Abstract--Global population aging presents increasing challenges to healthcare systems, with coronary artery disease (CAD) responsible for approximately 17.8 million deaths annually, making it a leading cause of global mortality . As CAD is largely preventable, early detection and proactive management are essential. In this work, we introduce DigitalShadow, an advanced early warning system for CAD, powered by a fine-tuned facial foundation model. The system is pre-trained on 21 million facial images and subsequently fine-tuned into LiveCAD, a specialized CAD risk assessment model trained on 7,004 facial images from 1,751 subjects across four hospitals in China. DigitalShadow functions passively and contactlessly, extracting facial features from live video streams without requiring active user engagement. Integrated with a personalized database, it generates natural language risk reports and individualized health recommendations. With privacy as a core design principle, DigitalShadow supports local deployment to ensure secure handling of user data. The world's population is rapidly ageing [1], with significant implications for the prevalence of chronic diseases such as Coronary Artery Disease (CAD) [2], affecting not only individuals but also families and societies at large [3]. The number of older people is increasing at an unprecedented rate, projected to grow from approximately 761 million in 2021 to 1.6 billion by 2050, which would represent nearly 16% of the global population, according to the UN's W orld Social Report 2023 [4]. The aging population presents numerous challenges, including increased pressure on healthcare systems, pension schemes, and long-term care facilities, alongside potential economic consequences, which together fuel growing demand for healthcare services [5], [6]. With advancing age, people become more vulnerable to various critical diseases [7], such as CAD [8], stroke [9], cancer [10], and Parkinson's disease (PD) [11], [12], [13], leading to considerable morbidity and mortality [14].

Recent advances have established the identifiability of a directed acyclic graph (DAG) under additive noise models (ANMs), spurring the development of various causal discovery methods. However, most existing methods make restrictive model assumptions, rely heavily on general independence tests, or require substantial computational time. To address these limitations, we propose a sequential procedure to orient undirected edges in a completed partial DAG (CPDAG), representing an equivalence class of DAGs, by leveraging the pairwise additive noise model (PANM) to identify their causal directions. We prove that this procedure can recover the true causal DAG assuming a restricted ANM. Building on this result, we develop a novel constraint-based algorithm for learning causal DAGs under nonlinear ANMs. Given an estimated CPDAG, we develop a ranking procedure that sorts undirected edges by their adherence to the PANM, which defines an evaluation order of the edges. To determine the edge direction, we devise a statistical test that compares the log-likelihood values, evaluated with respect to the competing directions, of a sub-graph comprising just the candidate nodes and their identified parents in the partial DAG. We further establish the structural learning consistency of our algorithm in the large-sample limit. Extensive experiments on synthetic and real-world datasets demonstrate that our method is computationally efficient, robust to model misspecification, and consistently outperforms many existing nonlinear DAG learning methods.

Depression is one of the most prevalent mental health disorders worldwide, with estimates indicating that around 5% of the worlds' adult population [1, 2] suffers from this condition. The primary methods for screening and monitoring depression rely on self-reported questionnaires, such as the Patient Health Questionnaire (PHQ-9) [3], Beck's Depression Inventory (BDI) [4] and Hamilton Depression Ratings Scale (HDRS) [5]. While these questionnaires are effective to varying degrees at screening patients for depression, they provide only limited information about the affected underlying neuro-cognitive processes in individuals, limiting the ability to personalize treatments. Given the heterogeneity of depressive symptomatology across patient populations [6, 7], it is crucial to elucidate the underlying neurophysiological mechanisms to support the development of more effective and individualized procedures for screening, monitoring, and treatment. Prior functional imaging studies have identified increased activity in anterior cin-gulate cortex (especially the subgenual anterior cingulate) during presentation of emotional stimuli, altered connectivity in prefrontal cortical areas, and default mode network as potential differentiating markers in depressed participants [8-13].

Growing concerns over data privacy and security highlight the importance of machine unlearning--removing specific data influences from trained models without full retraining. Techniques like Membership Inference Attacks (MIAs) are widely used to externally assess successful unlearning. However, existing methods face two key limitations: (1) maximizing MIA effectiveness (e.g., via online attacks) requires prohibitive computational resources, often exceeding retraining costs; (2) MIAs, designed for binary inclusion tests, struggle to capture granular changes in approximate unlearning. To address these challenges, we propose the Interpolated Approximate Measurement (IAM), a framework natively designed for unlearning inference. IAM quantifies sample-level unlearning completeness by interpolating the model's generalization-fitting behavior gap on queried samples. IAM achieves strong performance in binary inclusion tests for exact unlearning and high correlation for approximate unlearning--scalable to LLMs using just one pre-trained shadow model. We theoretically analyze how IAM's scoring mechanism maintains performance efficiently. We then apply IAM to recent approximate unlearning algorithms, revealing general risks of both over-unlearning and under-unlearning, underscoring the need for stronger safeguards in approximate unlearning systems. The code is available at https://github.com/Happy2Git/Unlearning_Inference_IAM.

Large Language Models (LLMs) have revolutionized Natural Language Processing by excelling at interpreting, reasoning about, and generating human language. However, their reliance on large-scale, often proprietary datasets poses a critical challenge: unauthorized usage of such data can lead to copyright infringement and significant financial harm. Existing dataset-inference methods typically depend on log probabilities to detect suspicious training material, yet many leading LLMs have begun withholding or obfuscating these signals. This reality underscores the pressing need for label-only approaches capable of identifying dataset membership without relying on internal model logits. We address this gap by introducing CatShift, a label-only dataset-inference framework that capitalizes on catastrophic forgetting: the tendency of an LLM to overwrite previously learned knowledge when exposed to new data. If a suspicious dataset was previously seen by the model, fine-tuning on a portion of it triggers a pronounced post-tuning shift in the model's outputs; conversely, truly novel data elicits more modest changes. By comparing the model's output shifts for a suspicious dataset against those for a known non-member validation set, we statistically determine whether the suspicious set is likely to have been part of the model's original training corpus. Extensive experiments on both open-source and API-based LLMs validate CatShift's effectiveness in logit-inaccessible settings, offering a robust and practical solution for safeguarding proprietary data.

Large language models frequently encounter conflicts between their parametric knowledge and contextual input, often resulting in factual inconsistencies or hallucinations. We propose Self-Reflective Debate for Contextual Reliability (SR-DCR), a lightweight framework that integrates token-level self-confidence with an asymmetric multi-agent debate to adjudicate such conflicts. A critic, deprived of context, challenges a defender who argues from the given passage; a judge model evaluates the debate and determines the context's reliability. The final answer is selected by combining the verdict with model confidence. Experiments on the ClashEval benchmark demonstrate that SR-DCR consistently enhances robustness to misleading context while maintaining accuracy on trustworthy inputs, outperforming both classical debate and confidence-only baselines with minimal computational overhead. The code is available at https://github.com/smiles724/Self-Reflective-Debates.

Watermarking becomes one of the pivotal solutions to trace and verify the origin of synthetic images generated by artificial intelligence models, but it is not free of risks. Recent studies demonstrate the capability to forge watermarks from a target image onto cover images via adversarial optimization without knowledge of the target generative model and watermark schemes. In this paper, we uncover a greater risk of an optimization-free and universal watermark forgery that harnesses existing regenerative diffusion models. Our proposed forgery attack, PnP (Plug-and-Plant), seamlessly extracts and integrates the target watermark via regenerating the image, without needing any additional optimization routine. It allows for universal watermark forgery that works independently of the target image's origin or the watermarking model used. We explore the watermarked latent extracted from the target image and visual-textual context of cover images as priors to guide sampling of the regenerative process. Extensive evaluation on 24 scenarios of model-data-watermark combinations demonstrates that PnP can successfully forge the watermark (up to 100% detectability and user attribution), and maintain the best visual perception. By bypassing model retraining and enabling adaptability to any image, our approach significantly broadens the scope of forgery attacks, presenting a greater challenge to the security of current watermarking techniques for diffusion models and the authority of watermarking schemes in synthetic data generation and governance.

To what extent do vision-and-language foundation models possess a realistic world model (observation $\times$ action $\rightarrow$ observation) and a dynamics model (observation $\times$ observation $\rightarrow$ action), when actions are expressed through language? While open-source foundation models struggle with both, we find that fine-tuning them to acquire a dynamics model through supervision is significantly easier than acquiring a world model. In turn, dynamics models can be used to bootstrap world models through two main strategies: 1) weakly supervised learning from synthetic data and 2) inference time verification. Firstly, the dynamics model can annotate actions for unlabelled pairs of video frame observations to expand the training data. We further propose a new objective, where image tokens in observation pairs are weighted by their importance, as predicted by a recognition model. Secondly, the dynamics models can assign rewards to multiple samples of the world model to score them, effectively guiding search at inference time. We evaluate the world models resulting from both strategies through the task of action-centric image editing on Aurora-Bench. Our best model achieves a performance competitive with state-of-the-art image editing models, improving on them by a margin of $15\%$ on real-world subsets according to GPT4o-as-judge, and achieving the best average human evaluation across all subsets of Aurora-Bench.

Membership inference tests aim to determine whether a particular data point was included in a language model's training set. However, recent works have shown that such tests often fail under the strict definition of membership based on exact matching, and have suggested relaxing this definition to include semantic neighbors as members as well. In this work, we show that membership inference tests are still unreliable under this relaxation - it is possible to poison the training dataset in a way that causes the test to produce incorrect predictions for a target point. We theoretically reveal a trade-off between a test's accuracy and its robustness to poisoning. We also present a concrete instantiation of this poisoning attack and empirically validate its effectiveness. Our results show that it can degrade the performance of existing tests to well below random.

Competitive programming, due to its high reasoning difficulty and precise correctness feedback, has become a key task for both training and evaluating the reasoning capabilities of large language models (LLMs). However, while a large amount of public problem data, such as problem statements and solutions, is available, the test cases of these problems are often difficult to obtain. Therefore, test case generation is a necessary task for building large-scale datasets, and the quality of the test cases directly determines the accuracy of the evaluation. In this paper, we introduce an LLM-based agent system that creates high-quality test cases for competitive programming problems. We apply this system to the CodeContests dataset and propose a new version with improved test cases, named CodeContests+. We evaluated the quality of test cases in CodeContestsPlus. First, we used 1.72 million submissions with pass/fail labels to examine the accuracy of these test cases in evaluation. The results indicated that CodeContests+ achieves significantly higher accuracy than CodeContests, particularly with a notably higher True Positive Rate (TPR). Subsequently, our experiments in LLM Reinforcement Learning (RL) further confirmed that improvements in test case quality yield considerable advantages for RL.