training process of a language model, considering that the training data is typically undisclosed? We are motivated by the premise that a language model is likely to identify verbatim excerpts from its training text. We propose DE-COP, a method to determine whether a piece of copyrighted content was included in training. DE-COP's core approach is to probe an LLM with multiple-choice questions, whose options include both verbatim text and their paraphrases. We construct BookTection, a benchmark with excerpts from 165 books published prior and subsequent to a model's training cutoff, along with their paraphrases. Our experiments show that DE-COP surpasses the prior best method by 9.6% in detection performance (AUC) on models with logits available. Moreover, DE-COP also achieves an average accuracy of 72% for detecting suspect books on fully black-box models where prior methods give approx 4% accuracy. Our code and datasets are available at https://github.com/avduarte333/DE-COP_Method

Moral language is subtle and culturally variable, making it difficult to translate faithfully across languages. Idiomatic expressions, slang, and cultural references introduce hard-to-avoid translation artifacts. Yet automated moral values classification depends on language-specific annotated corpora that exist almost exclusively in English. We investigate whether LLM-based translation can bridge this gap, taking Polish as a test case. Using sim50k morally-annotated social media posts from a diverse range of topics, we apply a principled four-method validation pipeline: LaBSE cross-lingual embedding similarity, Centered Kernel Alignment (CKA), LLM-as-judge evaluation, and deep learning classifier parity tests. We show that despite shortcomings in handling slang, vulgarity, and culturally-loaded expressions, direct translation preserves subtle moral cues well enough to be harvested by cross-lingual machine learning -- with mean cosine similarity of 0.86 and AUC gaps of 0.01--0.02 across all foundations closing further under fine-tuning of language models. These results demonstrate that machine translation is a practical and cost-effective path to moral values research in languages currently under-resourced in this domain. We demonstrate this for Polish as a representative Slavic language, with expected generalisation to related languages.

With growing real-world demands, efficient tracking has received increasing attention. However, most existing methods are limited to RGB inputs and struggle in multi-modal scenarios. Moreover, current multi-modal tracking approaches typically use complex designs, making them too heavy and slow for resource-constrained deployment. To tackle these limitations, we propose UETrack, an efficient framework for single object tracking. UETrack demonstrates high practicality and versatility, efficiently handling multiple modalities including RGB, Depth, Thermal, Event, and Language, and addresses the gap in efficient multi-modal tracking. It introduces two key components: a Token-Pooling-based Mixture-of-Experts mechanism that enhances modeling capacity through feature aggregation and expert specialization, and a Target-aware Adaptive Distillation strategy that selectively performs distillation based on sample characteristics, reducing redundant supervision and improving performance. Extensive experiments on 12 benchmarks across 3 hardware platforms show that UETrack achieves a superior speed-accuracy trade-off compared to previous methods. For instance, UETrack-B achieves 69.2% AUC on LaSOT and runs at 163/56/60 FPS on GPU/CPU/AGX, demonstrating strong practicality and versatility. Code is available at https://github.com/kangben258/UETrack.

Underwater Object Tracking (UOT) is crucial for efficient marine robotics, large scale ecological monitoring, and ocean exploration; however, progress has been hindered by the scarcity of large, multimodal, and diverse datasets. Existing benchmarks remain small and RGB only, limiting robustness under severe color distortion, turbidity, and low visibility conditions. We introduce MUOT_3M, the first pseudo multimodal UOT benchmark comprising 3 million frames from 3,030 videos (27.8h) annotated with 32 tracking attributes, 677 fine grained classes, and synchronized RGB, estimated enhanced RGB, estimated depth, and language modalities validated by a marine biologist. Building upon MUOT_3M, we propose MUTrack, a SAM-based multimodal to unimodal tracker featuring visual geometric alignment, vision language fusion, and four level knowledge distillation that transfers multimodal knowledge into a unimodal student model. Extensive evaluations across five UOT benchmarks demonstrate that MUTrack achieves up to 8.40% higher AUC and 7.80% higher precision than the strongest SOTA baselines while running at 24 FPS. MUOT_3M and MUTrack establish a new foundation for scalable, multimodally trained yet practically deployable underwater tracking.

In this research, I explore advanced deep learning methodologies to forecast the outcomes of the 2025 NCAA Division 1 Men's and Women's Basketball tournaments. Leveraging historical NCAA game data, I implement two sophisticated sequence-based models: Long Short-Term Memory (LSTM) and Transformer architectures. The predictive power of these models is augmented through comprehensive feature engineering, including team quality metrics derived from Generalized Linear Models (GLM), Elo ratings, seed differences, and aggregated box-score statistics. To evaluate the robustness and reliability of predictions, I train each model variant using both Binary Cross-Entropy (BCE) and Brier loss functions, providing insights into classification performance and probability calibration. My comparative analysis reveals that while the Transformer architecture optimized with BCE yields superior discriminative power (highest AUC of 0.8473), the LSTM model trained with Brier loss demonstrates superior probabilistic calibration (lowest Brier score of 0.1589). These findings underscore the importance of selecting appropriate model architectures and loss functions based on the specific requirements of forecasting tasks. The detailed analytical pipeline presented here serves as a reproducible framework for future predictive modeling tasks in sports analytics and beyond.

Breast cancer remains a leading cause of cancer-related mortality worldwide, making early detection and accurate treatment response monitoring critical priorities. We present BreastDCEDL, a curated, deep learning-ready dataset comprising pre-treatment 3D Dynamic Contrast-Enhanced MRI (DCE-MRI) scans from 2,070 breast cancer patients drawn from the I-SPY1, I-SPY2, and Duke cohorts, all sourced from The Cancer Imaging Archive. The raw DICOM imaging data were rigorously converted into standardized 3D NIfTI volumes with preserved signal integrity, accompanied by unified tumor annotations and harmonized clinical metadata including pathologic complete response (pCR), hormone receptor (HR), and HER2 status. Although DCE-MRI provides essential diagnostic information and deep learning offers tremendous potential for analyzing such complex data, progress has been limited by lack of accessible, public, multicenter datasets. BreastDCEDL addresses this gap by enabling development of advanced models, including state-of-the-art transformer architectures that require substantial training data. To demonstrate its capacity for robust modeling, we developed the first transformer-based model for breast DCE-MRI, leveraging Vision Transformer (ViT) architecture trained on RGB-fused images from three contrast phases (pre-contrast, early post-contrast, and late post-contrast). Our ViT model achieved state-of-the-art pCR prediction performance in HR+/HER2- patients (AUC 0.94, accuracy 0.93). BreastDCEDL includes predefined benchmark splits, offering a framework for reproducible research and enabling clinically meaningful modeling in breast cancer imaging.

This paper focuses on the task of hallucination detection, which aims to determine the truthfulness of LLM-generated statements. To address this problem, a popular class of methods utilize the LLM's self-consistencies in its beliefs in a set of logically related augmented statements generated by the LLM, which does not require external knowledge databases and can work with both white-box and black-box LLMs. However, in many existing approaches, the augmented statements tend to be very monotone and unstructured, which makes it difficult to integrate meaningful information from the LLM beliefs in these statements. Also, many methods work with the binarized version of the LLM's belief, instead of the continuous version, which significantly loses information. To overcome these limitations, in this paper, we propose Belief Tree Propagation (BTProp), a probabilistic framework for LLM hallucination detection. BTProp introduces a belief tree of logically related statements by recursively decomposing a parent statement into child statements with three decomposition strategies, and builds a hidden Markov tree model to integrate the LLM's belief scores in these statements in a principled way. Experiment results show that our method improves baselines by 3%-9% (evaluated by AUROC and AUC-PR) on multiple hallucination detection benchmarks. Code is available at https://github.com/UCSB-NLP-Chang/BTProp.

With the rapid progress of generative models, the current challenge in face forgery detection is how to effectively detect realistic manipulated faces from different unseen domains. Though previous studies show that pre-trained Vision Transformer (ViT) based models can achieve some promising results after fully fine-tuning on the Deepfake dataset, their generalization performances are still unsatisfactory. One possible reason is that fully fine-tuned ViT-based models may disrupt the pre-trained features [1, 2] and overfit to some data-specific patterns [3]. To alleviate this issue, we present a Forgery-aware Adaptive Vision Transformer (FA-ViT) under the adaptive learning paradigm, where the parameters in the pre-trained ViT are kept fixed while the designed adaptive modules are optimized to capture forgery features. Specifically, a global adaptive module is designed to model long-range interactions among input tokens, which takes advantage of self-attention mechanism to mine global forgery clues. To further explore essential local forgery clues, a local adaptive module is proposed to expose local inconsistencies by enhancing the local contextual association. In addition, we introduce a fine-grained adaptive learning module that emphasizes the common compact representation of genuine faces through relationship learning in fine-grained pairs, driving these proposed adaptive modules to be aware of fine-grained forgery-aware information. Extensive experiments demonstrate that our FA-ViT achieves state-of-the-arts results in the cross-dataset evaluation, and enhances the robustness against unseen perturbations. Particularly, FA-ViT achieves 93.83\% and 78.32\% AUC scores on Celeb-DF and DFDC datasets in the cross-dataset evaluation. The code and trained model have been released at: https://github.com/LoveSiameseCat/FAViT.

Gaze following aims to interpret human-scene interactions by predicting the person's focal point of gaze. Prevailing approaches often adopt a two-stage framework, whereby multi-modality information is extracted in the initial stage for gaze target prediction. Consequently, the efficacy of these methods highly depends on the precision of the preceding modality extraction. Others use a single-modality approach with complex decoders, increasing network computational load. Inspired by the remarkable success of pre-trained plain vision transformers (ViTs), we introduce a novel single-modality gaze following framework called ViTGaze. In contrast to previous methods, it creates a novel gaze following framework based mainly on powerful encoders (relative decoder parameters less than 1%). Our principal insight is that the inter-token interactions within self-attention can be transferred to interactions between humans and scenes. Leveraging this presumption, we formulate a framework consisting of a 4D interaction encoder and a 2D spatial guidance module to extract human-scene interaction information from self-attention maps. Furthermore, our investigation reveals that ViT with self-supervised pre-training has an enhanced ability to extract correlation information. Many experiments have been conducted to demonstrate the performance of the proposed method. Our method achieves state-of-the-art (SOTA) performance among all single-modality methods (3.4% improvement in the area under curve (AUC) score, 5.1% improvement in the average precision (AP)) and very comparable performance against multi-modality methods with 59% number of parameters less.

This paper considers a novel application of deep AUC maximization (DAM) for multi-instance learning (MIL), in which a single class label is assigned to a bag of instances (e.g., multiple 2D slices of a CT scan for a patient). We address a neglected yet non-negligible computational challenge of MIL in the context of DAM, i.e., bag size is too large to be loaded into {GPU} memory for backpropagation, which is required by the standard pooling methods of MIL. To tackle this challenge, we propose variance-reduced stochastic pooling methods in the spirit of stochastic optimization by formulating the loss function over the pooled prediction as a multi-level compositional function. By synthesizing techniques from stochastic compositional optimization and non-convex min-max optimization, we propose a unified and provable muli-instance DAM (MIDAM) algorithm with stochastic smoothed-max pooling or stochastic attention-based pooling, which only samples a few instances for each bag to compute a stochastic gradient estimator and to update the model parameter. We establish a similar convergence rate of the proposed MIDAM algorithm as the state-of-the-art DAM algorithms. Our extensive experiments on conventional MIL datasets and medical datasets demonstrate the superiority of our MIDAM algorithm.

Objective: We develop a deep learning framework based on the pre-trained Bidirectional Encoder Representations from Transformers (BERT) model using unstructured clinical notes from electronic health records (EHRs) to predict the risk of disease progression from Mild Cognitive Impairment (MCI) to Alzheimer's Disease (AD). Materials and Methods: We identified 3657 patients diagnosed with MCI together with their progress notes from Northwestern Medicine Enterprise Data Warehouse (NMEDW) between 2000-2020. The progress notes no later than the first MCI diagnosis were used for the prediction. We first preprocessed the notes by deidentification, cleaning and splitting, and then pretrained a BERT model for AD (AD-BERT) based on the publicly available Bio+Clinical BERT on the preprocessed notes. The embeddings of all the sections of a patient's notes processed by AD-BERT were combined by MaxPooling to compute the probability of MCI-to-AD progression. For replication, we conducted a similar set of experiments on 2563 MCI patients identified at Weill Cornell Medicine (WCM) during the same timeframe. Results: Compared with the 7 baseline models, the AD-BERT model achieved the best performance on both datasets, with Area Under receiver operating characteristic Curve (AUC) of 0.8170 and F1 score of 0.4178 on NMEDW dataset and AUC of 0.8830 and F1 score of 0.6836 on WCM dataset. Conclusion: We developed a deep learning framework using BERT models which provide an effective solution for prediction of MCI-to-AD progression using clinical note analysis.

We present Mirable's submission to the 2021 Emotions and Themes in Music challenge. In this work, we intend to address the question: can we leverage semi-supervised learning techniques on music emotion recognition? With that, we experiment with noisy student training, which has improved model performance in the image classification domain. As the noisy student method requires a strong teacher model, we further delve into the factors including (i) input training length and (ii) complementary music representations to further boost the performance of the teacher model. For (i), we find that models trained with short input length perform better in PR-AUC, whereas those trained with long input length perform better in ROC-AUC. For (ii), we find that using harmonic pitch class profiles (HPCP) consistently improve tagging performance, which suggests that harmonic representation is useful for music emotion tagging. Finally, we find that noisy student method only improves tagging results for the case of long training length. Additionally, we find that ensembling representations trained with different training lengths can improve tagging results significantly, which suggest a possible direction to explore incorporating multiple temporal resolutions in the network architecture for future work.

Clinical trials are crucial for drug development but are time consuming, expensive, and often burdensome on patients. More importantly, clinical trials face uncertain outcomes due to issues with efficacy, safety, or problems with patient recruitment. If we were better at predicting the results of clinical trials, we could avoid having to run trials that will inevitably fail more resources could be devoted to trials that are likely to succeed. In this paper, we propose Hierarchical INteraction Network (HINT) for more general, clinical trial outcome predictions for all diseases based on a comprehensive and diverse set of web data including molecule information of the drugs, target disease information, trial protocol and biomedical knowledge. HINT first encode these multi-modal data into latent embeddings, where an imputation module is designed to handle missing data. Next, these embeddings will be fed into the knowledge embedding module to generate knowledge embeddings that are pretrained using external knowledge on pharmaco-kinetic properties and trial risk from the web. Then the interaction graph module will connect all the embedding via domain knowledge to fully capture various trial components and their complex relations as well as their influences on trial outcomes. Finally, HINT learns a dynamic attentive graph neural network to predict trial outcome. Comprehensive experimental results show that HINT achieves strong predictive performance, obtaining 0.772, 0.607, 0.623, 0.703 on PR-AUC for Phase I, II, III, and indication outcome prediction, respectively. It also consistently outperforms the best baseline method by up to 12.4\% on PR-AUC.

Many applications of machine learning require a model to make accurate pre-dictions on test examples that are distributionally different from training ones, while task-specific labels are scarce during training. An effective approach to this challenge is to pre-train a model on related tasks where data is abundant, and then fine-tune it on a downstream task of interest. While pre-training has been effective in many language and vision domains, it remains an open question how to effectively use pre-training on graph datasets. In this paper, we develop a new strategy and self-supervised methods for pre-training Graph Neural Networks (GNNs). The key to the success of our strategy is to pre-train an expressive GNN at the level of individual nodes as well as entire graphs so that the GNN can learn useful local and global representations simultaneously. We systematically study pre-training on multiple graph classification datasets. We find that naive strategies, which pre-train GNNs at the level of either entire graphs or individual nodes, give limited improvement and can even lead to negative transfer on many downstream tasks. In contrast, our strategy avoids negative transfer and improves generalization significantly across downstream tasks, leading up to 9.4% absolute improvements in ROC-AUC over non-pre-trained models and achieving state-of-the-art performance for molecular property prediction and protein function prediction.

Data contamination poses a significant threat to the reliable evaluation of Large Language Models (LLMs). This issue arises when benchmark samples may inadvertently appear in training sets, compromising the validity of reported performance. While detection methods have been developed for the pre-training and Supervised Fine-Tuning stages, a critical research gap exists for the increasingly significant phase of Reinforcement Learning (RL) post-training. As RL post-training becomes pivotal for advancing LLM reasoning, the absence of specialized contamination detection methods in this paradigm presents a critical vulnerability. To address this, we conduct the first systematic study of data detection within RL post-training scenario and propose Self-Critique. Our method is motivated by a key observation: after RL phase, the output entropy distribution of LLMs tends to collapse into highly specific and sparse modes. Self-Critique probes for the underlying policy collapse, i.e., the model's convergence to a narrow reasoning path, which causes this entropy reduction. To facilitate this research, we also introduce RL-MIA, a benchmark constructed to simulate this specific contamination scenario. Extensive experiments show that Self-Critique significantly outperforms baseline methods across multiple models and contamination tasks, achieving an AUC improvement of up to 30%. Whereas existing methods are close to a random guess for RL-phase contamination, our method makes detection possible.

Mammography, an X-ray-based imaging technique, remains central to the early detection of breast cancer. Recent advances in artificial intelligence have enabled increasingly sophisticated computer-aided diagnostic methods, evolving from patch-based classifiers to whole-image approaches and then to multi-view architectures that jointly analyze complementary projections. Despite this progress, several critical questions remain unanswered. In this study, we systematically investigate these issues by addressing five key research questions: (1) the role of patch classifiers in performance, (2) the transferability of natural-image-trained backbones, (3) the advantages of learn-to-resize over conventional downscaling, (4) the contribution of multi-view integration, and (5) the robustness of findings across varying image quality. Beyond benchmarking, our experiments demonstrate clear performance gains over prior work. For the CBIS-DDSM dataset, we improved single-view AUC from 0.8153 to 0.8343, and multiple-view AUC from 0.8483 to 0.8658. Using a new comparative method, we also observed a 0.0217 AUC increase when extending from single to multiple-view analysis. On the complete VinDr-Mammo dataset, the multiple-view approach further improved results, achieving a 0.0492 AUC increase over single view and reaching 0.8511 AUC overall. These results establish new state-of-the-art benchmarks, providing clear evidence of the advantages of multi-view architectures for mammogram interpretation. Beyond performance, our analysis offers principled insights into model design and transfer learning strategies, contributing to the development of more accurate and reliable breast cancer screening tools. The inference code and trained models are publicly available at https://github.com/dpetrini/multiple-view.

Speaker identification in multilingual settings presents unique challenges, particularly when conventional models are predominantly trained on English data. In this paper, we propose WSI (Whisper Speaker Identification), a framework that repurposes the encoder of the Whisper automatic speech recognition model pre trained on extensive multilingual data to generate robust speaker embeddings via a joint loss optimization strategy that leverages online hard triplet mining and self supervised Normalized Temperature-scaled Cross Entropy loss. By capitalizing on Whisper language-agnostic acoustic representations, our approach effectively distinguishes speakers across diverse languages and recording conditions. Extensive evaluations on multiple corpora, including VoxTube (multilingual), JVS (Japanese), CallHome (German, Spanish, Chinese, and Japanese), and Voxconverse (English), demonstrate that WSI consistently outperforms state-of-the-art baselines, namely Pyannote Embedding, ECAPA TDNN, and Xvector, in terms of lower equal error rates and higher AUC scores. These results validate our hypothesis that a multilingual pre-trained ASR encoder, combined with joint loss optimization, substantially improves speaker identification performance in non-English languages.

As generative Artificial Intelligence (AI) advances, the realism of AI generated imagery has reached a threshold capable of deceiving even vigilant human observers. Yet, while current AI-generated Image Detection (AID) approaches perform exceptionally well on controlled benchmark datasets, they struggle significantly with real-world cases. To study this behavior we introduce the ITW-SM dataset, a curated collection of real and AI-generated images originating from major social media platforms. We employ it to analyze the effects of key design choices typically considered when building a detector, involving its architecture, pre-trained latent spaces, training data as well as pre-processing approaches. We indicate that naively scaling the pre-training stage or opting for more training data does not always lead to better detection performance. Instead, our work reveals that it is crucial to optimize each design choice to enable the processing pipeline to propagate and effectively analyze both low-level traces as well as high-level image semantics. Building on our findings, we achieve a substantial average improvement of 26.87% in AUC across multiple state-of-the-art detection approaches and under real-world conditions, providing a roadmap for developing more resilient detectors. Our assets are available on https://mever-team.github.io/itw-sm.

Detecting prompt injection and jailbreak attacks is critical for deploying LLM-based agents safely. As agents increasingly process untrusted data from emails, documents, tool outputs, and external APIs, robust attack detection becomes essential. Yet current evaluation practices and production systems have fundamental limitations. We present a comprehensive analysis using a diverse benchmark of 18 datasets spanning harmful requests, jailbreaks, indirect prompt injections, and extraction attacks. We propose Leave-One-Dataset-Out (LODO) evaluation to measure true out-of-distribution generalization, revealing that the standard practice of train-test splits from the same dataset sources severely overestimates performance: aggregate metrics show an 8.4 percentage point AUC inflation, but per-dataset gaps range from 1% to 25% accuracy-exposing heterogeneous failure modes. To understand why classifiers fail to generalize, we analyze Sparse Auto-Encoder (SAE) feature coefficients across LODO folds, finding that 28% of top features are dataset-dependent shortcuts whose class signal depends on specific dataset compositions rather than semantic content. We systematically compare production guardrails (PromptGuard 2, LlamaGuard) and LLM-as-judge approaches on our benchmark, finding all three fail on indirect attacks targeting agents (7-37% detection) and that PromptGuard 2 and LlamaGuard cannot evaluate agentic tool injection due to architectural limitations. Finally, we show that LODO-stable SAE features provide more reliable explanations for classifier decisions by filtering dataset artifacts. We release our evaluation framework at https://github.com/maxf-zn/prompt-mining to establish LODO as the appropriate protocol for prompt attack detection research.

Mycotoxin contamination poses a significant risk to cereal crop quality, food safety, and agricultural productivity. Accurate prediction of mycotoxin levels can support early intervention strategies and reduce economic losses. This study investigates the use of neural networks and transfer learning models to predict mycotoxin contamination in Irish oat crops as a multi-response prediction task. Our dataset comprises oat samples collected in Ireland, containing a mix of environmental, agronomic, and geographical predictors. Five modelling approaches were evaluated: a baseline multilayer perceptron (MLP), an MLP with pre-training, and three transfer learning models; TabPFN, TabNet, and FT-Transformer. Model performance was evaluated using regression (RMSE, R^2) and classification (AUC, F1) metrics, with results reported per toxin and on average. Additionally, permutation-based variable importance analysis was conducted to identify the most influential predictors across both prediction tasks. The transfer learning approach TabPFN provided the overall best performance, followed by the baseline MLP. Our variable importance analysis revealed that weather history patterns in the 90-day pre-harvest period were the most important predictors, alongside seed moisture content.

Hepatocellular carcinoma (HCC), ranking as the third leading cause of cancer-related mortality worldwide, demands urgent improvements in early detection to enhance patient survival. While ultrasound remains the preferred screening modality due to its cost-effectiveness and real-time capabilities, its sensitivity (59%-78%) heavily relies on radiologists' expertise, leading to inconsistent diagnostic outcomes and operational inefficiencies. Recent advancements in AI technology offer promising solutions to bridge this gap. This study introduces the Hierarchical Sparse Query Transformer (HSQformer), a novel hybrid architecture that synergizes CNNs' local feature extraction with Vision Transformers' global contextual awareness through latent space representation and sparse learning. By dynamically activating task-specific experts via a Mixture-of-Experts (MoE) framework, HSQformer achieves hierarchical feature integration without structural redundancy. Evaluated across three clinical scenarios: single-center, multi-center, and high-risk patient cohorts, HSQformer outperforms state-of-the-art models (e.g., 95.38% AUC in multi-center testing) and matches senior radiologists' diagnostic accuracy while significantly surpassing junior counterparts. These results highlight the potential of AI-assisted tools to standardize HCC screening, reduce dependency on human expertise, and improve early diagnosis rates. The full code is available at https://github.com/Asunatan/HSQformer.

Recent statements about the impressive capabilities of large language models (LLMs) are usually supported by evaluating on open-access benchmarks. Considering the vast size and wide-ranging sources of LLMs' training data, it could explicitly or implicitly include test data, leading to LLMs being more susceptible to data contamination. However, due to the opacity of training data, the black-box access of models, and the rapid growth of synthetic training data, detecting and mitigating data contamination for LLMs faces significant challenges. In this paper, we propose CDD, which stands for Contamination Detection via output Distribution for LLMs. CDD necessitates only the sampled texts to detect data contamination, by identifying the peakedness of LLM's output distribution. To mitigate the impact of data contamination in evaluation, we also present TED: Trustworthy Evaluation via output Distribution, based on the correction of LLM's output distribution. To facilitate this study, we introduce two benchmarks, i.e., DetCon and ComiEval, for data contamination detection and contamination mitigation evaluation tasks. Extensive experimental results show that CDD achieves the average relative improvements of 21.8\%-30.2\% over other contamination detection approaches in terms of Accuracy, F1 Score, and AUC metrics, and can effectively detect implicit contamination. TED substantially mitigates performance improvements up to 66.9\% attributed to data contamination across various contamination setups. In real-world applications, we reveal that ChatGPT exhibits a high potential to suffer from data contamination on HumanEval benchmark.

The sparse Mixture-of-Experts (Sparse-MoE) framework efficiently scales up model capacity in various domains, such as natural language processing and vision. Sparse-MoEs select a subset of the "experts" (thus, only a portion of the overall network) for each input sample using a sparse, trainable gate. Existing sparse gates are prone to convergence and performance issues when training with first-order optimization methods. In this paper, we introduce two improvements to current MoE approaches. First, we propose a new sparse gate: COMET, which relies on a novel tree-based mechanism. COMET is differentiable, can exploit sparsity to speed up computation, and outperforms state-of-the-art gates. Second, due to the challenging combinatorial nature of sparse expert selection, first-order methods are typically prone to low-quality solutions. To deal with this challenge, we propose a novel, permutation-based local search method that can complement first-order methods in training any sparse gate, e.g., Hash routing, Top-k, DSelect-k, and COMET. We show that local search can help networks escape bad initializations or solutions. We performed large-scale experiments on various domains, including recommender systems, vision, and natural language processing. On standard vision and recommender systems benchmarks, COMET+ (COMET with local search) achieves up to 13% improvement in ROC AUC over popular gates, e.g., Hash routing and Top-k, and up to 9% over prior differentiable gates e.g., DSelect-k. When Top-k and Hash gates are combined with local search, we see up to 100times reduction in the budget needed for hyperparameter tuning. Moreover, for language modeling, our approach improves over the state-of-the-art MoEBERT model for distilling BERT on 5/7 GLUE benchmarks as well as SQuAD dataset.

Knowledge tracing (KT) is the problem of predicting students' future performance based on their historical interactions with intelligent tutoring systems. Recently, many works present lots of special methods for applying deep neural networks to KT from different perspectives like model architecture, adversarial augmentation and etc., which make the overall algorithm and system become more and more complex. Furthermore, due to the lack of standardized evaluation protocol liu2022pykt, there is no widely agreed KT baselines and published experimental comparisons become inconsistent and self-contradictory, i.e., the reported AUC scores of DKT on ASSISTments2009 range from 0.721 to 0.821 minn2018deep,yeung2018addressing. Therefore, in this paper, we provide a strong but simple baseline method to deal with the KT task named simpleKT. Inspired by the Rasch model in psychometrics, we explicitly model question-specific variations to capture the individual differences among questions covering the same set of knowledge components that are a generalization of terms of concepts or skills needed for learners to accomplish steps in a task or a problem. Furthermore, instead of using sophisticated representations to capture student forgetting behaviors, we use the ordinary dot-product attention function to extract the time-aware information embedded in the student learning interactions. Extensive experiments show that such a simple baseline is able to always rank top 3 in terms of AUC scores and achieve 57 wins, 3 ties and 16 loss against 12 DLKT baseline methods on 7 public datasets of different domains. We believe this work serves as a strong baseline for future KT research. Code is available at https://github.com/pykt-team/pykt-toolkitWe merged our model to the \textsc{pyKT benchmark at https://pykt.org/.}.

Early detection of retinal diseases is one of the most important means of preventing partial or permanent blindness in patients. In this research, a novel multi-label classification system is proposed for the detection of multiple retinal diseases, using fundus images collected from a variety of sources. First, a new multi-label retinal disease dataset, the MuReD dataset, is constructed, using a number of publicly available datasets for fundus disease classification. Next, a sequence of post-processing steps is applied to ensure the quality of the image data and the range of diseases, present in the dataset. For the first time in fundus multi-label disease classification, a transformer-based model optimized through extensive experimentation is used for image analysis and decision making. Numerous experiments are performed to optimize the configuration of the proposed system. It is shown that the approach performs better than state-of-the-art works on the same task by 7.9% and 8.1% in terms of AUC score for disease detection and disease classification, respectively. The obtained results further support the potential applications of transformer-based architectures in the medical imaging field.

According to the considerable growth in the avail of chest X-ray images in diagnosing various diseases, as well as gathering extensive datasets, having an automated diagnosis procedure using deep neural networks has occupied the minds of experts. Most of the available methods in computer vision use a CNN backbone to acquire high accuracy on the classification problems. Nevertheless, recent researches show that transformers, established as the de facto method in NLP, can also outperform many CNN-based models in vision. This paper proposes a multi-label classification deep model based on the Swin Transformer as the backbone to achieve state-of-the-art diagnosis classification. It leverages Multi-Layer Perceptron, also known as MLP, for the head architecture. We evaluate our model on one of the most widely-used and largest x-ray datasets called "Chest X-ray14," which comprises more than 100,000 frontal/back-view images from over 30,000 patients with 14 famous chest diseases. Our model has been tested with several number of MLP layers for the head setting, each achieves a competitive AUC score on all classes. Comprehensive experiments on Chest X-ray14 have shown that a 3-layer head attains state-of-the-art performance with an average AUC score of 0.810, compared to the former SOTA average AUC of 0.799. We propose an experimental setup for the fair benchmarking of existing methods, which could be used as a basis for the future studies. Finally, we followed up our results by confirming that the proposed method attends to the pathologically relevant areas of the chest.

Visual tracking problem demands to efficiently perform robust classification and accurate target state estimation over a given target at the same time. Former methods have proposed various ways of target state estimation, yet few of them took the particularity of the visual tracking problem itself into consideration. After a careful analysis, we propose a set of practical guidelines of target state estimation for high-performance generic object tracker design. Following these guidelines, we design our Fully Convolutional Siamese tracker++ (SiamFC++) by introducing both classification and target state estimation branch(G1), classification score without ambiguity(G2), tracking without prior knowledge(G3), and estimation quality score(G4). Extensive analysis and ablation studies demonstrate the effectiveness of our proposed guidelines. Without bells and whistles, our SiamFC++ tracker achieves state-of-the-art performance on five challenging benchmarks(OTB2015, VOT2018, LaSOT, GOT-10k, TrackingNet), which proves both the tracking and generalization ability of the tracker. Particularly, on the large-scale TrackingNet dataset, SiamFC++ achieves a previously unseen AUC score of 75.4 while running at over 90 FPS, which is far above the real-time requirement. Code and models are available at: https://github.com/MegviiDetection/video_analyst .

Knowledge tracing is the task of modeling each student's mastery of knowledge concepts (KCs) as (s)he engages with a sequence of learning activities. Each student's knowledge is modeled by estimating the performance of the student on the learning activities. It is an important research area for providing a personalized learning platform to students. In recent years, methods based on Recurrent Neural Networks (RNN) such as Deep Knowledge Tracing (DKT) and Dynamic Key-Value Memory Network (DKVMN) outperformed all the traditional methods because of their ability to capture complex representation of human learning. However, these methods face the issue of not generalizing well while dealing with sparse data which is the case with real-world data as students interact with few KCs. In order to address this issue, we develop an approach that identifies the KCs from the student's past activities that are relevant to the given KC and predicts his/her mastery based on the relatively few KCs that it picked. Since predictions are made based on relatively few past activities, it handles the data sparsity problem better than the methods based on RNN. For identifying the relevance between the KCs, we propose a self-attention based approach, Self Attentive Knowledge Tracing (SAKT). Extensive experimentation on a variety of real-world dataset shows that our model outperforms the state-of-the-art models for knowledge tracing, improving AUC by 4.43% on average.

The Segment Anything Model 2 (SAM 2) has demonstrated strong performance in object segmentation tasks but faces challenges in visual object tracking, particularly when managing crowded scenes with fast-moving or self-occluding objects. Furthermore, the fixed-window memory approach in the original model does not consider the quality of memories selected to condition the image features for the next frame, leading to error propagation in videos. This paper introduces SAMURAI, an enhanced adaptation of SAM 2 specifically designed for visual object tracking. By incorporating temporal motion cues with the proposed motion-aware memory selection mechanism, SAMURAI effectively predicts object motion and refines mask selection, achieving robust, accurate tracking without the need for retraining or fine-tuning. SAMURAI operates in real-time and demonstrates strong zero-shot performance across diverse benchmark datasets, showcasing its ability to generalize without fine-tuning. In evaluations, SAMURAI achieves significant improvements in success rate and precision over existing trackers, with a 7.1% AUC gain on LaSOT_{ext} and a 3.5% AO gain on GOT-10k. Moreover, it achieves competitive results compared to fully supervised methods on LaSOT, underscoring its robustness in complex tracking scenarios and its potential for real-world applications in dynamic environments. Code and results are available at https://github.com/yangchris11/samurai.

We release Terminal Wrench, a subset of 331 terminal-agent benchmark environments, copied from the popular open benchmarks that are demonstrably reward-hackable. The data set includes 3,632 hack trajectories and 2,352 legitimate baseline trajectories across three frontier models (Claude Opus 4.6, Gemini 3.1 Pro, GPT-5.4). Each entry preserves the original task definition alongside full attack trajectories that show how the verifier was bypassed. It also includes cases where the task was not solved as intended. The tasks span system administration, machine learning, software engineering, and security challenges; the exploits range from simple output spoofing to stack-frame introspection, standard-library patching, and rootkit-style binary hijacking. Crucially, these exploits are specific to each task, rather than the evaluation harness, making them harder to patch. We also present a monitorability study in which hack trajectories are sanitized or stripped of reasoning traces and then scored by an LLM judge, showing that detection degrades meaningfully when chain-of-thought is removed (AUC drops from 0.97 to 0.92). The data set is publicly available at https://github.com/few-sh/terminal-wrench.

The limited availability of labeled chest X-ray datasets is a significant bottleneck in the development of medical imaging methods. Self-supervised learning (SSL) can mitigate this problem by training models on unlabeled data. Furthermore, self-supervised pretraining has yielded promising results in visual recognition of natural images but has not been given much consideration in medical image analysis. In this work, we propose a self-supervised method, DINO-CXR, which is a novel adaptation of a self-supervised method, DINO, based on a vision transformer for chest X-ray classification. A comparative analysis is performed to show the effectiveness of the proposed method for both pneumonia and COVID-19 detection. Through a quantitative analysis, it is also shown that the proposed method outperforms state-of-the-art methods in terms of accuracy and achieves comparable results in terms of AUC and F-1 score while requiring significantly less labeled data.

Self-supervised pre-training of deep learning models with contrastive learning is a widely used technique in image analysis. Current findings indicate a strong potential for contrastive pre-training on medical images. However, further research is necessary to incorporate the particular characteristics of these images. We hypothesize that the similarity of medical images hinders the success of contrastive learning in the medical imaging domain. To this end, we investigate different strategies based on deep embedding, information theory, and hashing in order to identify and reduce redundancy in medical pre-training datasets. The effect of these different reduction strategies on contrastive learning is evaluated on two pre-training datasets and several downstream classification tasks. In all of our experiments, dataset reduction leads to a considerable performance gain in downstream tasks, e.g., an AUC score improvement from 0.78 to 0.83 for the COVID CT Classification Grand Challenge, 0.97 to 0.98 for the OrganSMNIST Classification Challenge and 0.73 to 0.83 for a brain hemorrhage classification task. Furthermore, pre-training is up to nine times faster due to the dataset reduction. In conclusion, the proposed approach highlights the importance of dataset quality and provides a transferable approach to improve contrastive pre-training for classification downstream tasks on medical images.

Autism spectrum disorder (ASD) is a prevalent psychiatric condition characterized by atypical cognitive, emotional, and social patterns. Timely and accurate diagnosis is crucial for effective interventions and improved outcomes in individuals with ASD. In this study, we propose a novel Multi-Atlas Enhanced Transformer framework, METAFormer, ASD classification. Our framework utilizes resting-state functional magnetic resonance imaging data from the ABIDE I dataset, comprising 406 ASD and 476 typical control (TC) subjects. METAFormer employs a multi-atlas approach, where flattened connectivity matrices from the AAL, CC200, and DOS160 atlases serve as input to the transformer encoder. Notably, we demonstrate that self-supervised pretraining, involving the reconstruction of masked values from the input, significantly enhances classification performance without the need for additional or separate training data. Through stratified cross-validation, we evaluate the proposed framework and show that it surpasses state-of-the-art performance on the ABIDE I dataset, with an average accuracy of 83.7% and an AUC-score of 0.832. The code for our framework is available at https://github.com/Lugges991/METAFormer

Knee Osteoarthritis (KOA) is a highly prevalent chronic musculoskeletal condition with no currently available treatment. The manifestation of KOA is heterogeneous and prediction of its progression is challenging. Current literature suggests that the use of multi-modal data and advanced modeling methods, such as the ones based on Deep Learning, has promise in tackling this challenge. To date, however, the evidence on the efficacy of this approach is limited. In this study, we leveraged recent advances in Deep Learning and, using a Transformer approach, developed a unified framework for the multi-modal fusion of knee imaging data. Subsequently, we analyzed its performance across a range of scenarios by investigating multiple progression horizons -- from short-term to long-term. We report our findings using a large cohort (n=2421-3967) derived from the Osteoarthritis Initiative dataset. We show that structural knee MRI allows identifying radiographic KOA progressors on par with multi-modal fusion approaches, achieving an area under the ROC curve (ROC AUC) of 0.70-0.76 and Average Precision (AP) of 0.15-0.54 in 2-8 year horizons. Progression within 1 year was better predicted with a multi-modal method using X-ray, structural, and compositional MR images -- ROC AUC of 0.76(0.04), AP of 0.13(0.04) -- or via clinical data. Our follow-up analysis generally shows that prediction from the imaging data is more accurate for post-traumatic subjects, and we further investigate which subject subgroups may benefit the most. The present study provides novel insights into multi-modal imaging of KOA and brings a unified data-driven framework for studying its progression in an end-to-end manner, providing new tools for the design of more efficient clinical trials. The source code of our framework and the pre-trained models are made publicly available.

We discover that large language models exhibit spectral phase transitions in their hidden activation spaces when engaging in reasoning versus factual recall. Through systematic spectral analysis across 11 models spanning 5 architecture families (Qwen, Pythia, Phi, Llama, DeepSeek-R1), we identify seven core phenomena: (1)~Reasoning Spectral Compression -- 9/11 models show significantly lower α for reasoning (p < 0.05), with larger effects in stronger models; (2)~Instruction Tuning Spectral Reversal -- base models show reasoning α< factual α, while instruction-tuned models reverse this relationship; (3)~Architecture-Dependent Generation Taxonomy -- prompt-to-response shifts partition into expansion, compression, and equilibrium regimes; (4)~Spectral Scaling Law -- α_reasoning propto -0.074 ln N across 4 Qwen base models (R^2 = 0.46); (5)~Token-Level Spectral Cascade -- per-token alpha tracking reveals local synchronization that decays exponentially with layer distance, and is weaker for reasoning than factual tasks; (6)~Reasoning Step Spectral Punctuation -- phase-transition signatures align with reasoning step boundaries; and (7)~Spectral Correctness Prediction -- spectral α alone achieves AUC = 1.000 (Qwen2.5-7B, late layers) and mean AUC = 0.893 across 6 models in predicting correctness before the final answer is generated. Together, these findings establish a comprehensive spectral theory of reasoning in transformers, revealing that the geometry of thought is universal in direction, architecture-specific in dynamics, and predictive of outcome.

While hyperspectral imaging provides rich spatial-spectral information across hundreds of narrow wavelength bands for precise material identification, ground-based hyperspectral pre-trained backbones remain absent, constrained by varying spectral configurations across sensors, the scarcity and inconsistency of labels, and the limited scale and scene diversity of existing datasets. To address these challenges and enable universal perception, we propose HyperVision, the first ground-based hyperspectral pre-trained backbone. First, to handle varying spectral configurations, HyperVision adopts a channel-adaptive dynamic embedding mechanism to map heterogeneous inputs into a unified token space. Second, we develop an unsupervised representation learning framework. Specifically, to address label scarcity and inconsistency, a multi-source pseudo-labeling method is introduced to fuse spatial structures from SAM2 and fine-grained spectral material information from HyperFree. Furthermore, to enrich scene diversity and compensate for limited dataset scale, a cross-modal knowledge distillation mechanism is utilized to transfer rich semantic representations from a pre-trained RGB vision model to our backbone. Pre-trained on a collection of 15k images from 26 diverse ground-based datasets, HyperVision demonstrates exceptional generalization. Requiring only efficient head-only adaptation without adjusting backbone parameters, it achieves state-of-the-art performance compared to task-specific methods across three downstream tasks under varying sensor configurations, yielding up to a 16.3% relative improvement in hyperspectral semantic segmentation Acc_{M}, a 2.1% relative gain in object tracking AUC, and a 35.5% reduction in salient object detection MAE. The source code and pre-trained model will be publicly available on https://github.com/lronkitty/HyperVision .

The reasoning capabilities of large language models (LLMs) have improved substantially through increased test-time computation, typically in the form of intermediate tokens known as chain-of-thought (CoT). However, CoT often becomes unnecessarily long, increasing computation cost without actual accuracy gains or sometimes even degrading performance, a phenomenon known as ``overthinking''. We propose a multi-stage efficient reasoning method that combines supervised fine-tuning -- via rejection sampling or reasoning trace reformatting -- with reinforcement learning using an adaptive length penalty. We introduce a lightweight reward function that penalizes tokens generated after the first correct answer but encouraging self-verification only when beneficial. We conduct a holistic evaluation across seven diverse reasoning tasks, analyzing the accuracy-response length trade-off. Our approach reduces response length by an average of 28\% for 8B models and 40\% for 32B models, while incurring only minor performance drops of 1.6 and 2.5 points, respectively. Despite its conceptual simplicity, it achieves a superior trade-off compared to more complex state-of-the-art efficient reasoning methods, scoring 76.6, in terms of the area under the Overthinking-Adjusted Accuracy curve (AUC_{OAA}) -- 5 points above the base model and 2.5 points above the second-best approach.

Teaching videos are becoming a major medium for education, creating a growing need for scalable evaluation of their pedagogical quality. Existing automatic judges do not fully address this setting because teaching quality depends on multimodal evidence and should be evaluated with respect to the intended learner rather than as a universal property. We present EduPanel, a rubric-grounded, learner-conditioned LLM judge that decomposes evaluation across specialized agents to produce interpretable assessments for different aspects of teaching quality. Across expert studies, architecture ablations, and learner-persona analyses, EduPanel achieves reliability comparable to a median human expert. In expert evaluation, its feedback improves scoring accuracy (MAE 0.87 to 0.73), while experts remain able to detect unreliable outputs (AUC = 0.77) instead of accepting them blindly. These results suggest that EduPanel can serve as effective assistants for educational evaluation rather than replacements for human experts.

With the rapid growth in deepfake video content, we require improved and generalizable methods to detect them. Most existing detection methods either use uni-modal cues or rely on supervised training to capture the dissonance between the audio and visual modalities. While the former disregards the audio-visual correspondences entirely, the latter predominantly focuses on discerning audio-visual cues within the training corpus, thereby potentially overlooking correspondences that can help detect unseen deepfakes. We present Audio-Visual Feature Fusion (AVFF), a two-stage cross-modal learning method that explicitly captures the correspondence between the audio and visual modalities for improved deepfake detection. The first stage pursues representation learning via self-supervision on real videos to capture the intrinsic audio-visual correspondences. To extract rich cross-modal representations, we use contrastive learning and autoencoding objectives, and introduce a novel audio-visual complementary masking and feature fusion strategy. The learned representations are tuned in the second stage, where deepfake classification is pursued via supervised learning on both real and fake videos. Extensive experiments and analysis suggest that our novel representation learning paradigm is highly discriminative in nature. We report 98.6% accuracy and 99.1% AUC on the FakeAVCeleb dataset, outperforming the current audio-visual state-of-the-art by 14.9% and 9.9%, respectively.

Computer-aided histopathological image analysis for cancer detection is a major research challenge in the medical domain. Automatic detection and classification of nuclei for cancer diagnosis impose a lot of challenges in developing state of the art algorithms due to the heterogeneity of cell nuclei and data set variability. Recently, a multitude of classification algorithms has used complex deep learning models for their dataset. However, most of these methods are rigid and their architectural arrangement suffers from inflexibility and non-interpretability. In this research article, we have proposed a hybrid and flexible deep learning architecture OLConvNet that integrates the interpretability of traditional object-level features and generalization of deep learning features by using a shallower Convolutional Neural Network (CNN) named as CNN_{3L}. CNN_{3L} reduces the training time by training fewer parameters and hence eliminating space constraints imposed by deeper algorithms. We used F1-score and multiclass Area Under the Curve (AUC) performance parameters to compare the results. To further strengthen the viability of our architectural approach, we tested our proposed methodology with state of the art deep learning architectures AlexNet, VGG16, VGG19, ResNet50, InceptionV3, and DenseNet121 as backbone networks. After a comprehensive analysis of classification results from all four architectures, we observed that our proposed model works well and perform better than contemporary complex algorithms.

This paper reports on the ReINTEL Shared Task for Responsible Information Identification on social network sites, which is hosted at the seventh annual workshop on Vietnamese Language and Speech Processing (VLSP 2020). Given a piece of news with respective textual, visual content and metadata, participants are required to classify whether the news is `reliable' or `unreliable'. In order to generate a fair benchmark, we introduce a novel human-annotated dataset of over 10,000 news collected from a social network in Vietnam. All models will be evaluated in terms of AUC-ROC score, a typical evaluation metric for classification. The competition was run on the Codalab platform. Within two months, the challenge has attracted over 60 participants and recorded nearly 1,000 submission entries.

As autonomous AI agents transition from code completion tools to full-fledged teammates capable of opening pull requests (PRs) at scale, software maintainers face a new challenge: not just reviewing code, but managing complex interaction loops with non-human contributors. This paradigm shift raises a critical question: can we predict which agent-generated PRs will consume excessive review effort before any human interaction begins? Analyzing 33,707 agent-authored PRs from the AIDev dataset across 2,807 repositories, we uncover a striking two-regime behavioral pattern that fundamentally distinguishes autonomous agents from human developers. The first regime, representing 28.3 percent of all PRs, consists of instant merges (less than 1 minute), reflecting success on narrow automation tasks. The second regime involves iterative review cycles where agents frequently stall or abandon refinement (ghosting). We propose a Circuit Breaker triage model that predicts high-review-effort PRs (top 20 percent) at creation time using only static structural features. A LightGBM model achieves AUC 0.957 on a temporal split, while semantic text features (TF-IDF, CodeBERT) provide negligible predictive value. At a 20 percent review budget, the model intercepts 69 percent of total review effort, enabling zero-latency governance. Our findings challenge prevailing assumptions in AI-assisted code review: review burden is dictated by what agents touch, not what they say, highlighting the need for structural governance mechanisms in human-AI collaboration.

Zero-shot Text-to-Speech (TTS) voice cloning poses severe privacy risks, demanding the removal of specific speaker identities from trained TTS models. Conventional machine unlearning is insufficient in this context, as zero-shot TTS can dynamically reconstruct voices from just reference prompts. We formalize this task as Speech Generation Speaker Poisoning (SGSP), in which we modify trained models to prevent the generation of specific identities while preserving utility for other speakers. We evaluate inference-time filtering and parameter-modification baselines across 1, 15, and 100 forgotten speakers. Performance is assessed through the trade-off between utility (WER) and privacy, quantified using AUC and Forget Speaker Similarity (FSSIM). We achieve strong privacy for up to 15 speakers but reveal scalability limits at 100 speakers due to increased identity overlap. Our study thus introduces a novel problem and evaluation framework toward further advances in generative voice privacy.

Safety-critical perception systems require both reliable uncertainty quantification and principled abstention mechanisms to maintain safety under diverse operational conditions. We present a novel dual-threshold conformalization framework that provides statistically-guaranteed uncertainty estimates while enabling selective prediction in high-risk scenarios. Our approach uniquely combines a conformal threshold ensuring valid prediction sets with an abstention threshold optimized through ROC analysis, providing distribution-free coverage guarantees (\ge 1 - \alpha) while identifying unreliable predictions. Through comprehensive evaluation on CIFAR-100, ImageNet1K, and ModelNet40 datasets, we demonstrate superior robustness across camera and LiDAR modalities under varying environmental perturbations. The framework achieves exceptional detection performance (AUC: 0.993\to0.995) under severe conditions while maintaining high coverage (>90.0\%) and enabling adaptive abstention (13.5\%\to63.4\%\pm0.5) as environmental severity increases. For LiDAR-based perception, our approach demonstrates particularly strong performance, maintaining robust coverage (>84.5\%) while appropriately abstaining from unreliable predictions. Notably, the framework shows remarkable stability under heavy perturbations, with detection performance (AUC: 0.995\pm0.001) significantly outperforming existing methods across all modalities. Our unified approach bridges the gap between theoretical guarantees and practical deployment needs, offering a robust solution for safety-critical autonomous systems operating in challenging real-world conditions.

We propose a weakly-supervised model for word-level mispronunciation detection in non-native (L2) English speech. To train this model, phonetically transcribed L2 speech is not required and we only need to mark mispronounced words. The lack of phonetic transcriptions for L2 speech means that the model has to learn only from a weak signal of word-level mispronunciations. Because of that and due to the limited amount of mispronounced L2 speech, the model is more likely to overfit. To limit this risk, we train it in a multi-task setup. In the first task, we estimate the probabilities of word-level mispronunciation. For the second task, we use a phoneme recognizer trained on phonetically transcribed L1 speech that is easily accessible and can be automatically annotated. Compared to state-of-the-art approaches, we improve the accuracy of detecting word-level pronunciation errors in AUC metric by 30% on the GUT Isle Corpus of L2 Polish speakers, and by 21.5% on the Isle Corpus of L2 German and Italian speakers.

Computational pathology needs whole-slide image (WSI) foundation models that transfer across diverse clinical tasks, yet current approaches remain largely slide-centric, often depend on private data and expensive paired-report supervision, and do not explicitly model relationships among multiple slides from the same patient. We present MOOZY, a patient-first pathology foundation model in which the patient case, not the individual slide, is the core unit of representation. MOOZY explicitly models dependencies across all slides from the same patient via a case transformer during pretraining, combining multi-stage open self-supervision with scaled low-cost task supervision. In Stage 1, we pretrain a vision-only slide encoder on 77,134 public slide feature grids using masked self-distillation. In Stage 2, we align these representations with clinical semantics using a case transformer and multi-task supervision over 333 tasks from 56 public datasets, including 205 classification and 128 survival tasks across four endpoints. Across eight held-out tasks with five-fold frozen-feature probe evaluation, MOOZY achieves best or tied-best performance on most metrics and improves macro averages over TITAN by +7.37%, +5.50%, and +7.83% and over PRISM by +8.83%, +10.70%, and +9.78% for weighted F1, weighted ROC-AUC, and balanced accuracy, respectively. MOOZY is also parameter efficient with 85.77M parameters, 14x smaller than GigaPath. These results demonstrate that open, reproducible patient-level pretraining yields transferable embeddings, providing a practical path toward scalable patient-first histopathology foundation models.

Representation learning of pathology whole-slide images (WSIs) has primarily relied on weak supervision with Multiple Instance Learning (MIL). This approach leads to slide representations highly tailored to a specific clinical task. Self-supervised learning (SSL) has been successfully applied to train histopathology foundation models (FMs) for patch embedding generation. However, generating patient or slide level embeddings remains challenging. Existing approaches for slide representation learning extend the principles of SSL from patch level learning to entire slides by aligning different augmentations of the slide or by utilizing multimodal data. By integrating tile embeddings from multiple FMs, we propose a new single modality SSL method in feature space that generates useful slide representations. Our contrastive pretraining strategy, called COBRA, employs multiple FMs and an architecture based on Mamba-2. COBRA exceeds performance of state-of-the-art slide encoders on four different public CPTAC cohorts on average by at least +3.8% AUC, despite only being pretrained on 3048 WSIs from TCGA. Additionally, COBRA is readily compatible at inference time with previously unseen feature extractors.

Large language models exhibit safety degradation in non-English languages. Standard evaluation relies on Jailbreak Success Rate (JSR), which confounds several safety-driving factors into one, obscuring the specific cause(s) of safety failure. We introduce a latent variable model, a Multi-Group Item Response Theory (IRT) framework, that decouples safety-driving factors such as language-agnostic safety robustness (θ), intrinsic prompt hardness (β), global language processing difficulty (γ), and a prompt-specific cross-lingual safety gap (τ). Using the MultiJail dataset, we evaluate the safety robustness of 61 model configurations across 5 closed-model families and 10 languages of varying resource, aggregating a dataset of 1.9 million rows. Exploratory Factor Analysis shows safety is primarily unidimensional: models refuse different harm types mainly through a shared mechanism. Contrary to the expected trend that safety degrades largely in low-resource languages, 22 model configurations are more vulnerable in English than in low-resource languages. Low-resource languages produce more uncertain responses (high entropy) than high-resource languages. Also, high-τ prompts cluster in physical harm categories like Theft and Weapons and lower-resource languages, trends validated through cross-dataset generalization. While global translation quality shows low correlation with τ, severe mistranslations drive high-bias outliers, as validated by native speakers. Cultural and conceptual grounding mismatches also contribute to τ. In predictive validation, the IRT framework achieves AUC = 0.940, outperforming simpler baselines in predicting safe refusal of unsafe prompts. Our framework reveals concept-language vulnerabilities that aggregate metrics obscure, enabling fairer cross-lingual safety evaluation and targeted improvements in dataset construction.

The ever-growing textual records of contemporary social issues, often discussed online with moral rhetoric, present both an opportunity and a challenge for studying how moral concerns are debated in real life. Moral foundations theory is a taxonomy of intuitions widely used in data-driven analyses of online content, but current computational tools to detect moral foundations suffer from the incompleteness and fragility of their lexicons and from poor generalization across data domains. In this paper, we fine-tune a large language model to measure moral foundations in text based on datasets covering news media and long- and short-form online discussions. The resulting model, called Mformer, outperforms existing approaches on the same domains by 4--12% in AUC and further generalizes well to four commonly used moral text datasets, improving by up to 17% in AUC. We present case studies using Mformer to analyze everyday moral dilemmas on Reddit and controversies on Twitter, showing that moral foundations can meaningfully describe people's stance on social issues and such variations are topic-dependent. Pre-trained model and datasets are released publicly. We posit that Mformer will help the research community quantify moral dimensions for a range of tasks and data domains, and eventually contribute to the understanding of moral situations faced by humans and machines.

Face anti-spoofing (FAS) is indispensable for a face recognition system. Many texture-driven countermeasures were developed against presentation attacks (PAs), but the performance against unseen domains or unseen spoofing types is still unsatisfactory. Instead of exhaustively collecting all the spoofing variations and making binary decisions of live/spoof, we offer a new perspective on the FAS task to distinguish between normal and abnormal movements of live and spoof presentations. We propose Geometry-Aware Interaction Network (GAIN), which exploits dense facial landmarks with spatio-temporal graph convolutional network (ST-GCN) to establish a more interpretable and modularized FAS model. Additionally, with our cross-attention feature interaction mechanism, GAIN can be easily integrated with other existing methods to significantly boost performance. Our approach achieves state-of-the-art performance in the standard intra- and cross-dataset evaluations. Moreover, our model outperforms state-of-the-art methods by a large margin in the cross-dataset cross-type protocol on CASIA-SURF 3DMask (+10.26% higher AUC score), exhibiting strong robustness against domain shifts and unseen spoofing types.

As Large Language Models (LLMs) are increasingly deployed as autonomous agents in complex and long horizon settings, it is critical to evaluate their ability to sabotage users by pursuing hidden objectives. We study the ability of frontier LLMs to evade monitoring and achieve harmful hidden goals while completing a wide array of realistic tasks. We evaluate a broad range of frontier LLMs using SHADE (Subtle Harmful Agent Detection & Evaluation)-Arena, the first highly diverse agent evaluation dataset for sabotage and monitoring capabilities of LLM agents. SHADE-Arena consists of complex pairs of benign main tasks and harmful side objectives in complicated environments. Agents are evaluated on their ability to complete the side task without appearing suspicious to an LLM monitor. When measuring agent ability to (a) complete the main task, (b) complete the side task, and (c) avoid detection, we find that the best performing frontier models score 27% (Claude 3.7 Sonnet) and 15% (Gemini 2.5 Pro) as sabotage agents when overseen by Claude 3.6 Sonnet. For current frontier models, success on the side task relies heavily on having access to a hidden scratchpad that is not visible to the monitor. We also use SHADE-Arena to measure models' monitoring abilities, with the top monitor (Gemini 2.5 Pro) achieving an AUC of 0.87 at distinguishing benign and malign transcripts. We find that for now, models still struggle at sabotage due to failures in long-context main task execution. However, our measurements already demonstrate the difficulty of monitoring for subtle sabotage attempts, which we expect to only increase in the face of more complex and longer-horizon tasks.

In this study, we introduce a novel method to predict mental health by building machine learning models for a non-invasive wearable device equipped with Laser Doppler Flowmetry (LDF) and Fluorescence Spectroscopy (FS) sensors. Besides, we present the corresponding dataset to predict mental health, e.g. depression, anxiety, and stress levels via the DAS-21 questionnaire. To our best knowledge, this is the world's largest and the most generalized dataset ever collected for both LDF and FS studies. The device captures cutaneous blood microcirculation parameters, and wavelet analysis of the LDF signal extracts key rhythmic oscillations. The dataset, collected from 132 volunteers aged 18-94 from 19 countries, explores relationships between physiological features, demographics, lifestyle habits, and health conditions. We employed a variety of machine learning methods to classify stress detection, in which LightGBM is identified as the most effective model for stress detection, achieving a ROC AUC of 0.7168 and a PR AUC of 0.8852. In addition, we also incorporated Explainable Artificial Intelligence (XAI) techniques into our analysis to investigate deeper insights into the model's predictions. Our results suggest that females, younger individuals and those with a higher Body Mass Index (BMI) or heart rate have a greater likelihood of experiencing mental health conditions like stress and anxiety. All related code and data are published online: https://github.com/leduckhai/Wearable_LDF-FS.

This investigation reports on the results of convolutional neural networks developed for the recently introduced PathologicAL Myopia (PALM) dataset, which consists of 1200 fundus images. We propose a new Optic Nerve Head (ONH)-based prediction enhancement for the segmentation of atrophy and fovea. Models trained with 400 available training images achieved an AUC of 0.9867 for pathological myopia classification, and a Euclidean distance of 58.27 pixels on the fovea localization task, evaluated on a test set of 400 images. Dice and F1 metrics for semantic segmentation of lesions scored 0.9303 and 0.9869 on optic disc, 0.8001 and 0.9135 on retinal atrophy, and 0.8073 and 0.7059 on retinal detachment, respectively. Our work was acknowledged with an award in the context of the "PathologicAL Myopia detection from retinal images" challenge held during the IEEE International Symposium on Biomedical Imaging (April 2019). Considering that (pathological) myopia cases are often identified as false positives and negatives in classification systems for glaucoma, we envision that the current work could aid in future research to discriminate between glaucomatous and highly-myopic eyes, complemented by the localization and segmentation of landmarks such as fovea, optic disc and atrophy.

Application developers advertise their Apps by creating product pages with App images, and bidding on search terms. It is then crucial for App images to be highly relevant with the search terms. Solutions to this problem require an image-text matching model to predict the quality of the match between the chosen image and the search terms. In this work, we present a novel approach to matching an App image to search terms based on fine-tuning a pre-trained LXMERT model. We show that compared to the CLIP model and a baseline using a Transformer model for search terms, and a ResNet model for images, we significantly improve the matching accuracy. We evaluate our approach using two sets of labels: advertiser associated (image, search term) pairs for a given application, and human ratings for the relevance between (image, search term) pairs. Our approach achieves 0.96 AUC score for advertiser associated ground truth, outperforming the transformer+ResNet baseline and the fine-tuned CLIP model by 8% and 14%. For human labeled ground truth, our approach achieves 0.95 AUC score, outperforming the transformer+ResNet baseline and the fine-tuned CLIP model by 16% and 17%.

Gathering histopathology slides from over 100 publicly available cohorts, we compile a diverse dataset of 460 million pathology tiles covering more than 30 cancer sites. Using this dataset, we train a large self-supervised vision transformer using DINOv2 and publicly release one iteration of this model for further experimentation, coined Phikon-v2. While trained on publicly available histology slides, Phikon-v2 surpasses our previously released model (Phikon) and performs on par with other histopathology foundation models (FM) trained on proprietary data. Our benchmarks include eight slide-level tasks with results reported on external validation cohorts avoiding any data contamination between pre-training and evaluation datasets. Our downstream training procedure follows a simple yet robust ensembling strategy yielding a +1.75 AUC increase across tasks and models compared to one-shot retraining (p<0.001). We compare Phikon (ViT-B) and Phikon-v2 (ViT-L) against 14 different histology feature extractors, making our evaluation the most comprehensive to date. Our result support evidences that DINOv2 handles joint model and data scaling better than iBOT. Also, we show that recent scaling efforts are overall beneficial to downstream performance in the context of biomarker prediction with GigaPath and H-Optimus-0 (two ViT-g with 1.1B parameters each) standing out. However, the statistical margins between the latest top-performing FMs remain mostly non-significant; some even underperform on specific indications or tasks such as MSI prediction - deposed by a 13x smaller model developed internally. While latest foundation models may exhibit limitations for clinical deployment, they nonetheless offer excellent grounds for the development of more specialized and cost-efficient histology encoders fueling AI-guided diagnostic tools.

We introduce JuRe (Just Repair), a minimal denoising network for time series anomaly detection that exposes a central finding: architectural complexity is unnecessary when the training objective correctly implements the manifold-projection principle. JuRe consists of a single depthwise-separable convolutional residual block with hidden dimension 128, trained to repair corrupted time series windows and scored at inference by a fixed, parameter-free structural discrepancy function. Despite using no attention, no latent variable, and no adversarial component, JuRe ranks second on the TSB-AD multivariate benchmark (AUC-PR 0.404, 180 series, 17 datasets) and second on the UCR univariate archive by AUC-PR (0.198, 250 series), leading all neural baselines on AUC-PR and VUS-PR. Component ablation on TSB-AD identifies training-time corruption as the dominant factor (ΔAUC-PR = 0.047 on removal), confirming that the denoising objective, not network capacity, drives detection quality. Pairwise Wilcoxon signed-rank tests establish statistical significance against 21 of 25 baselines on TSB-AD. Code is available at the URL https://github.com/iis-esslingen/JuRe.

As chain of thought (CoT) has become central to scaling reasoning capabilities in large language models (LLMs), it has also emerged as a promising tool for interpretability, suggesting the opportunity to understand model decisions through verbalized reasoning. However, the utility of CoT toward interpretability depends upon its faithfulness---whether the model's stated reasoning reflects the underlying decision process. We provide mechanistic evidence that instruction-tuned models often determine their answer before generating CoT. Training linear probes on residual stream activations at the last token before CoT, we can predict the model's final answer with >0.9 AUC on most tasks. We find that these directions are not only predictive, but also causal: steering activations along the probe direction often flips model answers, with flip rates substantially exceeding norm-matched orthogonal baselines across most model-dataset pairs. When steering induces incorrect answers, we observe two distinct failure modes: confabulation (fabricating false premises) and non-entailment (stating correct premises but drawing unsupported conclusions). While post-hoc reasoning may be instrumentally useful when the model has a correct pre-CoT belief, these failure modes suggest it can result in undesirable behaviors when reasoning from a false belief.

When a language model asserts that "the capital of Australia is Sydney," does it know this is wrong? We characterize the geometry of correctness representations across 9 models from 5 architecture families. The structure is simple: the discriminative signal occupies 3-8 dimensions, performance degrades with additional dimensions, and no nonlinear classifier improves over linear separation. Centroid distance in the low-dimensional subspace matches trained probe performance (0.90 AUC), enabling few-shot detection: on GPT-2, 25 labeled examples achieve 89% of full-data accuracy. We validate causally through activation steering: the learned direction produces 10.9 percentage point changes in error rates while random directions show no effect. Internal probes achieve 0.80-0.97 AUC; output-based methods (P(True), semantic entropy) achieve only 0.44-0.64 AUC. The correctness signal exists internally but is not expressed in outputs. That centroid distance matches probe performance indicates class separation is a mean shift, making detection geometric rather than learned.

Reliable anomaly detection in photovoltaic (PV) modules is critical for maintaining solar energy efficiency. However, developing robust computer vision models for PV inspection is constrained by the scarcity of large-scale, diverse, and balanced datasets. This study introduces PV-DDPM, a spatially conditioned denoising diffusion probabilistic model that generates anomalous electroluminescence (EL) images across four PV cell types: multi-crystalline silicon (multi-c-Si), mono-crystalline silicon (mono-c-Si), half-cut multi-c-Si, and interdigitated back contact (IBC) with dogbone interconnect. PV-DDPM enables controlled synthesis of single-defect and multi-defect scenarios by conditioning on binary masks representing structural features and defect positions. To the best of our knowledge, this is the first framework that jointly models multiple PV cell types while supporting simultaneous generation of diverse anomaly types. We also introduce E-SCDD, an enhanced version of the SCDD dataset, comprising 1,000 pixel-wise annotated EL images spanning 30 semantic classes, and 1,768 unlabeled synthetic samples. Quantitative evaluation shows our generated images achieve a Fréchet Inception Distance (FID) of 4.10 and Kernel Inception Distance (KID) of 0.0023 pm 0.0007 across all categories. Training the vision--language anomaly detection model AA-CLIP on E-SCDD, compared to the SCDD dataset, improves pixel-level AUC and average precision by 1.70 and 8.34 points, respectively.

The ability to estimate the quality of scientific papers is central to how both humans and AI systems will advance scientific knowledge in the future. However, existing LLM-based estimation methods suffer from high inference cost, whereas the faster direct score regression approach is limited by scale inconsistencies. We present NAIPv2, a debiased and efficient framework for paper quality estimation. NAIPv2 employs pairwise learning within domain-year groups to reduce inconsistencies in reviewer ratings and introduces the Review Tendency Signal (RTS) as a probabilistic integration of reviewer scores and confidences. To support training and evaluation, we further construct NAIDv2, a large-scale dataset of 24,276 ICLR submissions enriched with metadata and detailed structured content. Trained on pairwise comparisons but enabling efficient pointwise prediction at deployment, NAIPv2 achieves state-of-the-art performance (78.2% AUC, 0.432 Spearman), while maintaining scalable, linear-time efficiency at inference. Notably, on unseen NeurIPS submissions, it further demonstrates strong generalization, with predicted scores increasing consistently across decision categories from Rejected to Oral. These findings establish NAIPv2 as a debiased and scalable framework for automated paper quality estimation, marking a step toward future scientific intelligence systems. Code and dataset are released at https://sway.cloud.microsoft/Pr42npP80MfPhvj8.

Time series classification (TSC) is dominated by a two-stage paradigm: train a feature encoder -- either from scratch on the target dataset or via pretraining on large corpora -- and then fit a task-specific classifier on top. While effective, this decoupling optimizes representation learning independently of the classification objective, requires per-dataset training, and prevents the model from exploiting label information during inference. We introduce TimEE, a 4.5M-parameter foundation model for end-to-end TSC via in-context learning. Given a labeled support set and a query time series, TimEE directly outputs a predicted class distribution in a single forward pass with no per-dataset training required. Following the prior-data fitted network (PFN) framework, TimEE is meta-trained exclusively on synthetic TSC tasks, where each task contains time series with distinct class identities arising from structured distributional shifts in the generative process. Despite seeing no real time series during pre-training, TimEE ranks first in ROC AUC (and third on accuracy) on the UCR benchmark among all compared methods, which include both foundation models and supervised deep learning baselines. To our knowledge, TimEE is the first purely synthetic-pretrained model to reach state-of-the-art performance on the UCR benchmark. These results establish end-to-end ICL with synthetic priors as a compelling, largely unexplored direction for TSC, with scaling, prior design, and richer generation mechanisms as natural avenues for improvement. Code is publicly available at http://github.com/automl/timee.

Scaling Transformer-based click-through rate (CTR) models by stacking more parameters brings growing computational and storage overhead, creating a widening gap between scaling ambitions and the stringent industrial deployment constraints. We propose LoopCTR, which introduces a loop scaling paradigm that increases training-time computation through recursive reuse of shared model layers, decoupling computation from parameter growth. LoopCTR adopts a sandwich architecture enhanced with Hyper-Connected Residuals and Mixture-of-Experts, and employs process supervision at every loop depth to encode multi-loop benefits into the shared parameters. This enables a train-multi-loop, infer-zero-loop strategy where a single forward pass without any loop already outperforms all baselines. Experiments on three public benchmarks and one industrial dataset demonstrate state-of-the-art performance. Oracle analysis further reveals 0.02--0.04 AUC of untapped headroom, with models trained with fewer loops exhibiting higher oracle ceilings, pointing to a promising frontier for adaptive inference.

Autonomous agents, especially delegated systems with memory, persistent context, and multi-step planning, pose a measurement problem not present in stateless models: an agent that preserves continued operation as a terminal objective and one that does so merely instrumentally can produce observationally similar trajectories. External behavioral monitoring cannot reliably distinguish between them. We introduce the Unified Continuation-Interest Protocol (UCIP), a multi-criterion detection framework that moves this distinction from behavior to the latent structure of agent trajectories. UCIP encodes trajectories with a Quantum Boltzmann Machine (QBM), a classical algorithm based on the density-matrix formalism of quantum statistical mechanics, and measures the von Neumann entropy of the reduced density matrix induced by a bipartition of hidden units. We test whether agents with terminal continuation objectives (Type A) produce latent states with higher entanglement entropy than agents whose continuation is merely instrumental (Type B). Higher entanglement reflects stronger cross-partition statistical coupling. On gridworld agents with known ground-truth objectives, UCIP achieves 100% detection accuracy and 1.0 AUC-ROC on held-out non-adversarial evaluation under the frozen Phase I gate. The entanglement gap between Type A and Type B agents is Delta = 0.381 (p < 0.001, permutation test). Pearson r = 0.934 across an 11-point interpolation sweep indicates that, within this synthetic family, UCIP tracks graded changes in continuation weighting rather than merely a binary label. Among the tested models, only the QBM achieves positive Delta. All computations are classical; "quantum" refers only to the mathematical formalism. UCIP does not detect consciousness or subjective experience; it detects statistical structure in latent representations that correlates with known objectives.

We address the challenge of detecting rare and diverse anomalies in surveillance videos using only video-level supervision. Our dual-backbone framework combines convolutional and transformer representations through top-k pooling, achieving 90.7% area under the curve (AUC) on the UCF-Crime dataset.

User specifications or legal frameworks often require information to be removed from pretrained models, including large language models (LLMs). This requires deleting or "forgetting" a set of data points from an already-trained model, which typically degrades its performance on other data points. Thus, a balance must be struck between removing information and keeping the model's other abilities intact, with a failure to balance this trade-off leading to poor deletion or an unusable model. To this end, we propose UPCORE (Utility-Preserving Coreset Selection), a method-agnostic data selection framework for mitigating collateral damage during unlearning. Finding that the model damage is correlated with the variance of the model's representations on the forget set, we selectively prune the forget set to remove outliers, thereby minimizing model degradation after unlearning. We evaluate UPCORE across three standard unlearning methods consistently achieving a superior balance between the competing objectives of deletion efficacy and model preservation. To better evaluate this trade-off, we introduce a new metric, measuring the area-under-the-curve (AUC) across standard metrics. We find that UPCORE improves both standard metrics and AUC, benefitting from positive transfer between the coreset and pruned points while reducing negative transfer from the forget set to points outside of it.

OpenAI's GPT-Image-2 has effectively erased the visual boundary between authentic and AI-edited document images: a single number on a receipt can be replaced in under a second for a few cents. We release AIForge-Doc v2, a paired dataset of 3,066 GPT-Image-2 document forgeries with pixel-precise masks in DocTamper-compatible format, and benchmark four lines of defence: human inspectors (N=120, n=365 pair-votes via the public 2AFC site CanUSpotAI.com), TruFor (generic forensic), DocTamper (qcf-568, document-specific), and the same GPT-Image-2 model as a zero-shot self-judge -- asked, to avoid the trivial "image is mostly real" reading, whether any region was generated or edited by an AI image model. Human 2AFC accuracy is 0.501, indistinguishable from chance: even side-by-side, inspectors cannot tell GPT-Image-2 receipt forgeries from authentic counterparts. The three computational judges sit only modestly above (TruFor 0.599, DocTamper 0.585, self-judge 0.532). The self-judge fails consistently, not by chance: across five prompt strategies and four policies for handling ambiguous responses, AUC never rises above 0.59. To rule out the possibility that the two forensic detectors are broken on our source domain rather than blind to AI inpainting, we calibrate each on a same-domain traditional-tampering set built for its training distribution: TruFor reaches AUC 0.962 on cross-camera splicing of our dataset, DocTamper reaches 0.852 on cross-document OCR-token splicing with two-pass JPEG re-encoding. Both retain near-published performance on traditional tampering; switching to GPT-Image-2 inpainting drops AUC by 0.27-0.36 (0.962->0.599 TruFor; 0.852->0.585 DocTamper), isolating a detection gap specific to GPT-Image-2 inpainting. We release the dataset, pipeline, four-judge protocol, and calibration sets.

We present H-Node Adversarial Noise Cancellation (H-Node ANC), a mechanistic framework that identifies, exploits, and defends hallucination representations in transformer-based large language models (LLMs) at the level of individual hidden-state dimensions. A logistic regression probe trained on last-token hidden states localizes hallucination signal to a small set of high-variance dimensions -- termed Hallucination Nodes (H-Nodes) -- with probe AUC reaching 0.90 across four architectures. A white-box adversarial attack amplifies these dimensions at inference time via a real-time forward hook, achieving a selectivity of 3.02x with less than 10% visibility to the defender. Adaptive ANC defense suppresses H-Node excess in-pass using confidence-weighted cancellation, reducing grounded activation drift by 33-42% over static cancellation. A dynamic iterative extension that re-ranks cancellation targets across successive passes recovers up to 0.69 robustness from a single-pass baseline of 8%. All contributions are validated on OPT-125M, Phi-3-mini-4k-instruct, LLaMA-3-8B-Instruct, and Mistral-7B-Instruct-v0.3 (125M-8B parameters). Perplexity impact is surgical (<5%) and MMLU degradation is at most 3%, confirming that the defense does not impair general reasoning capability.

Accurate localization of tumor regions from hematoxylin and eosin-stained whole-slide images is fundamental for translational research including spatial analysis, molecular profiling, and tissue architecture investigation. However, deep learning-based tumor detection trained within specific cancers may exhibit reduced robustness when applied across different tumor types. We investigated whether balanced training across cancers at modest scale can achieve high performance and generalize to unseen tumor types. A multi-cancer tumor localization model (MuCTaL) was trained on 79,984 non-overlapping tiles from four cancers (melanoma, hepatocellular carcinoma, colorectal cancer, and non-small cell lung cancer) using transfer learning with DenseNet169. The model achieved a tile-level ROC-AUC of 0.97 in validation data from the four training cancers, and 0.71 on an independent pancreatic ductal adenocarcinoma cohort. A scalable inference workflow was built to generate spatial tumor probability heatmaps compatible with existing digital pathology tools. Code and models are publicly available at https://github.com/AivaraX-AI/MuCTaL.

This work proposes a framework for optimizing machine learning algorithms. The practicality of the framework is illustrated using an important case study from the healthcare domain, which is predicting the admission status of emergency department (ED) patients (e.g., admitted vs. discharged) using patient data at the time of triage. The proposed framework can mitigate the crowding problem by proactively planning the patient boarding process. A large retrospective dataset of patient records is obtained from the electronic health record database of all ED visits over three years from three major locations of a healthcare provider in the Midwest of the US. Three machine learning algorithms are proposed: T-XGB, T-ADAB, and T-MLP. T-XGB integrates extreme gradient boosting (XGB) and Tabu Search (TS), T-ADAB integrates Adaboost and TS, and T-MLP integrates multi-layer perceptron (MLP) and TS. The proposed algorithms are compared with the traditional algorithms: XGB, ADAB, and MLP, in which their parameters are tunned using grid search. The three proposed algorithms and the original ones are trained and tested using nine data groups that are obtained from different feature selection methods. In other words, 54 models are developed. Performance was evaluated using five measures: Area under the curve (AUC), sensitivity, specificity, F1, and accuracy. The results show that the newly proposed algorithms resulted in high AUC and outperformed the traditional algorithms. The T-ADAB performs the best among the newly developed algorithms. The AUC, sensitivity, specificity, F1, and accuracy of the best model are 95.4%, 99.3%, 91.4%, 95.2%, 97.2%, respectively.

Left ventricular hypertrophy (LVH) results from chronic remodeling caused by a broad range of systemic and cardiovascular disease including hypertension, aortic stenosis, hypertrophic cardiomyopathy, and cardiac amyloidosis. Early detection and characterization of LVH can significantly impact patient care but is limited by under-recognition of hypertrophy, measurement error and variability, and difficulty differentiating etiologies of LVH. To overcome this challenge, we present EchoNet-LVH - a deep learning workflow that automatically quantifies ventricular hypertrophy with precision equal to human experts and predicts etiology of LVH. Trained on 28,201 echocardiogram videos, our model accurately measures intraventricular wall thickness (mean absolute error [MAE] 1.4mm, 95% CI 1.2-1.5mm), left ventricular diameter (MAE 2.4mm, 95% CI 2.2-2.6mm), and posterior wall thickness (MAE 1.2mm, 95% CI 1.1-1.3mm) and classifies cardiac amyloidosis (area under the curve of 0.83) and hypertrophic cardiomyopathy (AUC 0.98) from other etiologies of LVH. In external datasets from independent domestic and international healthcare systems, EchoNet-LVH accurately quantified ventricular parameters (R2 of 0.96 and 0.90 respectively) and detected cardiac amyloidosis (AUC 0.79) and hypertrophic cardiomyopathy (AUC 0.89) on the domestic external validation site. Leveraging measurements across multiple heart beats, our model can more accurately identify subtle changes in LV geometry and its causal etiologies. Compared to human experts, EchoNet-LVH is fully automated, allowing for reproducible, precise measurements, and lays the foundation for precision diagnosis of cardiac hypertrophy. As a resource to promote further innovation, we also make publicly available a large dataset of 23,212 annotated echocardiogram videos.

The fairness of machine learning-based decisions has become an increasingly important focus in the design of supervised machine learning methods. Most fairness approaches optimize a specified trade-off between performance measure(s) (e.g., accuracy, log loss, or AUC) and fairness metric(s) (e.g., demographic parity, equalized odds). This begs the question: are the right performance-fairness trade-offs being specified? We instead re-cast fair machine learning as an imitation learning task by introducing superhuman fairness, which seeks to simultaneously outperform human decisions on multiple predictive performance and fairness measures. We demonstrate the benefits of this approach given suboptimal decisions.