Self-supervised learning (SSL) is now a standard way to pretrain medical image models, but performance is still mostly judged by downstream accuracy. For safety-critical screening tasks such as diabetic retinopathy grading, this is not enough: a model must also know when its predictions are unreliable and defer uncertain cases for clinical review. In this work, we examine how the length of SSL pretraining influences confidence calibration and confidence-based abstention. We evaluate multiple SSL checkpoints under a fixed fine-tuning protocol and assess calibrated confidence, coverage, selective accuracy, and selective macro-F1. Across datasets and data regimes, SSL pretraining improves selective prediction compared to training from scratch. Unlike prior SSL studies that primarily evaluate downstream accuracy or AUROC, we analyze how SSL pretraining duration influences calibration and selective prediction behavior under confidence-based abstention. However, once accuracy saturates, selective performance can still change markedly across checkpoints, and longer pretraining does not consistently improve reliability. These results underscore the importance of abstention-aware evaluation and suggest that pretraining length should be treated as an important reliability-related design choice rather than only a computational detail. Code is available at https://github.com/29 muskaan712/ijcai-knowing-when-not-to-predict.
论文检索
输入标题、作者或关键词，从 100,903 篇学术成果中精准定位
Contemporary glioma diagnosis integrates molecular features with histopathology to guide clinical decision-making. However, in clinical settings, divergent imaging protocols result in incomplete MRI sequences, leading to two primary challenges: forcing existing frameworks to discard a large portion of clinical data during training and consequently limiting their clinical applicability. To address these limitations, we propose GMENet, a Generative Mixture of Experts Network for multi-center glioma diagnosis with incomplete imaging sequences. Firstly, we design a Cross-attention-based Gated Generation Module that synthesizes missing sequence features from available sequences via cross-attention and dynamic gating mechanisms, incorporating a cycle-consistency loss to preserve semantic integrity. Secondly, we introduce a Dynamically Weighted Experts Fusion Module that performs mixture-of-experts interaction and confidence-aware fusion over original and synthesized dual-sequence features for multi-task prediction. We evaluate GMENet on a multi-center cohort of 1,241 subjects from four in-house datasets and two public repositories. Experiments show that GMENet expands clinically usable training data by 97%, relative to complete-sequence-only data. Furthermore, it consistently outperforms state-of-the-art methods trained on complete data, demonstrating improved robustness under cross-center distribution shifts. Code is available at: https://github.com/spf-sd/GMENet.
Existing physiological foundation models face two limitations: rigid hard-negative sampling indiscriminately repels morphologically similar samples, distorting the natural manifold; and coarse discretization strategies sever the intrinsic continuity of physiological states, inducing precision loss. To address these challenges, we propose DNA-PPG, a novel pre-training framework anchored in Dual Neighborhood Alignment. DNA-PPG integrates the Morphology-Aware Self-Supervised Branch using Time-Frequency Soft Weighting to capture universal signal dynamics shared among diverse subjects, with the Physiological Semantic Alignment Branch that projects physiological indicators into continuous semantic space to explicitly embed precise physiological priors into the representation space. We scale the pre-training to 10.7 million PPG segments from over 8,400 subjects to ensure robust generalization. Extensive evaluations on six downstream benchmarks demonstrate that DNA-PPG significantly outperforms state-of-the-art baselines, achieving an 18% reduction in regression error and an 11.5% improvement in classification performance. These results validate DNA-PPG as a robust, universal feature extractor for diverse photoplethysmography applications.
Identifying novel hypotheses is essential to scientific research, yet this process risks being overwhelmed by the sheer volume and complexity of available information. Existing automated methods often struggle to generate novel and evidence-grounded hypotheses, lack robust iterative refinement and rarely undergo rigorous temporal evaluation for future discovery potential. To address this, we propose BIODISCO, a multi-agent framework that draws upon language model-based reasoning and a dual-mode evidence system (biomedical knowledge graphs and automated literature retrieval) for grounded novelty, integrates an internal scoring and feedback loop for iterative refinement, and validates performance through pioneering temporal and human evaluations and a Bradley-Terry paired comparison model for statistical assessment. Evaluations suggest improved novelty and significance relative to ablated configurations and a generalist biomedical agent. Designed for flexibility and modularity, BIODISCO allows seamless integration of custom language models or knowledge graphs, and can be run with just a few lines of code.
Medical image segmentation models based on Vision Mamba architectures have demonstrated strong performance, along with improved efficiency compared to convolutional and transformer-based approaches. However, existing lightweight and ultralight variants often suffer from boundary degradation and inconsistent shape prediction, thereby limiting their clinical reliability. We propose TvaraNet, an extremely lightweight segmentation network designed to preserve boundary fidelity under strict efficiency constraints. TvaraNet contains only 0.037M parameters and requires only 0.060 GFLOPs, enabling efficient deployment in resource-constrained environments. The architecture introduces two core lightweight modules. Parallel Skip Mamba Averaging (PASMA) enhances long-range dependency modeling by injecting globally aggregated channel context into mamba blocks. Dilated Asymmetric Spatial Mixer Attention (DASMA) improves boundary-aware feature refinement in skip connections through efficient multi-scale spatial modulation. In addition, training-time regularization strategies, including an Adaptive Gated Gradient Reversal Layer (AGGRL), an adaptive Singular Value Decomposition (SVD) loss, and a Boundary-weighted Cross-Entropy loss (BWCE), are employed to suppress redundant features and emphasize object boundaries without adding inference overhead. Experiments on ISIC-17, ISIC-18, and a spinal segmentation dataset demonstrate that TvaraNet consistently outperforms existing lightweight models and achieves competitive or superior boundary-aware performance compared to heavier architectures. Notably, it improves Mean Intersection over Union (mIoU) by +1.50, +1.42, and +3.35 over Ultralight VMUNet on the respective datasets, establishing TvaraNet as a practical solution for boundary-consistent medical image segmentation on edge and low-resource platforms. Code is available at: https://github.com/MindsLab-GitHub/TvaraNet
Foundation models are reshaping medical imaging, yet their application in echocardiography remains limited, hindered by a heavy reliance on private datasets that prevent reproducible comparison. Echocardiography poses unique challenges, including noisy acquisitions, high frame redundancy, and limited diverse public datasets. To address this, we introduce CardioBench, a comprehensive benchmark for echocardiography foundation models. Specifically, CardioBench unifies eight publicly available datasets into a standardized suite spanning four regression and five classification tasks, covering functional, structural, diagnostic, and view recognition endpoints. Leveraging this framework, we evaluate several leading foundation models, including cardiac-specific, biomedical, and general-purpose encoders, under consistent zero-shot, probing, and alignment protocols. Our analysis reveals that while general-purpose encoders transfer well and often close the gap with probing, they struggle significantly with fine-grained distinctions like view classification and subtle pathology recognition. Results indicate that models capturing temporal cardiac dynamics perform best on functional tasks, while retrieval-based approaches generalize more consistently across datasets. By releasing preprocessing, splits, and public evaluation pipelines, CardioBench establishes a reproducible reference point to guide the architectural design of future echocardiography and possibly other medical imaging foundation models.
Electroencephalogram (EEG) signals are vital for automated seizure detection, but their inherent noise makes robust representation learning challenging. Existing graph construction methods, whether correlation-based or learning-based, often generate redundant or irrelevant edges due to the noisy nature of EEG data. This significantly impairs the quality of graph representations and limits downstream task performance. Motivated by the remarkable reasoning and contextual understanding capabilities of large language models (LLMs), we explore the idea of using LLMs as graph edge refiners. Specifically, we propose a two-stage framework: we first verify that LLM-based edge refinement can effectively identify and remove redundant connections, leading to significant improvements in seizure detection accuracy and more meaningful graph structures. Building on this insight, we further develop a robust solution where the initial graph is constructed using a transformer-based edge predictor and multilayer perceptron, assigning probability scores to potential edges and applying a threshold to determine their existence. The LLM then acts as an edge set refiner, making informed decisions based on both textual and statistical features of node pairs to validate the remaining connections. Extensive experiments on TUSZ dataset demonstrate that our LLM-refined graph learning framework not only enhances task performance but also yields cleaner and more interpretable graph representations.
Physiological signals are widely used for health assessment in clinical and daily-life settings. Established physiological signals collected for inpatient and clinical use are often impractical for patients' daily home use due to their complexity and resource demands. In contrast, wearable signals enable continuous monitoring in everyday life, but many have limited reliability and are not widely understood or accepted in clinical practices. To leverage complementary strengths of clinical and wearable physiological signals, we propose PhysioGMC, a Generalizable Multi-modal Coordination framework for Physiological signals that explicitly accounts for their strong inter-subject variability. PhysioGMC incorporates both clinical and wearable modalities into the training process to improve cross-subject performance when only a single wearable modality is available at deployment. The framework introduces a cross-modal contrastive learning module comprising two contrastive losses to jointly learn label-relevant, subject-agnostic representations across modalities. The self-supervised contrastive loss aligns latent features across modalities, while the supervised contrastive loss encourages learning label-discriminative features that are invariant to subject identity. Experiments on cardiovascular health monitoring and sleep staging tasks demonstrate that PhysioGMC consistently outperforms existing methods, achieving superior cross-subject performance at test time using only wearable modalities, such as photoplethysmography (PPG).
Predicting single-cell gene responses to chemical perturbations is vital for personalized therapy, yet existing deep learning models face significant hurdles. Standard regression-based approaches suffer from "mean prediction bias," failing to capture cellular heterogeneity, while current Flow Matching (FM) methods struggle with the extreme sparsity of single-cell data and the dimensionality mismatches inherent in transformer-based generation. To address these challenges, we introduce sc2Flow, a dual-stage framework that unifies discrete and continuous flow matching. sc2Flow first predicts the binary mask of expressed genes and subsequently models their quantitative levels using a scalable Transformer. This decoupling effectively resolves dimensionality conflicts and eliminates parameter redundancy. Extensive benchmarks on Sci-Plex3 and other datasets demonstrate that sc2Flow significantly outperforms state-of-the-art baselines on distribution matching, successfully mitigating mean bias to preserve critical biological heterogeneity.
Whole Slide Image (WSI) analysis faces challenges due to gigapixel resolutions and slide-level weak supervision. Multiple Instance Learning (MIL) serves as a pivotal method for this task. However, existing MIL frameworks often fail to exploit the inherent redundancy of tissue patterns or the semantic coherence among similar patches within a WSI. We propose GroupMIL, a novel framework that introduces a differentiable grouping mechanism into the MIL framework. This approach enables the automatic emergence of semantic segments using only slide-level labels. We specifically introduce a multi-stage grouping block and a hierarchical aggregator, which progressively fuse features within and across groups to construct a robust slide-level representation. Extensive experiments on multiple public datasets across cancer subtyping and survival prediction tasks demonstrate that GroupMIL consistently surpasses state-of-the-art performance.
Central nervous system (CNS) drug discovery is constrained by an immense and sparse chemical search space. Meanwhile, molecules that simultaneously achieve brain penetration, target efficacy, and synthesizability are extremely scarce. However, existing generative models rarely couple rigorous multi-objective control with robustness to distribution shift, limiting their reliability in realistic CNS design. We introduce D²G-TO, a task-aware and out of-distribution (OOD)-guided discrete graph diffusion framework that unifies multi-pharmacological properties with structural distribution guidance. A novel Structural Similarity Guidance mechanism steers generation toward in-distribution regions while repelling OOD modes, maintaining structural distributional consistency in realistic scenarios. Across BBBP, BACE, and QM9 benchmarks, D²G-TO achieves strong validity, diversity, and other metrics. In an Alzheimer's disease case study, we subject the generated molecules to cross-property pharmacological prediction and systematic ADMET profiling, followed by structure based molecular docking against BACE-1 to assess binding-mode plausibility. D²G-TO identifies candidates that jointly satisfy blood–brain barrier permeability, β-site amyloid precursor protein cleaving enzyme 1 inhibition, and synthetic accessibility. Thus, D²G-TO has the potential to serve as an efficient in silico engine for early-stage CNS drug design. The code is available at https://github.com/zhaix922/DDG_TO.
Predicting drug-disease associations (DDAs) plays a crucial role in drug development and disease treatment. However, existing researches predominantly focus on single DDAs prediction task, often overlooking the intricate relationships among different tasks, which can further improve the performance of methods for DDAs prediction. To address this limitation, a multi-task prediction framework, capable of simultaneously predicting drug-disease, drug-protein, and disease-protein associations, is proposed, named MTP-DDA. The framework constructs three distinct graphs to reflect different relationships between biological entities. Then, based on these graphs, two sub-views and one main-view are constructed. For sub-view, corruption strategy is adopted to generate corrupted view, and Graph Convolutional Network (GCN) is employed to extract features from both the original view and its corrupted version, with contrastive learning applied to enhance feature representations. For main-view, GCN and Node2Vec are utilized to extract low-order and high-order node features respectively, and an attention mechanism is utilized for feature fusion. Finally, the node features from three views above are integrated, and the dot product operation is applied to the node features of association pairs to derive association scores, thereby enabling multi-task association prediction. Under 10-fold cross-validation, the proposed framework outperforms current methods on public datasets, demonstrating its effectiveness and robustness.
Recent advances in vision-language models (VLMs) have shown remarkable performance in medical image classification tasks. However, applying VLMs to fetal cardiac ultrasound (FCU) remains challenging due to compound distribution shifts, including covariate shifts caused by cross-center heterogeneity and semantic shifts arising from clinically non-standard views. To address this issue, we propose Dual-Level Contrastive Learning (DLCL), the first prompt-based VLM framework for out-of-distribution (OOD) detection in FCU, to the best of our knowledge. DLCL explicitly shapes the vision-language representation space through complementary local-level and global-level contrastive objectives. Specifically, local contrastive learning aligns instance-level features to mitigate covariate shifts, whereas global contrastive learning regularizes global prototypes to address semantic shifts. We conduct extensive experiments on a private multi-center FCU dataset and the public ISIC-OOD dataset to validate the proposed approach. On the challenging FCU task, DLCL achieves an AUROC of 89.61% and a harmonic mean of 80.35%, significantly outperforming state-of-the-art methods.
Retrieving real time information is a fundamental capability for agents with search capabilities in real world applications. However, existing benchmarks are predominantly static and therefore fail to capture the temporal dynamics of information and the continuously evolving nature of real world knowledge. To address this limitation, we propose RT-QA, a dynamic evaluation framework that leverages executable code workflows to retrieve current answers at evaluation time. Specifically, we construct an agent based pipeline that autonomously generates code for web crawling and DOM based answer extraction to produce real time ground truth. To ensure robust evaluation over time, the pipeline further incorporates a repair mechanism to adapt to changes in web page structures. RT-QA spans 12 domains, such as Finance and Sports, with 320 Chinese questions categorized into three difficulty levels. Extensive evaluations of advanced models, such as GPT-5.2 and GLM-4.7, reveal significant limitations in real time adaptability: even the best models achieve only 46% accuracy. Our analysis highlights two primary failure modes: (1) Lazy Retrieval, where agents rely on search snippets instead of deeply scanning specific websites for information, accounting for 20% of failures; and (2) Temporal Confusion, a cognitive error where agents retrieve a historical date, for example an event in 2024, and fail to anchor again to the current time, 2026, for subsequent reasoning. These findings suggest that future agents require not just better retrieval strategies, but robust temporal state management.
Achieving an optimal balance between exploration and exploitation remains a fundamental challenge in reinforcement learning. This work revisits the exploration-exploitation dilemma through the lens of entropy, offering a novel perspective on this enduring problem. It establishes a theoretical connection between policy's entropy and exploratory behavior, using entropy as a rational measure to quantify the exploration-exploitation trade-off. Theoretical analyses demonstrate that a modified Bellman equation, augmented with a novelty-seeking term, ensures appropriate entropy adjustment and guarantees its globally monotonic decay. The derived policy optimization process inherently accommodates all three regimes of the exploration-exploitation spectrum, enabling a principled transition from exploration to exploitation. Building on these theoretical insights, this work introduces AdaZero, an adaptive deep architecture that dynamically balances exploration and exploitation. Extensive empirical evaluations highlight AdaZero's robust performance and validate the theory's feasibility.
Acquiring three-dimensional(3D) microstructural images of cement hydration reveals critical microscale features essential for understanding hydration mechanisms and advancing material development. Micro-computed tomography(µCT) is widely used to capture these images due to its non-destructive, repeatable 3D imaging capability, despite at high operational cost. However, µCT suffers from limited resolution, weak texture, and missing phase information. In contrast, backscattered electron(BSE) provides high-resolution, phase-sensitive, and texture-rich images but are confined to 2D form. Inspired by the idea of fusing BSE style into µCT images for computationally imaging 3D BSE data, this paper proposes a cross-device generation method, termed CT2BSE, to construct 3D BSE microstructural images from µCT for cement hydration. A Voxel Swin Transformer is proposed to extract 3D μCT features by improving Video Swin Transformer with a Content-Aware Positional Encoding–3D, making encoder better suited for homogeneous images such as cement. Furthermore, a Holistic Style Injector–3D is proposed to fuse 3D μCT features with 2D BSE characteristics, thereby mitigating feature instability caused by the extreme intensity variations typically exhibited in cement images. Subsequently, the fused features are decoded into 3D volumes aligned with the BSE style while preserving original microstructure. Experimental results demonstrate that CT2BSE can generate high-fidelity 3D BSE images from μCT.
Modern neural networks can be accurate yet poorly calibrated. We present Random Geometric Calibration (RGC), a post hoc method that augments confidence estimation with a geometric signal derived from distances to training data in an aggregated semantic representation space. Prior geometric calibration methods in feature space often rely on architecture-specific layer choices, which can be unstable across models and datasets. To avoid this dependence, RGC samples a fixed number of intermediate blocks uniformly at random and reuses this set at both calibration and test time. We instantiate RGC with RGCL, which applies spatial pyramid pooling to each sampled activation, projects features to a fixed dimension, and computes a nearest-neighbor separation score that is mapped to probabilities via isotonic regression. We also present RGCC, which replaces pooling and projection with uniform coordinate sampling for improved computational efficiency. We evaluate RGCL and RGCC across multiple datasets, including CIFAR-10, CIFAR-100, and Tiny-ImageNet, and across a diverse set of architectures such as ResNet, DenseNet, and DINOv2. Our results show that RGCL achieves state-of-the-art or near state-of-the-art calibration performance, reducing expected calibration error by approximately 78% on average and by up to 97% for some models. Finally, we provide representation analysis that explains why RGCL and RGCC succeed without layer selection and characterize their computational overheads. We observe that layers exhibiting favorable geometric structure tend to emerge in later network blocks, and that randomized aggregation implicitly emphasizes these layers, leading to robust and reliable calibration performance.
In multi-turn interactions, large language models (LLMs) often exhibit a persistent influence from prior turns, even after an explicit topic switch. This behavior, which we term semantic inertia, can cause responses to deviate from the expected output distribution for an independent task, undermining task isolation and reliability. This paper introduces a rigorous experimental framework to systematically characterize the nature, form, and dynamics of semantic inertia. We propose an operational definition and a causal-contrastive method that isolates semantic carryover from confounding factors like context length. Through a series of experiments on five leading LLMs, we (i) confirm the existence of semantic inertia and identify its boundary conditions; (ii) model its decay over the course of generation, revealing a characteristic timescale and a heavy-tailed distribution; (iii) decompose its effects on three distinct channels—factual accuracy, structural integrity, and stylistic expression; and (iv) probe its controllability using prompt-based interventions. Our key findings show that inertia is not a simple length effect but an intrinsic dynamic, strongest in the initial part of a generation and decaying over a timescale of approximately 100-200 tokens. Its impact is most pronounced as a stylistic residue, while its effect on factual correctness is weaker and highly dependent on the task and domain switch. Crucially, we find that prompt-level ``reset'' instructions are unreliable and often counter-productive, while conflicting constraints consistently amplify, rather than resolve, output deviation. These results suggest that governing semantic inertia requires system-level state management mechanisms rather than relying on prompt engineering alone.
Real-world deployment of large language models (LLMs) necessitates a robust and low-latency approach to detect prompt injections; existing low-overhead methods fail to simultaneously boost robustness and reduce latency. Current defenses for prompt injection either rely on brittle heuristics or invoke costly auxiliary models, imposing a significant runtime burden. We introduce BERM, a lightweight framework that performs in-situ detection by modeling a host LLM’s internal representations extracted during prefill, adding negligible overhead. Our approach trains a lightweight classifier atop the LLM by learning a compact manifold of benign representations via joint contrastive learning to maximize the separation from malicious representations. At inference, this pre-trained classifier enables in-situ detection without invoking auxiliary guard models. On a diverse landscape of prompt injection attacks, our framework establishes a new state-of-the-art, achieving an F1-score 5.2 percentage points (pp) higher than the best prior work. Critically, BERM achieves this while being over 12x faster, reducing inference overhead to near-zero.
The Tournament Fixing Problem (TFP) asks whether a knockout tournament can be scheduled to guarantee that a given player v* wins. Although TFP is NP-hard in general, it is known to be fixed-parameter tractable (FPT) when parameterized by the feedback arc/vertex set number, or the in/out-degree of v*. However, it remained open whether TFP is FPT with respect to the subset FAS number of v* --- the minimum number of arcs intersecting all cycles containing v* --- a parameter that is never larger than the aforementioned ones. In this paper, we resolve this question negatively by proving that TFP stays NP-hard even when the subset FAS number of v* is constant ≥ 1 and either the subgraph induced by the in-neighbors D[N_{in}(v*)] or the out-neighbors D[N_{out}(v*)] is acyclic. Conversely, when both D[N_{in}(v*)] and D[N_{out}(v*)] are acyclic, we show that TFP becomes FPT parameterized by the subset FAS number of v*. Furthermore, we provide sufficient conditions under which v* can win even when this parameter is unbounded.