Authors: Zachary T. Rewolinski, Bin Yu
Abstract: Data science is a pillar of artificial intelligence (AI), which is transforming nearly every domain of human activity, from the social and physical sciences to engineering and medicine. While data-driven findings in AI offer unprecedented power to extract insights and guide decision-making, many are difficult or impossible to replicate. A key reason for this challenge is the uncertainty introduced by the many choices made throughout the data science life cycle (DSLC). Traditional statistical frameworks often fail to account for this uncertainty. The Predictability-Computability-Stability (PCS) framework for veridical (truthful) data science offers a principled approach to addressing this challenge throughout the DSLC. This paper presents an updated and streamlined PCS workflow, tailored for practitioners and enhanced with guided use of generative AI. We include a running example to display the PCS framework in action, and conduct a related case study which showcases the uncertainty in downstream predictions caused by judgment calls in the data cleaning stage.
Authors: Ziyang Zhang, Feifan Zhang, Weidong Tang, Lei Shi, Tailai Chen
Abstract: Nonlinear partial differential equations (PDEs) are pivotal in modeling complex physical systems, yet traditional Physics-Informed Neural Networks (PINNs) often struggle with unresolved residuals in critical spatiotemporal regions and violations of temporal causality. To address these limitations, we propose a novel Residual Guided Training strategy for Physics-Informed Transformer via Generative Adversarial Networks (GAN). Our framework integrates a decoder-only Transformer to inherently capture temporal correlations through autoregressive processing, coupled with a residual-aware GAN that dynamically identifies and prioritizes high-residual regions. By introducing a causal penalty term and an adaptive sampling mechanism, the method enforces temporal causality while refining accuracy in problematic domains. Extensive numerical experiments on the Allen-Cahn, Klein-Gordon, and Navier-Stokes equations demonstrate significant improvements, achieving relative MSE reductions of up to three orders of magnitude compared to baseline methods. This work bridges the gap between deep learning and physics-driven modeling, offering a robust solution for multiscale and time-dependent PDE systems.
Authors: Christina Butsko, Kristof Van Tricht, Gabriel Tseng, Giorgia Milli, David Rolnick, Ruben Cartuyvels, Inbal Becker Reshef, Zoltan Szantoi, Hannah Kerner
Abstract: The increasing availability of geospatial foundation models has the potential to transform remote sensing applications such as land cover classification, environmental monitoring, and change detection. Despite promising benchmark results, the deployment of these models in operational settings is challenging and rare. Standardized evaluation tasks often fail to capture real-world complexities relevant for end-user adoption such as data heterogeneity, resource constraints, and application-specific requirements. This paper presents a structured approach to integrate geospatial foundation models into operational mapping systems. Our protocol has three key steps: defining application requirements, adapting the model to domain-specific data and conducting rigorous empirical testing. Using the Presto model in a case study for crop mapping, we demonstrate that fine-tuning a pre-trained model significantly improves performance over conventional supervised methods. Our results highlight the model's strong spatial and temporal generalization capabilities. Our protocol provides a replicable blueprint for practitioners and lays the groundwork for future research to operationalize foundation models in diverse remote sensing applications. Application of the protocol to the WorldCereal global crop-mapping system showcases the framework's scalability.
Authors: Dmitriy Kashitsyn, Dmitriy Shabanov
Abstract: The article explores an encoding and structural information processing approach using sparse bit vectors and fixed-length linear vectors. The following are presented: a discrete method of speculative stochastic dimensionality reduction of multidimensional code and linear spaces with linear asymptotic complexity; a geometric method for obtaining discrete embeddings of an organised code space that reflect the internal structure of a given modality. The structure and properties of a code space are investigated using three modalities as examples: morphology of Russian and English languages, and immunohistochemical markers. Parallels are drawn between the resulting map of the code space layout and so-called pinwheels appearing on the mammalian neocortex. A cautious assumption is made about similarities between neocortex organisation and processes happening in our models.
Authors: Bakhtiyar Mammadli, Casim Yazici, Muhammed G\"urb\"uz, \.Irfan Kocaman, F. Javier Dominguez-Gutierrez, Fatih Mehmet \"Ozkal
Abstract: In this study, we present a machine learning (ML) framework to predict the axial load-bearing capacity, (kN), of cold-formed steel structural members. The methodology emphasizes robust model selection and interpretability, addressing the limitations of traditional analytical approaches in capturing the nonlinearities and geometrical complexities inherent to buckling behavior. The dataset, comprising key geometric and mechanical parameters of steel columns, was curated with appropriate pre-processing steps including removal of non-informative identifiers and imputation of missing values. A comprehensive suite of regression algorithms, ranging from linear models to kernel-based regressors and ensemble tree methods was evaluated. Among these, Gradient Boosting Regression exhibited superior predictive performance across multiple metrics, including the coefficient of determination (R2), root mean squared error (RMSE), and mean absolute error (MAE), and was consequently selected as the final model. Model interpretability was addressed using SHapley Additive exPlanations (SHAP), enabling insight into the relative importance and interaction of input features influencing the predicted axial capacity. To facilitate practical deployment, the model was integrated into an interactive, Python-based web interface via Streamlit. This tool allows end-users-such as structural engineers and designers, to input design parameters manually or through CSV upload, and to obtain real-time predictions of axial load capacity without the need for programming expertise. Applied to the context of steel storage rack columns, the framework demonstrates how data-driven tools can enhance design safety, streamline validation workflows, and inform decision-making in structural applications where buckling is a critical failure mode
Authors: Chao Yan, Babak Mafakheri
Abstract: Satellite connectivity is gaining increased attention as the demand for seamless internet access, especially in transportation and remote areas, continues to grow. For fast-moving objects such as aircraft, vehicles, or trains, satellite connectivity is critical due to their mobility and frequent presence in areas without terrestrial coverage. Maintaining reliable connectivity in these cases requires frequent switching between satellite beams, constellations, or orbits. To enhance user experience and address challenges like long switching times, Machine Learning (ML) algorithms can analyze historical connectivity data and predict network quality at specific locations. This allows for proactive measures, such as network switching before connectivity issues arise. In this paper, we analyze a real dataset of communication between a Geostationary Orbit (GEO) satellite and aircraft over multiple flights, using ML to predict signal quality. Our prediction model achieved an F1 score of 0.97 on the test data, demonstrating the accuracy of machine learning in predicting signal quality during flight. By enabling seamless broadband service, including roaming between different satellite constellations and providers, our model addresses the need for real-time predictions of signal quality. This approach can further be adapted to automate satellite and beam-switching mechanisms to improve overall communication efficiency. The model can also be retrained and applied to any moving object with satellite connectivity, using customized datasets, including connected vehicles and trains.
Authors: Moutaz Bellah Bentrad, Adel Ghoggal, Tahar Bahi, Abderaouf Bahi
Abstract: The diagnosis of induction machines has traditionally relied on model-based methods that require the development of complex dynamic models, making them difficult to implement and computationally expensive. To overcome these limitations, this paper proposes a model-free approach using Graph Neural Networks (GNNs) for fault diagnosis in induction machines. The focus is on detecting multiple fault types -- including eccentricity, bearing defects, and broken rotor bars -- under varying severity levels and load conditions. Unlike traditional approaches, raw current and vibration signals are used as direct inputs, eliminating the need for signal preprocessing or manual feature extraction. The proposed GNN-ASE model automatically learns and extracts relevant features from raw inputs, leveraging the graph structure to capture complex relationships between signal types and fault patterns. It is evaluated for both individual fault detection and multi-class classification of combined fault conditions. Experimental results demonstrate the effectiveness of the proposed model, achieving 92.5\% accuracy for eccentricity defects, 91.2\% for bearing faults, and 93.1\% for broken rotor bar detection. These findings highlight the model's robustness and generalization capability across different operational scenarios. The proposed GNN-based framework offers a lightweight yet powerful solution that simplifies implementation while maintaining high diagnostic performance. It stands as a promising alternative to conventional model-based diagnostic techniques for real-world induction machine monitoring and predictive maintenance.
Authors: Adil Mukhtar, Michael Hadwiger, Franz Wotawa, Gerald Schweiger
Abstract: Reproducibility is a cornerstone of scientific research, enabling independent verification and validation of empirical findings. The topic gained prominence in fields such as psychology and medicine, where concerns about non - replicable results sparked ongoing discussions about research practices. In recent years, the fast-growing field of Machine Learning (ML) has become part of this discourse, as it faces similar concerns about transparency and reliability. Some reproducibility issues in ML research are shared with other fields, such as limited access to data and missing methodological details. In addition, ML introduces specific challenges, including inherent nondeterminism and computational constraints. While reproducibility issues are increasingly recognized by the ML community and its major conferences, less is known about how these challenges manifest in applied disciplines. This paper contributes to closing this gap by analyzing the transparency and reproducibility standards of ML applications in building energy systems. The results indicate that nearly all articles are not reproducible due to insufficient disclosure across key dimensions of reproducibility. 72% of the articles do not specify whether the dataset used is public, proprietary, or commercially available. Only two papers share a link to their code - one of which was broken. Two-thirds of the publications were authored exclusively by academic researchers, yet no significant differences in reproducibility were observed compared to publications with industry-affiliated authors. These findings highlight the need for targeted interventions, including reproducibility guidelines, training for researchers, and policies by journals and conferences that promote transparency and reproducibility.
Authors: Vijja Wichitwechkarn, Charles Fox, Ruchi Choudhary
Abstract: Foundation models for natural language processing have many coherent definitions of hallucination and methods for its detection and mitigation. However, analogous definitions and methods do not exist for multi-variate time-series (MVTS) foundation models. We propose new definitions for MVTS hallucination, along with new detection and mitigation methods using a diffusion model to estimate hallucination levels. We derive relational datasets from popular time-series datasets to benchmark these relational hallucination levels. Using these definitions and models, we find that open-source pre-trained MVTS imputation foundation models relationally hallucinate on average up to 59.5% as much as a weak baseline. The proposed mitigation method reduces this by up to 47.7% for these models. The definition and methods may improve adoption and safe usage of MVTS foundation models.
Authors: Zhenan Lin, Yuni Lai, Wai Lun Lo, Richard Tai-Chiu Hsung, Harris Sik-Ho Tsang, Xiaoyu Xue, Kai Zhou, Yulin Zhu
Abstract: Time-evolving traffic flow forecasting are playing a vital role in intelligent transportation systems and smart cities. However, the dynamic traffic flow forecasting is a highly nonlinear problem with complex temporal-spatial dependencies. Although the existing methods has provided great contributions to mine the temporal-spatial patterns in the complex traffic networks, they fail to encode the globally temporal-spatial patterns and are prone to overfit on the pre-defined geographical correlations, and thus hinder the model's robustness on the complex traffic environment. To tackle this issue, in this work, we proposed a multi-grained temporal-spatial graph learning framework to adaptively augment the globally temporal-spatial patterns obtained from a crafted graph transformer encoder with the local patterns from the graph convolution by a crafted gated fusion unit with residual connection techniques. Under these circumstances, our proposed model can mine the hidden global temporal-spatial relations between each monitor stations and balance the relative importance of local and global temporal-spatial patterns. Experiment results demonstrate the strong representation capability of our proposed method and our model consistently outperforms other strong baselines on various real-world traffic networks.
Authors: Etienne Buehrle, Christoph Stiller
Abstract: The optimal control problem of stochastic systems is commonly solved via robust or scenario-based optimization methods, which are both challenging to scale to long optimization horizons. We cast the optimal control problem of a stochastic system as a convex optimization problem over occupation measures. We demonstrate our method on a set of synthetic and real-world scenarios, learning cost functions from data via Christoffel polynomials. The code for our experiments is available at https://github.com/ebuehrle/dpoc.
Authors: Binrui Shen, Yuan Liang, Shengxin Zhu
Abstract: Graph matching, typically formulated as a Quadratic Assignment Problem (QAP), seeks to establish node correspondences between two graphs. To address the NP-hardness of QAP, some existing methods adopt projection-based relaxations that embed the problem into the convex hull of the discrete domain. However, these relaxations inevitably enlarge the feasible set, introducing two sources of error: numerical scale sensitivity and geometric misalignment between the relaxed and original domains. To alleviate these errors, we propose a novel relaxation framework by reformulating the projection step as a Frobenius-regularized Linear Assignment (FRA) problem, where a tunable regularization term mitigates feasible region inflation. This formulation enables normalization-based operations to preserve numerical scale invariance without compromising accuracy. To efficiently solve FRA, we propose the Scaling Doubly Stochastic Normalization (SDSN) algorithm. Building on its favorable computational properties, we develop a theoretically grounded mixed-precision architecture to achieve substantial acceleration. Comprehensive CPU-based benchmarks demonstrate that FRAM consistently outperforms all baseline methods under identical precision settings. When combined with a GPU-based mixed-precision architecture, FRAM achieves up to 370X speedup over its CPU-FP64 counterpart, with negligible loss in solution accuracy.
Authors: Amir Hossein Kalantari, Eleonora Papadimitriou, Amir Pooyan Afghari
Abstract: Naturalistic driving studies offer a powerful means for observing and quantifying real-world driving behaviour. One of their prominent applications in traffic safety is the continuous monitoring and classification of risky driving behaviour. However, many existing frameworks rely on fixed time windows and static thresholds for distinguishing between safe and risky behaviour - limiting their ability to respond to the stochastic nature of real-world driving. This study proposes a dynamic and individualised framework for identifying risky driving behaviour using Belgian naturalistic driving data. The approach leverages a rolling time window and bi-level optimisation to dynamically calibrate both risk thresholds and model hyperparameters, capturing subtle behavioural shifts. Two safety indicators, speed-weighted headway and harsh driving events, were evaluated using three data-driven models: Random Forest, XGBoost, and Deep Neural Network (DNN). The DNN demonstrated strong capability in capturing subtle changes in driving behaviour, particularly excelling in high-recall tasks, making it promising for early-stage risk detection. XGBoost provided the most balanced and stable performance across different thresholds and evaluation metrics. While random forest showed more variability, it responded sensitively to dynamic threshold adjustments, which may be advantageous during model adaptation or tuning. Speed-weighted headway emerged as a more stable and context-sensitive risk indicator than harsh driving events, likely due to its robustness to label sparsity and contextual variation. Overall, the findings support the value of adaptive, personalised risk detection approaches for enhancing real-time safety feedback and tailoring driver support in intelligent transport systems.
Authors: Julien Simon de Kergunic (CRIStAL), Rony Abecidan (CRIStAL), Patrick Bas (CRIStAL), Vincent Itier (IMT Nord Europe, CRIStAL)
Abstract: Despite recent progress in splicing detection, deep learning-based forensic tools remain difficult to deploy in practice due to their high sensitivity to training conditions. Even mild post-processing applied to evaluation images can significantly degrade detector performance, raising concerns about their reliability in operational contexts. In this work, we show that the same deep architecture can react very differently to unseen post-processing depending on the learned weights, despite achieving similar accuracy on in-distribution test data. This variability stems from differences in the latent spaces induced by training, which affect how samples are separated internally. Our experiments reveal a strong correlation between the distribution of latent margins and a detector's ability to generalize to post-processed images. Based on this observation, we propose a practical strategy for building more robust detectors: train several variants of the same model under different conditions, and select the one that maximizes latent margins.
Authors: Ruichen Xu, Kexin Chen
Abstract: Modern large language models excel in knowledge-intensive tasks, yet how transformers acquire (store) knowledge during pre-training and extract (retrieve) it during post-fine-tuning inference remains theoretically opaque. While prior theoretical work has begun to investigate these questions through the analysis of training dynamics, such studies are limited to single-layer, attention-only architectures. However, most existing studies suggest that MLPs are the most contributing components for storing knowledge in transformer-based language models. Meanwhile, our empirical investigations reveal that such simplified models, when trained using standard next-token prediction objectives, may be incapable of acquiring or extracting factual knowledge. To overcome this limitation, we introduce a tractable one-layer transformer framework that crucially incorporates both self-attention and MLP modules. By tracking its gradient dynamics, we establish convergence and generalization guarantees that illuminate the ability of knowledge acquisition and extraction. We prove that 1) Transformers can achieve near-optimal training loss during pre-training, signifying effective knowledge acquisition; 2) With a large fine-tuning dataset and specific data multiplicity conditions met, transformers can achieve low generalization error when tested on factual knowledge learned during pre-training but not reinforced during the fine-tuning, indicating successful knowledge extraction; 3) When the conditions are not satisfied, transformers exhibit high generalization loss, resulting in hallucinations. Our analysis includes both full fine-tuning and low-rank fine-tuning. Furthermore, our analysis offers theoretical insights into several pertinent empirical phenomena, such as the role of learning rate schedules. Experiments on synthetic and real-world PopQA datasets with GPT-2 and Llama-3.2-1B validate our results.
Authors: Advey Nandan, Cheng-Ting Chou, Amrit Kurakula, Cole Blondin, Kevin Zhu, Vasu Sharma, Sean O'Brien
Abstract: We investigate the phenomenon of neuron universality in independently trained GPT-2 Small models, examining how these universal neurons-neurons with consistently correlated activations across models-emerge and evolve throughout training. By analyzing five GPT-2 models at three checkpoints (100k, 200k, 300k steps), we identify universal neurons through pairwise correlation analysis of activations over a dataset of 5 million tokens. Ablation experiments reveal significant functional impacts of universal neurons on model predictions, measured via loss and KL divergence. Additionally, we quantify neuron persistence, demonstrating high stability of universal neurons across training checkpoints, particularly in deeper layers. These findings suggest stable and universal representational structures emerge during neural network training.
Authors: Aitor S\'anchez-Ferrera, Usue Mori, Borja Calvo, Jose A. Lozano
Abstract: Time series anomaly detection plays a critical role in a wide range of real-world applications. Among unsupervised approaches, self-supervised learning has gained traction for modeling normal behavior without the need of labeled data. However, many existing methods rely on a single proxy task, limiting their ability to capture meaningful patterns in normal data. Moreover, they often depend on handcrafted transformations tailored specific domains, hindering their generalization accross diverse problems. To address these limitations, we introduce NeuCoReClass AD, a self-supervised multi-task time series anomaly detection framework that combines contrastive, reconstruction, and classification proxy tasks. Our method employs neural transformation learning to generate augmented views that are informative, diverse, and coherent, without requiring domain-specific knowledge. We evaluate NeuCoReClass AD across a wide range of benchmarks, demonstrating that it consistently outperforms both classical baselines and most deep-learning alternatives. Furthermore, it enables the characterization of distinct anomaly profiles in a fully unsupervised manner.
Authors: Ziyao Wang, Guoheng Sun, Yexiao He, Zheyu Shen, Bowei Tian, Ang Li
Abstract: Commercial LLM services often conceal internal reasoning traces while still charging users for every generated token, including those from hidden intermediate steps, raising concerns of token inflation and potential overbilling. This gap underscores the urgent need for reliable token auditing, yet achieving it is far from straightforward: cryptographic verification (e.g., hash-based signature) offers little assurance when providers control the entire execution pipeline, while user-side prediction struggles with the inherent variance of reasoning LLMs, where token usage fluctuates across domains and prompt styles. To bridge this gap, we present PALACE (Predictive Auditing of LLM APIs via Reasoning Token Count Estimation), a user-side framework that estimates hidden reasoning token counts from prompt-answer pairs without access to internal traces. PALACE introduces a GRPO-augmented adaptation module with a lightweight domain router, enabling dynamic calibration across diverse reasoning tasks and mitigating variance in token usage patterns. Experiments on math, coding, medical, and general reasoning benchmarks show that PALACE achieves low relative error and strong prediction accuracy, supporting both fine-grained cost auditing and inflation detection. Taken together, PALACE represents an important first step toward standardized predictive auditing, offering a practical path to greater transparency, accountability, and user trust.
Authors: Khaled Eskaf
Abstract: SmartDate is an AI-powered system for automated sorting and quality control of date fruits. It combines deep learning, genetic algorithms, and reinforcement learning to improve classification accuracy and predict shelf life. The system uses high-resolution imaging and Visible-Near-Infrared (VisNIR) spectral sensors to evaluate key features such as moisture, sugar content, and texture. Reinforcement learning enables real-time adaptation to production conditions, while genetic algorithms optimize model parameters. SmartDate achieved 94.5 percent accuracy, 93.1 percent F1-score, and an AUC-ROC of 0.96. The system reduces waste and ensures that only high-quality dates reach the market, setting a new benchmark in smart agriculture.
Authors: Jinsoo Bae, Seoung Bum Kim, Hyungrok Do
Abstract: Semi-supervised learning (SSL) uses unlabeled data to improve the performance of machine learning models when labeled data is scarce. However, its real-world applications often face the label distribution mismatch problem, in which the unlabeled dataset includes instances whose ground-truth labels are absent from the labeled training dataset. Recent studies, referred to as safe SSL, have addressed this issue by using both classification and out-of-distribution (OOD) detection. However, the existing methods may suffer from overconfidence in deep neural networks, leading to increased SSL errors because of high confidence in incorrect pseudo-labels or OOD detection. To address this, we propose a novel method, CaliMatch, which calibrates both the classifier and the OOD detector to foster safe SSL. CaliMatch presents adaptive label smoothing and temperature scaling, which eliminates the need to manually tune the smoothing degree for effective calibration. We give a theoretical justification for why improving the calibration of both the classifier and the OOD detector is crucial in safe SSL. Extensive evaluations on CIFAR-10, CIFAR-100, SVHN, TinyImageNet, and ImageNet demonstrate that CaliMatch outperforms the existing methods in safe SSL tasks.
Authors: Jiazhen Pan (Cherise), Bailiang Jian (Cherise), Paul Hager (Cherise), Yundi Zhang (Cherise), Che Liu (Cherise), Friedrike Jungmann (Cherise), Hongwei Bran Li (Cherise), Chenyu You (Cherise), Junde Wu (Cherise), Jiayuan Zhu (Cherise), Fenglin Liu (Cherise), Yuyuan Liu (Cherise), Niklas Bubeck (Cherise), Christian Wachinger (Cherise), Chen (Cherise), Chen (Cherise), Zhenyu Gong, Cheng Ouyang, Georgios Kaissis, Benedikt Wiestler, Daniel Rueckert
Abstract: Ensuring the safety and reliability of large language models (LLMs) in clinical practice is critical to prevent patient harm and promote trustworthy healthcare applications of AI. However, LLMs are advancing so rapidly that static safety benchmarks often become obsolete upon publication, yielding only an incomplete and sometimes misleading picture of model trustworthiness. We demonstrate that a Dynamic, Automatic, and Systematic (DAS) red-teaming framework that continuously stress-tests LLMs can reveal significant weaknesses of current LLMs across four safety-critical domains: robustness, privacy, bias/fairness, and hallucination. A suite of adversarial agents is applied to autonomously mutate test cases, identify/evolve unsafe-triggering strategies, and evaluate responses, uncovering vulnerabilities in real time without human intervention. Applying DAS to 15 proprietary and open-source LLMs revealed a stark contrast between static benchmark performance and vulnerability under adversarial pressure. Despite a median MedQA accuracy exceeding 80\%, 94\% of previously correct answers failed our dynamic robustness tests. We observed similarly high failure rates across other domains: privacy leaks were elicited in 86\% of scenarios, cognitive-bias priming altered clinical recommendations in 81\% of fairness tests, and we identified hallucination rates exceeding 66\% in widely used models. Such profound residual risks are incompatible with routine clinical practice. By converting red-teaming from a static checklist into a dynamic stress-test audit, DAS red-teaming offers the surveillance that hospitals/regulators/technology vendors require as LLMs become embedded in patient chatbots, decision-support dashboards, and broader healthcare workflows. Our framework delivers an evolvable, scalable, and reliable safeguard for the next generation of medical AI.
Authors: Feng Xu, Hui Wang, Yuting Huang, Danwei Zhang, Zizhu Fan
Abstract: Modeling temporal multimodal data poses significant challenges in classification tasks, particularly in capturing long-range temporal dependencies and intricate cross-modal interactions. Audiovisual data, as a representative example, is inherently characterized by strict temporal order and diverse modalities. Effectively leveraging the temporal structure is essential for understanding both intra-modal dynamics and inter-modal correlations. However, most existing approaches treat each modality independently and rely on shallow fusion strategies, which overlook temporal dependencies and hinder the model's ability to represent complex structural relationships. To address the limitation, we propose the hybrid hypergraph network (HHN), a novel framework that models temporal multimodal data via a segmentation-first, graph-later strategy. HHN splits sequences into timestamped segments as nodes in a heterogeneous graph. Intra-modal structures are captured via hyperedges guided by a maximum entropy difference criterion, enhancing node heterogeneity and structural discrimination, followed by hypergraph convolution to extract high-order dependencies. Inter-modal links are established through temporal alignment and graph attention for semantic fusion. HHN achieves state-of-the-art (SOTA) results on four multimodal datasets, demonstrating its effectiveness in complex classification tasks.
Authors: M. Ontivero-Ortega, A. Fania, A. Lacalamita, R. Bellotti, A. Monaco, S. Stramaglia
Abstract: Leveraging recent advances in the analysis of synergy and redundancy in systems of random variables, an adaptive version of the widely used metric Leave One Covariate Out (LOCO) has been recently proposed to quantify cooperative effects in feature importance (Hi-Fi), a key technique in explainable artificial intelligence (XAI), so as to disentangle high-order effects involving a particular input feature in regression problems. Differently from standard feature importance tools, where a single score measures the relevance of each feature, each feature is here characterized by three scores, a two-body (unique) score and higher-order scores (redundant and synergistic). This paper presents a framework to assign those three scores (unique, redundant, and synergistic) to each individual pattern of the data set, while comparing it with the well-known measure of feature importance named {\it Shapley effect}. To illustrate the potential of the proposed framework, we focus on a One-Health application: the relation between air pollutants and Alzheimer's disease mortality rate. Our main result is the synergistic association between features related to $O_3$ and $NO_2$ with mortality, especially in the provinces of Bergamo e Brescia; notably also the density of urban green areas displays synergistic influence with pollutants for the prediction of AD mortality. Our results place local Hi-Fi as a promising tool of wide applicability, which opens new perspectives for XAI as well as to analyze high-order relationships in complex systems.
Authors: Hanchen Yang, Jiaqi Wang, Jiannong Cao, Wengen Li, Jialun Zheng, Yangning Li, Chunyu Miao, Jihong Guan, Shuigeng Zhou, Philip S. Yu
Abstract: Sea surface temperature (SST) prediction is a critical task in ocean science, supporting various applications, such as weather forecasting, fisheries management, and storm tracking. While existing data-driven methods have demonstrated significant success, they often neglect to leverage the rich domain knowledge accumulated over the past decades, limiting further advancements in prediction accuracy. The recent emergence of large language models (LLMs) has highlighted the potential of integrating domain knowledge for downstream tasks. However, the application of LLMs to SST prediction remains underexplored, primarily due to the challenge of integrating ocean domain knowledge and numerical data. To address this issue, we propose Ocean Knowledge Graph-enhanced LLM (OKG-LLM), a novel framework for global SST prediction. To the best of our knowledge, this work presents the first systematic effort to construct an Ocean Knowledge Graph (OKG) specifically designed to represent diverse ocean knowledge for SST prediction. We then develop a graph embedding network to learn the comprehensive semantic and structural knowledge within the OKG, capturing both the unique characteristics of individual sea regions and the complex correlations between them. Finally, we align and fuse the learned knowledge with fine-grained numerical SST data and leverage a pre-trained LLM to model SST patterns for accurate prediction. Extensive experiments on the real-world dataset demonstrate that OKG-LLM consistently outperforms state-of-the-art methods, showcasing its effectiveness, robustness, and potential to advance SST prediction. The codes are available in the online repository.
Authors: Andy Hu, Devika Prasad, Luiz Pizzato, Nicholas Foord, Arman Abrahamyan, Anna Leontjeva, Cooper Doyle, Dan Jermyn
Abstract: In machine learning, the process of feature selection involves finding a reduced subset of features that captures most of the information required to train an accurate and efficient model. This work presents FeatureCuts, a novel feature selection algorithm that adaptively selects the optimal feature cutoff after performing filter ranking. Evaluated on 14 publicly available datasets and one industry dataset, FeatureCuts achieved, on average, 15 percentage points more feature reduction and up to 99.6% less computation time while maintaining model performance, compared to existing state-of-the-art methods. When the selected features are used in a wrapper method such as Particle Swarm Optimization (PSO), it enables 25 percentage points more feature reduction, requires 66% less computation time, and maintains model performance when compared to PSO alone. The minimal overhead of FeatureCuts makes it scalable for large datasets typically seen in enterprise applications.
Authors: Yeong-Joon Ju, Seong-Whan Lee
Abstract: Multimodal Large Language Models (MLLMs) have emerged as a promising solution for universal embedding tasks, yet adapting their generative nature for discriminative representation learning remains a significant challenge. The dominant paradigm of large-scale contrastive pre-training suffers from critical inefficiencies, including prohibitive computational costs and a failure to leverage the intrinsic, instruction-following capabilities of MLLMs. To overcome these limitations, we propose an efficient framework for universal multimodal embeddings, which bridges this gap by centering on two synergistic components. First, our hierarchical embedding prompt template employs a two-level instruction architecture that forces the model to produce discriminative representations. Building on this strong foundation, our second component, self-aware hard negative sampling, redefines the fine-tuning process by leveraging the model's own understanding to efficiently mine challenging negatives while actively filtering out potential false negatives. Our comprehensive experiments show that our hierarchical prompt achieves zero-shot performance competitive with contrastively trained baselines and enhances the fine-tuning process by lifting a simple in-batch negative baseline by 4.8 points on the MMEB benchmark. We further boost the performance via our self-aware hard negative sampling, achieving the state-of-the-art performance without the contrative pre-training. Our work presents an effective and efficient pathway to adapt MLLMs for universal embedding tasks, significantly reducing training time.
Authors: Chuan He, Yang Chen, Wuliang Huang, Tianyi Zheng, Jianhu Chen, Bin Dou, Yice Luo, Yun Zhu, Baokun Wang, Yongchao Liu, Xing Fu, Yu Cheng, Chuntao Hong, Weiqiang Wang, Xin-Wei Yao
Abstract: Multi-source user representation learning plays a critical role in enabling personalized services on web platforms (e.g., Alipay). While prior works have adopted late-fusion strategies to combine heterogeneous data sources, they suffer from three key limitations: lack of unified representation frameworks, scalability and storage issues in data compression, and inflexible cross-task generalization. To address these challenges, we propose U^2QT (Unified User Quantized Tokenizers), a novel framework that integrates cross-domain knowledge transfer with early fusion of heterogeneous domains. Our framework employs a two-stage architecture: first, a causal Q-Former projects domain-specific features into a shared causal representation space to preserve inter-modality dependencies; second, a multi-view RQ-VAE discretizes causal embeddings into compact tokens through shared and source-specific codebooks, enabling efficient storage while maintaining semantic coherence. Experimental results showcase U^2QT's advantages across diverse downstream tasks, outperforming task-specific baselines in future behavior prediction and recommendation tasks while achieving efficiency gains in storage and computation. The unified tokenization framework enables seamless integration with language models and supports industrial-scale applications.
Authors: Amrit Rajeev, Udayaadithya Avadhanam, Harshula Tulapurkar, SaiBarath Sundar
Abstract: Zero-shot text classification remains a difficult task in domains with evolving knowledge and ambiguous category boundaries, such as ticketing systems. Large language models (LLMs) often struggle to generalize in these scenarios due to limited topic separability, while few-shot methods are constrained by insufficient data diversity. We propose a classification framework that combines iterative topic refinement, contrastive prompting, and active learning. Starting with a small set of labeled samples, the model generates initial topic labels. Misclassified or ambiguous samples are then used in an iterative contrastive prompting process to refine category distinctions by explicitly teaching the model to differentiate between closely related classes. The framework features a human-in-the-loop component, allowing users to introduce or revise category definitions in natural language. This enables seamless integration of new, unseen categories without retraining, making the system well-suited for real-world, dynamic environments. The evaluations on AGNews and DBpedia demonstrate strong performance: 91% accuracy on AGNews (3 seen, 1 unseen class) and 84% on DBpedia (8 seen, 1 unseen), with minimal accuracy shift after introducing unseen classes (82% and 87%, respectively). The results highlight the effectiveness of prompt-based semantic reasoning for fine-grained classification with limited supervision.
Authors: Rub\'en Mu\~noz-Sierra, Manuel Doblar\'e, Jacobo Ayensa-Jim\'enez
Abstract: Physically Guided Neural Networks with Internal Variables are SciML tools that use only observable data for training and and have the capacity to unravel internal state relations. They incorporate physical knowledge both by prescribing the model architecture and using loss regularization, thus endowing certain specific neurons with a physical meaning as internal state variables. Despite their potential, these models face challenges in scalability when applied to high-dimensional data such as fine-grid spatial fields or time-evolving systems. In this work, we propose some enhancements to the PGNNIV framework that address these scalability limitations through reduced-order modeling techniques. Specifically, we introduce alternatives to the original decoder structure using spectral decomposition, POD, and pretrained autoencoder-based mappings. These surrogate decoders offer varying trade-offs between computational efficiency, accuracy, noise tolerance, and generalization, while improving drastically the scalability. Additionally, we integrate model reuse via transfer learning and fine-tuning strategies to exploit previously acquired knowledge, supporting efficient adaptation to novel materials or configurations, and significantly reducing training time while maintaining or improving model performance. To illustrate these various techniques, we use a representative case governed by the nonlinear diffusion equation, using only observable data. Results demonstrate that the enhanced PGNNIV framework successfully identifies the underlying constitutive state equations while maintaining high predictive accuracy. It also improves robustness to noise, mitigates overfitting, and reduces computational demands. The proposed techniques can be tailored to various scenarios depending on data availability, resources, and specific modeling objectives, overcoming scalability challenges in all the scenarios.
Authors: Sudip K. Seal, Maksudul Alam, Jorge Ramirez, Sajal Dash, Hao Lu
Abstract: Energy efficiency of training and inferencing with large neural network models is a critical challenge facing the future of sustainable large-scale machine learning workloads. This paper introduces an alternative strategy, called phantom parallelism, to minimize the net energy consumption of traditional tensor (model) parallelism, the most energy-inefficient component of large neural network training. The approach is presented in the context of feed-forward network architectures as a preliminary, but comprehensive, proof-of-principle study of the proposed methodology. We derive new forward and backward propagation operators for phantom parallelism, implement them as custom autograd operations within an end-to-end phantom parallel training pipeline and compare its parallel performance and energy-efficiency against those of conventional tensor parallel training pipelines. Formal analyses that predict lower bandwidth and FLOP counts are presented with supporting empirical results on up to 256 GPUs that corroborate these gains. Experiments are shown to deliver ~50% reduction in the energy consumed to train FFNs using the proposed phantom parallel approach when compared with conventional tensor parallel methods. Additionally, the proposed approach is shown to train smaller phantom models to the same model loss on smaller GPU counts as larger tensor parallel models on larger GPU counts offering the possibility for even greater energy savings.
Authors: Xiang Li, Penglei Sun, Wanyun Zhou, Zikai Wei, Yongqi Zhang, Xiaowen Chu
Abstract: Individual investors are significantly outnumbered and disadvantaged in financial markets, overwhelmed by abundant information and lacking professional analysis. Equity research reports stand out as crucial resources, offering valuable insights. By leveraging these reports, large language models (LLMs) can enhance investors' decision-making capabilities and strengthen financial analysis. However, two key challenges limit their effectiveness: (1) the rapid evolution of market events often outpaces the slow update cycles of existing knowledge bases, (2) the long-form and unstructured nature of financial reports further hinders timely and context-aware integration by LLMs. To address these challenges, we tackle both data and methodological aspects. First, we introduce the Event-Enhanced Automated Construction of Financial Knowledge Graph (FinKario), a dataset comprising over 305,360 entities, 9,625 relational triples, and 19 distinct relation types. FinKario automatically integrates real-time company fundamentals and market events through prompt-driven extraction guided by professional institutional templates, providing structured and accessible financial insights for LLMs. Additionally, we propose a Two-Stage, Graph-Based retrieval strategy (FinKario-RAG), optimizing the retrieval of evolving, large-scale financial knowledge to ensure efficient and precise data access. Extensive experiments show that FinKario with FinKario-RAG achieves superior stock trend prediction accuracy, outperforming financial LLMs by 18.81% and institutional strategies by 17.85% on average in backtesting.
Authors: Timothy Oladunni, Alex Wong
Abstract: This study proposes a novel perspective on multimodal deep learning for biomedical signal classification, systematically analyzing how complementary feature domains impact model performance. While fusing multiple domains often presumes enhanced accuracy, this work demonstrates that adding modalities can yield diminishing returns, as not all fusions are inherently advantageous. To validate this, five deep learning models were designed, developed, and rigorously evaluated: three unimodal (1D-CNN for time, 2D-CNN for time-frequency, and 1D-CNN-Transformer for frequency) and two multimodal (Hybrid 1, which fuses 1D-CNN and 2D-CNN; Hybrid 2, which combines 1D-CNN, 2D-CNN, and a Transformer). For ECG classification, bootstrapping and Bayesian inference revealed that Hybrid 1 consistently outperformed the 2D-CNN baseline across all metrics (p-values < 0.05, Bayesian probabilities > 0.90), confirming the synergistic complementarity of the time and time-frequency domains. Conversely, Hybrid 2's inclusion of the frequency domain offered no further improvement and sometimes a marginal decline, indicating representational redundancy; a phenomenon further substantiated by a targeted ablation study. This research redefines a fundamental principle of multimodal design in biomedical signal analysis. We demonstrate that optimal domain fusion isn't about the number of modalities, but the quality of their inherent complementarity. This paradigm-shifting concept moves beyond purely heuristic feature selection. Our novel theoretical contribution, "Complementary Feature Domains in Multimodal ECG Deep Learning," presents a mathematically quantifiable framework for identifying ideal domain combinations, demonstrating that optimal multimodal performance arises from the intrinsic information-theoretic complementarity among fused domains.
Authors: Roie Kazoom, Ofir Cohen, Rami Puzis, Asaf Shabtai, Ofer Hadar
Abstract: We introduce VAULT, a fully automated adversarial RAG pipeline that systematically uncovers and remedies weaknesses in NLI models through three stages: retrieval, adversarial generation, and iterative retraining. First, we perform balanced few-shot retrieval by embedding premises with both semantic (BGE) and lexical (BM25) similarity. Next, we assemble these contexts into LLM prompts to generate adversarial hypotheses, which are then validated by an LLM ensemble for label fidelity. Finally, the validated adversarial examples are injected back into the training set at increasing mixing ratios, progressively fortifying a zero-shot RoBERTa-base model.On standard benchmarks, VAULT elevates RoBERTa-base accuracy from 88.48% to 92.60% on SNLI +4.12%, from 75.04% to 80.95% on ANLI +5.91%, and from 54.67% to 71.99% on MultiNLI +17.32%. It also consistently outperforms prior in-context adversarial methods by up to 2.0% across datasets. By automating high-quality adversarial data curation at scale, VAULT enables rapid, human-independent robustness improvements in NLI inference tasks.
Authors: Lucas Robinet, Ahmad Berjaoui, Elizabeth Cohen-Jonathan Moyal
Abstract: Self-supervised learning has driven major advances in computational pathology by enabling models to learn rich representations from hematoxylin and eosin (H&E)-stained cancer tissue. However, histopathology alone often falls short for molecular characterization and understanding clinical outcomes, as important information is contained in high-dimensional omics profiles like transcriptomics, methylomics, or genomics. In this work, we introduce MORPHEUS, a unified transformer-based pre-training framework that encodes both histopathology and multi-omics data into a shared latent space. At its core, MORPHEUS relies on a masked modeling objective applied to randomly selected omics portions, encouraging the model to learn biologically meaningful cross-modal relationships. The same pre-trained network can be applied to histopathology alone or in combination with any subset of omics modalities, seamlessly adapting to the available inputs. Additionally, MORPHEUS enables any-to-any omics generation, enabling one or more omics profiles to be inferred from any subset of modalities, including H&E alone. Pre-trained on a large pan-cancer cohort, MORPHEUS consistently outperforms state-of-the-art methods across diverse modality combinations and tasks, positioning itself as a promising framework for developing multimodal foundation models in oncology. The code is available at: https://github.com/Lucas-rbnt/MORPHEUS
Authors: Agrim Bari, Parikshit Hegde, Gustavo de Veciana
Abstract: With the growing use of Large Language Model (LLM)-based tools like ChatGPT, Perplexity, and Gemini across industries, there is a rising need for efficient LLM inference systems. These systems handle requests with a unique two-phase computation structure: a prefill-phase that processes the full input prompt and a decode-phase that autoregressively generates tokens one at a time. This structure calls for new strategies for routing and scheduling requests. In this paper, we take a comprehensive approach to this challenge by developing a theoretical framework that models routing and scheduling in LLM inference systems. We identify two key design principles-optimal tiling and dynamic resource allocation-that are essential for achieving high throughput. Guided by these principles, we propose the Resource-Aware Dynamic (RAD) scheduler and prove that it achieves throughput optimality under mild conditions. To address practical Service Level Objectives (SLOs) such as serving requests with different Time Between Token (TBT) constraints, we design the SLO-Aware LLM Inference (SLAI) scheduler. SLAI uses real-time measurements to prioritize decode requests that are close to missing their TBT deadlines and reorders prefill requests based on known prompt lengths to further reduce the Time To First Token (TTFT) delays. We evaluate SLAI on the Openchat ShareGPT4 dataset using the Mistral-7B model on an NVIDIA RTX ADA 6000 GPU. Compared to Sarathi-Serve, SLAI reduces the median TTFT by 53% and increases the maximum serving capacity by 26% such that median TTFT is below 0.5 seconds, while meeting tail TBT latency constraints.
Authors: Gnankan Landry Regis N'guessan
Abstract: Conventional deep learning models embed data in Euclidean space $\mathbb{R}^d$, a poor fit for strictly hierarchical objects such as taxa, word senses, or file systems. We introduce van der Put Neural Networks (v-PuNNs), the first architecture whose neurons are characteristic functions of p-adic balls in $\mathbb{Z}_p$. Under our Transparent Ultrametric Representation Learning (TURL) principle every weight is itself a p-adic number, giving exact subtree semantics. A new Finite Hierarchical Approximation Theorem shows that a depth-K v-PuNN with $\sum_{j=0}^{K-1}p^{\,j}$ neurons universally represents any K-level tree. Because gradients vanish in this discrete space, we propose Valuation-Adaptive Perturbation Optimization (VAPO), with a fast deterministic variant (HiPaN-DS) and a moment-based one (HiPaN / Adam-VAPO). On three canonical benchmarks our CPU-only implementation sets new state-of-the-art: WordNet nouns (52,427 leaves) 99.96% leaf accuracy in 16 min; GO molecular-function 96.9% leaf / 100% root in 50 s; NCBI Mammalia Spearman $\rho = -0.96$ with true taxonomic distance. The learned metric is perfectly ultrametric (zero triangle violations), and its fractal and information-theoretic properties are analyzed. Beyond classification we derive structural invariants for quantum systems (HiPaQ) and controllable generative codes for tabular data (Tab-HiPaN). v-PuNNs therefore bridge number theory and deep learning, offering exact, interpretable, and efficient models for hierarchical data.
Authors: Arjun Manoj, Anastasia S. Georgiou, Dimitris G. Giovanis, Themistoklis P. Sapsis, Ioannis G. Kevrekidis
Abstract: Multi-fidelity optimization employs surrogate models that integrate information from varying levels of fidelity to guide efficient exploration of complex design spaces while minimizing the reliance on (expensive) high-fidelity objective function evaluations. To advance Gaussian Process (GP)-based multi-fidelity optimization, we implement a proximity-based acquisition strategy that simplifies fidelity selection by eliminating the need for separate acquisition functions at each fidelity level. We also enable multi-fidelity Upper Confidence Bound (UCB) strategies by combining them with multi-fidelity GPs rather than the standard GPs typically used. We benchmark these approaches alongside other multi-fidelity acquisition strategies (including fidelity-weighted approaches) comparing their performance, reliance on high-fidelity evaluations, and hyperparameter tunability in representative optimization tasks. The results highlight the capability of the proximity-based multi-fidelity acquisition function to deliver consistent control over high-fidelity usage while maintaining convergence efficiency. Our illustrative examples include multi-fidelity chemical kinetic models, both homogeneous and heterogeneous (dynamic catalysis for ammonia production).
Authors: Jesse He, Akbar Rafiey, Gal Mishne, Yusu Wang
Abstract: In recent years, the remarkable success of graph neural networks (GNNs) on graph-structured data has prompted a surge of methods for explaining GNN predictions. However, the state-of-the-art for GNN explainability remains in flux. Different comparisons find mixed results for different methods, with many explainers struggling on more complex GNN architectures and tasks. This presents an urgent need for a more careful theoretical analysis of competing GNN explanation methods. In this work we take a closer look at GNN explanations in two different settings: input-level explanations, which produce explanatory subgraphs of the input graph, and layerwise explanations, which produce explanatory subgraphs of the computation graph. We establish the first theoretical connections between the popular perturbation-based and classical gradient-based methods, as well as point out connections between other recently proposed methods. At the input level, we demonstrate conditions under which GNNExplainer can be approximated by a simple heuristic based on the sign of the edge gradients. In the layerwise setting, we point out that edge gradients are equivalent to occlusion search for linear GNNs. Finally, we demonstrate how our theoretical results manifest in practice with experiments on both synthetic and real datasets.
Authors: Nicholas E. Corrado, Josiah P. Hanna
Abstract: Independent on-policy policy gradient algorithms are widely used for multi-agent reinforcement learning (MARL) in cooperative and no-conflict games, but they are known to converge suboptimally when each agent's policy gradient points toward a suboptimal equilibrium. In this work, we identify a subtler failure mode that arises \textit{even when the expected policy gradients of all agents point toward an optimal solution.} After collecting a finite set of trajectories, stochasticity in independent action sampling can cause the joint data distribution to deviate from the expected joint on-policy distribution. This \textit{sampling error} w.r.t. the joint on-policy distribution produces inaccurate gradient estimates that can lead agents to converge suboptimally. In this paper, we investigate if joint sampling error can be reduced through coordinated action selection and whether doing so improves the reliability of policy gradient learning in MARL. Toward this end, we introduce an adaptive action sampling approach to reduce joint sampling error. Our method, Multi-Agent Proximal Robust On-Policy Sampling (MA-PROPS), uses a centralized behavior policy that we continually adapt to place larger probability on joint actions that are currently under-sampled w.r.t. the current joint policy. We empirically evaluate MA-PROPS in a diverse range of multi-agent games and demonstrate that (1) MA-PROPS reduces joint sampling error more efficiently than standard on-policy sampling and (2) improves the reliability of independent policy gradient algorithms, increasing the fraction of training runs that converge to an optimal joint policy.
Authors: Xuan Liu, Siru Ouyang, Xianrui Zhong, Jiawei Han, Huimin Zhao
Abstract: Large language models (LLMs) have gained significant attention in chemistry. However, most existing datasets center on molecular-level property prediction and overlook the role of fine-grained functional group (FG) information. Incorporating FG-level data can provide valuable prior knowledge that links molecular structures with textual descriptions, which can be used to build more interpretable, structure-aware LLMs for reasoning on molecule-related tasks. Moreover, LLMs can learn from such fine-grained information to uncover hidden relationships between specific functional groups and molecular properties, thereby advancing molecular design and drug discovery. Here, we introduce FGBench, a dataset comprising 625K molecular property reasoning problems with functional group information. Functional groups are precisely annotated and localized within the molecule, which ensures the dataset's interoperability thereby facilitating further multimodal applications. FGBench includes both regression and classification tasks on 245 different functional groups across three categories for molecular property reasoning: (1) single functional group impacts, (2) multiple functional group interactions, and (3) direct molecular comparisons. In the benchmark of state-of-the-art LLMs on 7K curated data, the results indicate that current LLMs struggle with FG-level property reasoning, highlighting the need to enhance reasoning capabilities in LLMs for chemistry tasks. We anticipate that the methodology employed in FGBench to construct datasets with functional group-level information will serve as a foundational framework for generating new question-answer pairs, enabling LLMs to better understand fine-grained molecular structure-property relationships. The dataset and evaluation code are available at https://github.com/xuanliugit/FGBench.
Authors: Johann Birnick
Abstract: We explain how data-driven quantization of a linear unit in a neural network corresponds to solving the closest vector problem for a certain lattice generated by input data. We prove that the GPTQ algorithm is equivalent to Babai's well-known nearest-plane algorithm. We furthermore provide geometric intuition for both algorithms. Lastly, we note the consequences of these results, in particular hinting at the possibility for using lattice basis reduction for better quantization.
Authors: Siddharth Rout, Eldad Haber, Stephane Gaudreault
Abstract: Learning dynamical systems is crucial across many fields, yet applying machine learning techniques remains challenging due to missing variables and noisy data. Classical mathematical models often struggle in these scenarios due to the arose ill-posedness of the physical systems. Stochastic machine learning techniques address this challenge by enabling the modeling of such ill-posed problems. Thus, a single known input to the trained machine learning model may yield multiple plausible outputs, and all of the outputs are correct. In such scenarios, probabilistic forecasting is inherently meaningful. In this study, we introduce a variant of flow matching for probabilistic forecasting which estimates possible future states as a distribution over possible outcomes rather than a single-point prediction. Perturbation of complex dynamical states is not trivial. Community uses typical Gaussian or uniform perturbations to crucial variables to model uncertainty. However, not all variables behave in a Gaussian fashion. So, we also propose a generative machine learning approach to physically and logically perturb the states of complex high-dimensional dynamical systems. Finally, we establish the mathematical foundations of our method and demonstrate its effectiveness on several challenging dynamical systems, including a variant of the high-dimensional WeatherBench dataset, which models the global weather at a 5.625{\deg} meridional resolution.
Authors: Md Sultanul Islam Ovi, Jamal Hossain, Md Raihan Alam Rahi, Fatema Akter
Abstract: Student mental health is an increasing concern in academic institutions, where stress can severely impact well-being and academic performance. Traditional assessment methods rely on subjective surveys and periodic evaluations, offering limited value for timely intervention. This paper introduces a context-aware machine learning framework for classifying student stress using two complementary survey-based datasets covering psychological, academic, environmental, and social factors. The framework follows a six-stage pipeline involving preprocessing, feature selection (SelectKBest, RFECV), dimensionality reduction (PCA), and training with six base classifiers: SVM, Random Forest, Gradient Boosting, XGBoost, AdaBoost, and Bagging. To enhance performance, we implement ensemble strategies, including hard voting, soft voting, weighted voting, and stacking. Our best models achieve 93.09% accuracy with weighted hard voting on the Student Stress Factors dataset and 99.53% with stacking on the Stress and Well-being dataset, surpassing previous benchmarks. These results highlight the potential of context-integrated, data-driven systems for early stress detection and underscore their applicability in real-world academic settings to support student well-being.
Authors: Yang Liu, Xuejiao Kang, Sathya Iyer, Idris Malik, Ruixuan Li, Juan Wang, Xinchen Lu, Xiangxue Zhao, Dayong Wang, Menghan Liu, Isaac Liu, Feng Liang, Yinzhe Yu
Abstract: User attributes are essential in multiple stages of modern recommendation systems and are particularly important for mitigating the cold-start problem and improving the experience of new or infrequent users. We propose Behavior-based User Segmentation (BUS), a novel tree-based data structure that hierarchically segments the user universe with various users' categorical attributes based on the users' product-specific engagement behaviors. During the BUS tree construction, we use Normalized Discounted Cumulative Gain (NDCG) as the objective function to maximize the behavioral representativeness of marginal users relative to active users in the same segment. The constructed BUS tree undergoes further processing and aggregation across the leaf nodes and internal nodes, allowing the generation of popular social content and behavioral patterns for each node in the tree. To further mitigate bias and improve fairness, we use the social graph to derive the user's connection-based BUS segments, enabling the combination of behavioral patterns extracted from both the user's own segment and connection-based segments as the connection aware BUS-based recommendation. Our offline analysis shows that the BUS-based retrieval significantly outperforms traditional user cohort-based aggregation on ranking quality. We have successfully deployed our data structure and machine learning algorithm and tested it with various production traffic serving billions of users daily, achieving statistically significant improvements in the online product metrics, including music ranking and email notifications. To the best of our knowledge, our study represents the first list-wise learning-to-rank framework for tree-based recommendation that effectively integrates diverse user categorical attributes while preserving real-world semantic interpretability at a large industrial scale.
Authors: Ran Li, Lingshu Zeng
Abstract: Pseudo-random number generators (PRNGs) are high-nonlinear processes, and they are key blocks in optimization of Large language models. Transformers excel at processing complex nonlinear relationships. Thus it is reasonable to generate high-quality pseudo-random numbers based on transformers. In this paper, we explore this question from both theoretical and practical perspectives, highlighting the potential benefits and implications of Transformer in PRNGs. We theoretically demonstrate that decoder-only Transformer models with Chain-of-Thought can simulate both the Linear Congruential Generator (LCG) and Mersenne Twister (MT) PRNGs. Based on this, we conclude that the log-precision decoder-only Transformer can represent non-uniform $\text{AC}^0$. Our simulative theoretical findings are validated through experiments. The random numbers generated by Transformer-based PRNGs successfully pass the majority of NIST tests, whose heat maps exhibit clear statistical randomness. Finally, we assess their capability in prediction attacks.
Authors: Kotaro Yoshida, Yuji Naraki, Takafumi Horie, Ryotaro Shimizu, Hiroki Naganuma
Abstract: Model merging has emerged as an efficient and flexible paradigm for multi-task learning, with numerous methods being proposed in recent years. However, these state-of-the-art techniques are typically evaluated on benchmark suites that are highly favorable to model merging, and their robustness in more realistic settings remains largely unexplored. In this work, we first investigate the vulnerabilities of model-merging methods and pinpoint the source-model characteristics that critically underlie them. Specifically, we identify two factors that are particularly harmful to the merging process: (1) disparities in task vector norms, and (2) the low confidence of the source models. To address this issue, we propose DisTaC (Distillation for Task vector Conditioning), a novel method that pre-conditions these problematic task vectors before the merge. DisTaC leverages knowledge distillation to adjust a task vector's norm and increase source-model confidence while preserving its essential task-specific knowledge. Our extensive experiments demonstrate that by pre-conditioning task vectors with DisTaC, state-of-the-art merging techniques can successfully integrate models exhibiting the harmful traits -- where they would otherwise fail -- achieving significant performance gains.
Authors: Hongquan Zhang, Jingyu Gong, Zhizhong Zhang, Xin Tan, Yanyun Qu, Yuan Xie
Abstract: The main challenge in lifelong imitation learning lies in the balance between mitigating catastrophic forgetting of previous skills while maintaining sufficient capacity for acquiring new ones. However, current approaches typically address these aspects in isolation, overlooking their internal correlation in lifelong skill acquisition. We address this limitation with a unified framework named Tokenized Skill Scaling (T2S). Specifically, by tokenizing the model parameters, the linear parameter mapping of the traditional transformer is transformed into cross-attention between input and learnable tokens, thereby enhancing model scalability through the easy extension of new tokens. Additionally, we introduce language-guided skill scaling to transfer knowledge across tasks efficiently and avoid linearly growing parameters. Extensive experiments across diverse tasks demonstrate that T2S: 1) effectively prevents catastrophic forgetting (achieving an average NBT of 1.0% across the three LIBERO task suites), 2) excels in new skill scaling with minimal increases in trainable parameters (needing only 8.0% trainable tokens in an average of lifelong tasks), and 3) enables efficient knowledge transfer between tasks (achieving an average FWT of 77.7% across the three LIBERO task suites), offering a promising solution for lifelong imitation learning.
Authors: Jiayi Chen, Jing Li, Guiling Wang
Abstract: Reinforcement Learning (RL) has shown significant promise in automated portfolio management; however, effectively balancing risk and return remains a central challenge, as many models fail to adapt to dynamically changing market conditions. In this paper, we propose Meta-controlled Agents for a Risk-aware System (MARS), a novel RL framework designed to explicitly address this limitation through a multi-agent, risk-aware approach. Instead of a single monolithic model, MARS employs a Heterogeneous Agent Ensemble where each agent possesses a unique, intrinsic risk profile. This profile is enforced by a dedicated Safety-Critic network and a specific risk-tolerance threshold, allowing agents to specialize in behaviors ranging from capital preservation to aggressive growth. To navigate different market regimes, a high-level Meta-Adaptive Controller (MAC) learns to dynamically orchestrate the ensemble. By adjusting its reliance on conservative versus aggressive agents, the MAC effectively lowers portfolio volatility during downturns and seeks higher returns in bull markets, thus minimizing maximum drawdown and enhancing overall stability. This two-tiered structure allows MARS to generate a disciplined and adaptive portfolio that is robust to market fluctuations. The framework achieves a superior balance between risk and return by leveraging behavioral diversity rather than explicit market-feature engineering. Experiments on major international stock indexes, including periods of significant financial crisis, demonstrate the efficacy of our framework on risk-adjusted criteria, significantly reducing maximum drawdown and volatility while maintaining competitive returns.
Authors: Kaichen Zhang, Shenghao Gao, Yuzhong Hong, Haipeng Sun, Junwei Bao, Hongfei Jiang, Yang Song, Hong Dingqian, Hui Xiong
Abstract: Current large language model post-training optimizes a risk-neutral objective that maximizes expected reward, yet evaluation relies heavily on risk-seeking metrics like Pass@k (at least one success in k trials) and Max@k (maximum reward across k responses). This mismatch in risk preferences can inevitably lead to suboptimal performance. To bridge this gap, we propose Risk-Seeking Policy Optimization (RSPO), a novel method that directly targets Pass@k and Max@k during training. A key challenge in optimizing these metrics is the "hitchhiking" problem: low-reward responses are inadvertently reinforced if they co-occur with a high-reward response within a sample of k generations, resulting in inefficient optimization. RSPO addresses this problem by leveraging the closed-form probability that a given response is the maximum among k samplings. Despite the complexity of nested gradients over multiple responses, RSPO produces efficient, unbiased gradient estimators for both metrics. We validate our approach with both rigorous theoretical analysis and comprehensive experimental results.
Authors: Shiko Kudo
Abstract: The dominant paradigm in modern neural networks relies on simple, monotonically-increasing activation functions like ReLU. While effective, this paradigm necessitates large, massively-parameterized models to approximate complex functions. In this paper, we introduce the Periodic Linear Unit (PLU), a learnable sine-wave based activation with periodic non-monotonicity. PLU is designed for maximum expressive power and numerical stability, achieved through its formulation and a paired innovation we term Repulsive Reparameterization, which prevents the activation from collapsing into a non-expressive linear function. We demonstrate that a minimal MLP with only two PLU neurons can solve the spiral classification task, a feat impossible for equivalent networks using standard activations. This suggests a paradigm shift from networks as piecewise Taylor-like approximators to powerful Fourier-like function synthesizers, achieving exponential gains in parameter efficiency by placing intelligence in the neuron itself.
Authors: Zhuo Yang, Jiaqing Xie, Shuaike Shen, Daolang Wang, Yeyun Chen, Ben Gao, Shuzhou Sun, Biqing Qi, Dongzhan Zhou, Lei Bai, Linjiang Chen, Shufei Zhang, Jun Jiang, Tianfan Fu, Yuqiang Li
Abstract: Deep learning holds immense promise for spectroscopy, yet research and evaluation in this emerging field often lack standardized formulations. To address this issue, we introduce SpectrumLab, a pioneering unified platform designed to systematize and accelerate deep learning research in spectroscopy. SpectrumLab integrates three core components: a comprehensive Python library featuring essential data processing and evaluation tools, along with leaderboards; an innovative SpectrumAnnotator module that generates high-quality benchmarks from limited seed data; and SpectrumBench, a multi-layered benchmark suite covering 14 spectroscopic tasks and over 10 spectrum types, featuring spectra curated from over 1.2 million distinct chemical substances. Thorough empirical studies on SpectrumBench with 18 cutting-edge multimodal LLMs reveal critical limitations of current approaches. We hope SpectrumLab will serve as a crucial foundation for future advancements in deep learning-driven spectroscopy.
Authors: Kun Li, Zhennan Wu, Yida Xiong, Hongzhi Zhang, Longtao Hu, Zhonglie Liu, Junqi Zeng, Wenjie Wu, Mukun Chen, Jiameng Chen, Wenbin Hu
Abstract: Drug discovery is of great social significance in safeguarding human health, prolonging life, and addressing the challenges of major diseases. In recent years, artificial intelligence has demonstrated remarkable advantages in key tasks across bioinformatics and pharmacology, owing to its efficient data processing and data representation capabilities. However, most existing computational platforms cover only a subset of core tasks, leading to fragmented workflows and low efficiency. In addition, they often lack algorithmic innovation and show poor generalization to out-of-distribution (OOD) data, which greatly hinders the progress of drug discovery. To address these limitations, we propose Baishenglai (BSL), a deep learning-enhanced, open-access platform designed for virtual drug discovery. BSL integrates seven core tasks within a unified and modular framework, incorporating advanced technologies such as generative models and graph neural networks. In addition to achieving state-of-the-art (SOTA) performance on multiple benchmark datasets, the platform emphasizes evaluation mechanisms that focus on generalization to OOD molecular structures. Comparative experiments with existing platforms and baseline methods demonstrate that BSL provides a comprehensive, scalable, and effective solution for virtual drug discovery, offering both algorithmic innovation and high-precision prediction for real-world pharmaceutical research. In addition, BSL demonstrated its practical utility by discovering novel modulators of the GluN1/GluN3A NMDA receptor, successfully identifying three compounds with clear bioactivity in in-vitro electrophysiological assays. These results highlight BSL as a promising and comprehensive platform for accelerating biomedical research and drug discovery. The platform is accessible at https://www.baishenglai.net.
Authors: Sukwon Yun, Xin Liu, Yunhak Oh, Junseok Lee, Tianlong Chen, Tsuyoshi Murata, Chanyoung Park
Abstract: In real-world graphs, we often encounter missing feature situations where a few or the majority of node features, e.g., sensitive information, are missed. In such scenarios, directly utilizing Graph Neural Networks (GNNs) would yield sub-optimal results in downstream tasks such as node classification. Despite the emergence of a few GNN-based methods attempting to mitigate its missing situation, when only a few features are available, they rather perform worse than traditional structure-based models. To this end, we propose a novel framework that further illuminates the potential of classical Label Propagation (Oldie), taking advantage of Feature Propagation, especially when only a partial feature is available. Now called by GOODIE, it takes a hybrid approach to obtain embeddings from the Label Propagation branch and Feature Propagation branch. To do so, we first design a GNN-based decoder that enables the Label Propagation branch to output hidden embeddings that align with those of the FP branch. Then, GOODIE automatically captures the significance of structure and feature information thanks to the newly designed Structure-Feature Attention. Followed by a novel Pseudo-Label contrastive learning that differentiates the contribution of each positive pair within pseudo-labels originating from the LP branch, GOODIE outputs the final prediction for the unlabeled nodes. Through extensive experiments, we demonstrate that our proposed model, GOODIE, outperforms the existing state-of-the-art methods not only when only a few features are available but also in abundantly available situations. Source code of GOODIE is available at: https://github.com/SukwonYun/GOODIE.
Authors: Yile Li, Shandian Zhe
Abstract: Learning solution operators for partial differential equations (PDEs) has become a foundational task in scientific machine learning. However, existing neural operator methods require abundant training data for each specific PDE and lack the ability to generalize across PDE families. In this work, we propose MOFS: a unified multimodal framework for multi-operator few-shot learning, which aims to generalize to unseen PDE operators using only a few demonstration examples. Our method integrates three key components: (i) multi-task self-supervised pretraining of a shared Fourier Neural Operator (FNO) encoder to reconstruct masked spatial fields and predict frequency spectra, (ii) text-conditioned operator embeddings derived from statistical summaries of input-output fields, and (iii) memory-augmented multimodal prompting with gated fusion and cross-modal gradient-based attention. We adopt a two-stage training paradigm that first learns prompt-conditioned inference on seen operators and then applies end-to-end contrastive fine-tuning to align latent representations across vision, frequency, and text modalities. Experiments on PDE benchmarks, including Darcy Flow and Navier Stokes variants, demonstrate that our model outperforms existing operator learning baselines in few-shot generalization. Extensive ablations validate the contributions of each modality and training component. Our approach offers a new foundation for universal and data-efficient operator learning across scientific domains.
Authors: Juntong Chen, Huayuan Ye, He Zhu, Siwei Fu, Changbo Wang, Chenhui Li
Abstract: Accurate and reliable visualization of spatiotemporal sensor data such as environmental parameters and meteorological conditions is crucial for informed decision-making. Traditional spatial interpolation methods, however, often fall short of producing reliable interpolation results due to the limited and irregular sensor coverage. This paper introduces a novel spatial interpolation pipeline that achieves reliable interpolation results and produces a novel heatmap representation with uncertainty information encoded. We leverage imputation reference data from Graph Neural Networks (GNNs) to enhance visualization reliability and temporal resolution. By integrating Principal Neighborhood Aggregation (PNA) and Geographical Positional Encoding (GPE), our model effectively learns the spatiotemporal dependencies. Furthermore, we propose an extrinsic, static visualization technique for interpolation-based heatmaps that effectively communicates the uncertainties arising from various sources in the interpolated map. Through a set of use cases, extensive evaluations on real-world datasets, and user studies, we demonstrate our model's superior performance for data imputation, the improvements to the interpolant with reference data, and the effectiveness of our visualization design in communicating uncertainties.
Authors: Hung-Chieh Fang, Hsuan-Tien Lin, Irwin King, Yifei Zhang
Abstract: Federated Unsupervised Learning (FUL) aims to learn expressive representations in federated and self-supervised settings. The quality of representations learned in FUL is usually determined by uniformity, a measure of how uniformly representations are distributed in the embedding space. However, existing solutions perform well in achieving intra-client (local) uniformity for local models while failing to achieve inter-client (global) uniformity after aggregation due to non-IID data distributions and the decentralized nature of FUL. To address this issue, we propose Soft Separation and Distillation (SSD), a novel approach that preserves inter-client uniformity by encouraging client representations to spread toward different directions. This design reduces interference during client model aggregation, thereby improving global uniformity while preserving local representation expressiveness. We further enhance this effect by introducing a projector distillation module to address the discrepancy between loss optimization and representation quality. We evaluate SSD in both cross-silo and cross-device federated settings, demonstrating consistent improvements in representation quality and task performance across various training scenarios. Our results highlight the importance of inter-client uniformity in FUL and establish SSD as an effective solution to this challenge. Project page: https://ssd-uniformity.github.io/
Authors: Micah Rentschler, Jesse Roberts
Abstract: Ensuring sufficient exploration is a central challenge when training meta-reinforcement learning (meta-RL) agents to solve novel environments. Conventional solutions to the exploration-exploitation dilemma inject explicit incentives such as randomization, uncertainty bonuses, or intrinsic rewards to encourage exploration. In this work, we hypothesize that an agent trained solely to maximize a greedy (exploitation-only) objective can nonetheless exhibit emergent exploratory behavior, provided three conditions are met: (1) Recurring Environmental Structure, where the environment features repeatable regularities that allow past experience to inform future choices; (2) Agent Memory, enabling the agent to retain and utilize historical interaction data; and (3) Long-Horizon Credit Assignment, where learning propagates returns over a time frame sufficient for the delayed benefits of exploration to inform current decisions. Through experiments in stochastic multi-armed bandits and temporally extended gridworlds, we observe that, when both structure and memory are present, a policy trained on a strictly greedy objective exhibits information-seeking exploratory behavior. We further demonstrate, through controlled ablations, that emergent exploration vanishes if either environmental structure or agent memory is absent (Conditions 1 & 2). Surprisingly, removing long-horizon credit assignment (Condition 3) does not always prevent emergent exploration-a result we attribute to the pseudo-Thompson Sampling effect. These findings suggest that, under the right prerequisites, exploration and exploitation need not be treated as orthogonal objectives but can emerge from a unified reward-maximization process.
Authors: Shangshang Yang, Jialin Han, Xiaoshan Yu, Ziwen Wang, Hao Jiang, Haiping Ma, Xingyi Zhang, Geyong Min
Abstract: Online intelligent education platforms have generated a vast amount of distributed student learning data. This influx of data presents opportunities for cognitive diagnosis (CD) to assess students' mastery of knowledge concepts while also raising significant data privacy and security challenges. To cope with this issue, federated learning (FL) becomes a promising solution by jointly training models across multiple local clients without sharing their original data. However, the data quality problem, caused by the ability differences and educational context differences between different groups/schools of students, further poses a challenge to the fairness of models. To address this challenge, this paper proposes a fairness-aware federated cognitive diagnosis framework (FedCD) to jointly train CD models built upon a novel parameter decoupling-based personalization strategy, preserving privacy of data and achieving precise and fair diagnosis of students on each client. As an FL paradigm, FedCD trains a local CD model for the students in each client based on its local student learning data, and each client uploads its partial model parameters to the central server for parameter aggregation according to the devised innovative personalization strategy. The main idea of this strategy is to decouple model parameters into two parts: the first is used as locally personalized parameters, containing diagnostic function-related model parameters, to diagnose each client's students fairly; the second is the globally shared parameters across clients and the server, containing exercise embedding parameters, which are updated via fairness-aware aggregation, to alleviate inter-school unfairness. Experiments on three real-world datasets demonstrate the effectiveness of the proposed FedCD framework and the personalization strategy compared to five FL approaches under three CD models.
Authors: Joshua Dimasaka, Christian Gei{\ss}, Emily So
Abstract: Regional disaster resilience quantifies the changing nature of physical risks to inform policy instruments ranging from local immediate recovery to international sustainable development. While many existing state-of-practice methods have greatly advanced the dynamic mapping of exposure and hazard, our understanding of large-scale physical vulnerability has remained static, costly, limited, region-specific, coarse-grained, overly aggregated, and inadequately calibrated. With the significant growth in the availability of time-series satellite imagery and derived products for exposure and hazard, we focus our work on the equally important yet challenging element of the risk equation: physical vulnerability. We leverage machine learning methods that flexibly capture spatial contextual relationships, limited temporal observations, and uncertainty in a unified probabilistic spatiotemporal inference framework. We therefore introduce Graph Variational State-Space Model (GraphVSSM), a novel modular spatiotemporal approach that uniquely integrates graph deep learning, state-space modeling, and variational inference using time-series data and prior expert belief systems in a weakly supervised or coarse-to-fine-grained manner. We present three major results: a city-wide demonstration in Quezon City, Philippines; an investigation of sudden changes in the cyclone-impacted coastal Khurushkul community (Bangladesh) and mudslide-affected Freetown (Sierra Leone); and an open geospatial dataset, METEOR 2.5D, that spatiotemporally enhances the existing global static dataset for UN Least Developed Countries (2020). Beyond advancing regional disaster resilience assessment and improving our understanding global disaster risk reduction progress, our method also offers a probabilistic deep learning approach, contributing to broader urban studies that require compositional data analysis in weak supervision.
Authors: Vamsi Sai Krishna Malineni, Suresh Rajendran
Abstract: The utilization of Deep Neural Networks (DNNs) in physical science and engineering applications has gained traction due to their capacity to learn intricate functions. While large datasets are crucial for training DNN models in fields like computer vision and natural language processing, obtaining such datasets for engineering applications is prohibitively expensive. Physics-Informed Neural Networks (PINNs), a branch of Physics-Informed Machine Learning (PIML), tackle this challenge by embedding physical principles within neural network architectures. PINNs have been extensively explored for solving diverse forward and inverse problems in fluid mechanics. Nonetheless, there is limited research on employing PINNs for flow reconstruction from sparse data under constrained computational resources. Earlier studies were focused on forward problems with well-defined data. The present study attempts to develop models capable of reconstructing the flow field data from sparse datasets mirroring real-world scenarios. This study focuses on two cases: (a) two-dimensional (2D) unsteady laminar flow past a circular cylinder and (b) three-dimensional (3D) unsteady turbulent flow past an ultra-large container ship (ULCS). The first case compares the effectiveness of training methods like Standard PINN and Backward Compatible PINN (BC-PINN) and explores the performance enhancements through systematic relaxation of physics constraints and dynamic weighting of loss function components. The second case highlights the capability of PINN-based models to learn underlying physics from sparse data while accurately reconstructing the flow field for a highly turbulent flow.
Authors: Christopher Godwin Udomboso, Caston Sigauke, Ini Adinya
Abstract: Effective data partitioning is known to be crucial in machine learning. Traditional cross-validation methods like K-Fold Cross-Validation (KFCV) enhance model robustness but often compromise generalisation assessment due to high computational demands and extensive data shuffling. To address these issues, the integration of the Simple Random Sampling (SRS), which, despite providing representative samples, can result in non-representative sets with imbalanced data. The study introduces a hybrid model, Fusion Sampling Validation (FSV), combining SRS and KFCV to optimise data partitioning. FSV aims to minimise biases and merge the simplicity of SRS with the accuracy of KFCV. The study used three datasets of 10,000, 50,000, and 100,000 samples, generated with a normal distribution (mean 0, variance 1) and initialised with seed 42. KFCV was performed with five folds and ten repetitions, incorporating a scaling factor to ensure robust performance estimation and generalisation capability. FSV integrated a weighted factor to enhance performance and generalisation further. Evaluations focused on mean estimates (ME), variance estimates (VE), mean squared error (MSE), bias, the rate of convergence for mean estimates (ROC\_ME), and the rate of convergence for variance estimates (ROC\_VE). Results indicated that FSV consistently outperformed SRS and KFCV, with ME values of 0.000863, VE of 0.949644, MSE of 0.952127, bias of 0.016288, ROC\_ME of 0.005199, and ROC\_VE of 0.007137. FSV demonstrated superior accuracy and reliability in data partitioning, particularly in resource-constrained environments and extensive datasets, providing practical solutions for effective machine learning implementations.
Authors: Glen Berseth
Abstract: In the era of deep reinforcement learning, making progress is more complex, as the collected experience must be compressed into a deep model for future exploitation and sampling. Many papers have shown that training a deep learning policy under the changing state and action distribution leads to sub-optimal performance, or even collapse. This naturally leads to the concern that even if the community creates improved exploration algorithms or reward objectives, will those improvements fall on the \textit{deaf ears} of optimization difficulties. This work proposes a new \textit{practical} sub-optimality estimator to determine optimization limitations of deep reinforcement learning algorithms. Through experiments across environments and RL algorithms, it is shown that the difference between the best experience generated is 2-3$\times$ better than the policies' learned performance. This large difference indicates that deep RL methods only exploit half of the good experience they generate.
Authors: Xin Chen, Shuaijun Chen, Omid Tavallaie, Nguyen Tran, Shuhuang Xiang, Albert Zomaya
Abstract: Federated Learning (FL) enables collaborative model training across decentralized data sources while preserving data privacy. However, the growing size of Machine Learning (ML) models poses communication and computation challenges in FL. Low-Rank Adaptation (LoRA) has recently been introduced into FL as an efficient fine-tuning method, reducing communication overhead by updating only a small number of trainable parameters. Despite its effectiveness, how to aggregate LoRA-updated local models on the server remains a critical and understudied problem. In this paper, we provide a unified convergence analysis for LoRA-based FL. We first categories the current aggregation method into two major type: Sum-Product (SP) and Product-Sum (PS). Then we formally define the Aggregation-Broadcast Operator (ABO) and derive a general convergence condition under mild assumptions. Furthermore, we present several sufficient conditions that guarantee convergence of the global model. These theoretical analyze offer a principled understanding of various aggregation strategies. Notably, we prove that the SP and PS aggregation methods both satisfy our convergence condition, but differ in their ability to achieve the optimal convergence rate. Extensive experiments on standard benchmarks validate our theoretical findings.
Authors: R\'emi Bardenet, Myl\`ene Ma\"ida, Martin Rouault
Abstract: Gibbs measures, such as Coulomb gases, are popular in modelling systems of interacting particles. Recently, we proposed to use Gibbs measures as randomized numerical integration algorithms with respect to a target measure $\pi$ on $\mathbb R^d$, following the heuristics that repulsiveness between particles should help reduce integration errors. A major issue in this approach is to tune the interaction kernel and confining potential of the Gibbs measure, so that the equilibrium measure of the system is the target distribution $\pi$. Doing so usually requires another Monte Carlo approximation of the \emph{potential}, i.e. the integral of the interaction kernel with respect to $\pi$. Using the methodology of large deviations from Garcia--Zelada (2019), we show that a random approximation of the potential preserves the fast large deviation principle that guarantees the proposed integration algorithm to outperform independent or Markov quadratures. For non-singular interaction kernels, we make minimal assumptions on this random approximation, which can be the result of a computationally cheap Monte Carlo preprocessing. For the Coulomb interaction kernel, we need the approximation to be based on another Gibbs measure, and we prove in passing a control on the uniform convergence of the approximation of the potential.
Authors: Stefan Bielmeier, Gerald Friedland
Abstract: We investigate how feature correlations influence the capacity of Dense Associative Memory (DAM), a Transformer attention-like model. Practical machine learning scenarios involve feature-correlated data and learn representations in the input space, but current capacity analyses do not account for this. We develop an empirical framework to analyze the effects of data structure on capacity dynamics. Specifically, we systematically construct datasets that vary in feature correlation and pattern separation using Hamming distance from information theory, and compute the model's corresponding storage capacity using a simple binary search algorithm. Our experiments confirm that memory capacity scales exponentially with increasing separation in the input space. Feature correlations do not alter this relationship fundamentally, but reduce capacity slightly at constant separation. This effect is amplified at higher polynomial degrees in the energy function, suggesting that Associative Memory is more limited in depicting higher-order interactions between features than patterns. Our findings bridge theoretical work and practical settings for DAM, and might inspire more data-centric methods.
Authors: Hongwei Ma, Junbin Gao, Minh-Ngoc Tran
Abstract: Accurate, explainable and physically-credible forecasting remains a persistent challenge for multivariate time-series whose statistical properties vary across domains. We present CPformer, a Concept- and Physics-enhanced Transformer that channels every prediction through five self-supervised, domain-agnostic concepts while enforcing differentiable residuals drawn from first-principle constraints. Unlike prior efficiency-oriented Transformers that rely purely on sparsity or frequency priors , CPformer combines latent transparency with hard scientific guidance while retaining attention for long contexts. We tested CPformer on six publicly-available datasets: sub-hourly Electricity and Traffic, hourly ETT, high-dimensional Weather, weekly Influenza-like Illness, and minute-level Exchange Rate, and CPformer achieves the lowest error in eight of twelve MSE/MAE cells. Relative to the strongest Transformer baseline (FEDformer), CPformer reduces mean-squared-error by 23% on Electricity, 44% on Traffic and 61% on Illness, while matching performance on strictly periodic Weather and ETT series.
Authors: Md Zahidul Islam, Md Shafiqur Rahman, Md Sumsuzoha, Babul Sarker, Md Rafiqul Islam, Mahfuz Alam, Sanjib Kumar Shil
Abstract: Cryptocurrency markets are experiencing rapid growth, but this expansion comes with significant challenges, particularly in predicting cryptocurrency prices for traders in the U.S. In this study, we explore how deep learning and machine learning models can be used to forecast the closing prices of the XRP/USDT trading pair. While many existing cryptocurrency prediction models focus solely on price and volume patterns, they often overlook market liquidity, a crucial factor in price predictability. To address this, we introduce two important liquidity proxy metrics: the Volume-To-Volatility Ratio (VVR) and the Volume-Weighted Average Price (VWAP). These metrics provide a clearer understanding of market stability and liquidity, ultimately enhancing the accuracy of our price predictions. We developed four machine learning models, Linear Regression, Random Forest, XGBoost, and LSTM neural networks, using historical data without incorporating the liquidity proxy metrics, and evaluated their performance. We then retrained the models, including the liquidity proxy metrics, and reassessed their performance. In both cases (with and without the liquidity proxies), the LSTM model consistently outperformed the others. These results underscore the importance of considering market liquidity when predicting cryptocurrency closing prices. Therefore, incorporating these liquidity metrics is essential for more accurate forecasting models. Our findings offer valuable insights for traders and developers seeking to create smarter and more risk-aware strategies in the U.S. digital assets market.
Authors: Hang Ni, Weijia Zhang, Hao Liu
Abstract: Recent advancements in deep learning have led to the development of Foundation Models (FMs) for weather forecasting, yet their ability to predict extreme weather events remains limited. Existing approaches either focus on general weather conditions or specialize in specific-type extremes, neglecting the real-world atmospheric patterns of diversified extreme events. In this work, we identify two key characteristics of extreme events: (1) the spectral disparity against normal weather regimes, and (2) the hierarchical drivers and geographic blending of diverse extremes. Along this line, we propose UniExtreme, a universal extreme weather forecasting foundation model that integrates (1) an Adaptive Frequency Modulation (AFM) module that captures region-wise spectral differences between normal and extreme weather, through learnable Beta-distribution filters and multi-granularity spectral aggregation, and (2) an Event Prior Augmentation (EPA) module which incorporates region-specific extreme event priors to resolve hierarchical extreme diversity and composite extreme schema, via a dual-level memory fusion network. Extensive experiments demonstrate that UniExtreme outperforms state-of-the-art baselines in both extreme and general weather forecasting, showcasing superior adaptability across diverse extreme scenarios.
Authors: Shayan Alahyari, Shiva Mehdipour Ghobadlou, Mike Domaratzki
Abstract: Imbalanced regression arises when the target distribution is skewed, causing models to focus on dense regions and struggle with underrepresented (minority) samples. Despite its relevance across many applications, few methods have been designed specifically for this challenge. Existing approaches often rely on fixed, ad hoc thresholds to label samples as rare or common, overlooking the continuous complexity of the joint feature-target space and fail to represent the true underlying rare regions. To address these limitations, we propose a fully data-driven GAN-based augmentation framework that uses Mahalanobis-Gaussian Mixture Modeling (GMM) to automatically identify minority samples and employs deterministic nearest-neighbour matching to enrich sparse regions. Rather than preset thresholds, our method lets the data determine which observations are truly rare. Evaluation on 32 benchmark imbalanced regression datasets demonstrates that our approach consistently outperforms state-of-the-art data augmentation methods.
Authors: Mikhail Andronov, Natalia Andronova, Michael Wand, J\"urgen Schmidhuber, Djork-Arn\'e Clevert
Abstract: AI-based computer-aided synthesis planning (CASP) systems are in demand as components of AI-driven drug discovery workflows. However, the high latency of such CASP systems limits their utility for high-throughput synthesizability screening in de novo drug design. We propose a method for accelerating multi-step synthesis planning systems that rely on SMILES-to-SMILES transformers as single-step retrosynthesis models. Our approach reduces the latency of SMILES-to-SMILES transformers powering multi-step synthesis planning in AiZynthFinder through speculative beam search combined with a scalable drafting strategy called Medusa. Replacing standard beam search with our approach allows the CASP system to solve 26\% to 86\% more molecules under the same time constraints of several seconds. Our method brings AI-based CASP systems closer to meeting the strict latency requirements of high-throughput synthesizability screening and improving general user experience.
Authors: Ivan Karpukhin, Andrey Savchenko
Abstract: Deep learning has achieved remarkable success in modeling sequential data, including event sequences, temporal point processes, and irregular time series. Recently, transformers have largely replaced recurrent networks in these tasks. However, transformers often underperform RNNs in classification tasks where the objective is to predict future targets. The reason behind this performance gap remains largely unexplored. In this paper, we identify a key limitation of transformers: the absence of a single state vector that provides a compact and effective representation of the entire sequence. Additionally, we show that contrastive pretraining of embedding vectors fails to capture local context, which is crucial for accurate prediction. To address these challenges, we introduce history tokens, a novel concept that facilitates the accumulation of historical information during next-token prediction pretraining. Our approach significantly improves transformer-based models, achieving impressive results in finance, e-commerce, and healthcare tasks. The code is publicly available on GitHub.
Authors: Akhila Henry, Nithin Nagaraj
Abstract: Neurochaos Learning (NL) is a brain-inspired classification framework that employs chaotic dynamics to extract features from input data and yields state of the art performance on classification tasks. However, NL requires the tuning of multiple hyperparameters and computing of four chaotic features per input sample. In this paper, we propose AutochaosNet - a novel, hyperparameter-free variant of the NL algorithm that eliminates the need for both training and parameter optimization. AutochaosNet leverages a universal chaotic sequence derived from the Champernowne constant and uses the input stimulus to define firing time bounds for feature extraction. Two simplified variants - TM AutochaosNet and TM-FR AutochaosNet - are evaluated against the existing NL architecture - ChaosNet. Our results demonstrate that AutochaosNet achieves competitive or superior classification performance while significantly reducing training time due to reduced computational effort. In addition to eliminating training and hyperparameter tuning, AutochaosNet exhibits excellent generalisation capabilities, making it a scalable and efficient choice for real-world classification tasks. Future work will focus on identifying universal orbits under various chaotic maps and incorporating them into the NL framework to further enhance performance.
Authors: Aleksandr Dremov, Alexander H\"agele, Atli Kosson, Martin Jaggi
Abstract: Learning rate scheduling is essential in transformer training, where the final annealing plays a crucial role in getting the best performance. However, the mechanisms behind this cooldown phase, with its characteristic drop in loss, remain poorly understood. To address this, we provide a comprehensive analysis focusing solely on the cooldown phase in the Warmup-Stable-Decay (WSD) learning rate scheduler. Our analysis reveals that different cooldown shapes reveal a fundamental bias-variance trade-off in the resulting models, with shapes that balance exploration and exploitation consistently outperforming alternatives. Similarly, we find substantial performance variations $\unicode{x2013}$ comparable to those from cooldown shape selection $\unicode{x2013}$ when tuning AdamW hyperparameters. Notably, we observe consistent improvements with higher values of $\beta_2$ during cooldown. From a loss landscape perspective, we provide visualizations of the landscape during cooldown, supporting the river valley loss perspective empirically. These findings offer practical recommendations for configuring the WSD scheduler in transformer training, emphasizing the importance of optimizing the cooldown phase alongside traditional hyperparameter tuning.
Authors: Jiaxing Qiu, Dongliang Guo, Brynne Sullivan, Teague R. Henry, Tom Hartvigsen
Abstract: In time series editing, we aim to modify some properties of a given time series without altering others. For example, when analyzing a hospital patient's blood pressure, we may add a sudden early drop and observe how it impacts their future while preserving other conditions. Existing diffusion-based editors rely on rigid, predefined attribute vectors as conditions and produce all-or-nothing edits through sampling. This attribute- and sampling-based approach limits flexibility in condition format and lacks customizable control over editing strength. To overcome these limitations, we introduce Instruction-based Time Series Editing, where users specify intended edits using natural language. This allows users to express a wider range of edits in a more accessible format. We then introduce InstructTime, the first instruction-based time series editor. InstructTime takes in time series and instructions, embeds them into a shared multi-modal representation space, then decodes their embeddings to generate edited time series. By learning a structured multi-modal representation space, we can easily interpolate between embeddings to achieve varying degrees of edit. To handle local and global edits together, we propose multi-resolution encoders. In our experiments, we use synthetic and real datasets and find that InstructTime is a state-of-the-art time series editor: InstructTime achieves high-quality edits with controllable strength, can generalize to unseen instructions, and can be easily adapted to unseen conditions through few-shot learning.
Authors: Azaz-Ur-Rehman Nasir, Samroz Ahmad Shoaib, Muhammad Abdullah Hanif, Muhammad Shafique
Abstract: Hardware-aware Neural Architecture Search (NAS) is one of the most promising techniques for designing efficient Deep Neural Networks (DNNs) for resource-constrained devices. Surrogate models play a crucial role in hardware-aware NAS as they enable efficient prediction of performance characteristics (e.g., inference latency and energy consumption) of different candidate models on the target hardware device. In this paper, we focus on building hardware-aware latency prediction models. We study different types of surrogate models and highlight their strengths and weaknesses. We perform a systematic analysis to understand the impact of different factors that can influence the prediction accuracy of these models, aiming to assess the importance of each stage involved in the model designing process and identify methods and policies necessary for designing/training an effective estimation model, specifically for GPU-powered devices. Based on the insights gained from the analysis, we present a holistic framework that enables reliable dataset generation and efficient model generation, considering the overall costs of different stages of the model generation pipeline.
Authors: Zishan Shao, Yixiao Wang, Qinsi Wang, Ting Jiang, Zhixu Du, Hancheng Ye, Danyang Zhuo, Yiran Chen, Hai Li
Abstract: Singular Value Decomposition (SVD) has recently seen a surge of interest as a simple yet powerful tool for large language models (LLMs) compression, with a growing number of works demonstrating 20-80% parameter reductions at minimal accuracy loss. Previous SVD-based approaches have focused primarily on reducing the memory footprint of model weights, largely overlooking the additional activation memory overhead incurred during inference when applying truncated factors via standard dense CUDA kernels. Our experiments demonstrate that this activation overhead, scaling with sequence length and hidden dimension, prevents current SVD compression techniques from achieving any reduction in peak inference memory, thereby limiting their viability for real-world, on-device deployments. We introduce FlashSVD, a novel, end-to-end rank-aware streaming inference framework specifically designed for SVD-compressed large language models. FlashSVD can be seamlessly integrated with any model that employs SVD-based methods for parameter reduction. By fusing low-rank projection kernels directly into both the self-attention and feed-forward network (FFN) pipelines, FlashSVD avoid materializing full-size activation buffers. Instead, small tiles of the truncated factors are loaded into on-chip SRAM, multiplied and reduced on the fly, and immediately evicted, preserving high GPU occupancy and adding no extra latency. On standard encoder benchmarks (e.g., BERT-Base), FlashSVD cuts peak activation memory by up to 70.2% and intermediate transient memory by 75%, all while incur no accuracy loss with upstreaming compression methods, offering a practical path toward memory-constrained deployment of low-rank LLMs.
Authors: Menglin Kong, Vincent Zhihao Zheng, Lijun Sun
Abstract: Many real-world time series exhibit strong periodic structures arising from physical laws, human routines, or seasonal cycles. However, modern deep forecasting models often fail to capture these recurring patterns due to spectral bias and a lack of frequency-aware inductive priors. Motivated by this gap, we propose a simple yet effective method that enhances long-term forecasting by explicitly modeling periodicity through spectral initialization and frequency-constrained optimization. Specifically, we extract dominant low-frequency components via Fast Fourier Transform (FFT)-guided coordinate descent, initialize sinusoidal embeddings with these components, and employ a two-speed learning schedule to preserve meaningful frequency structure during training. Our approach is model-agnostic and integrates seamlessly into existing Transformer-based architectures. Extensive experiments across diverse real-world benchmarks demonstrate consistent performance gains--particularly at long horizons--highlighting the benefits of injecting spectral priors into deep temporal models for robust and interpretable long-range forecasting. Moreover, on synthetic data, our method accurately recovers ground-truth frequencies, further validating its interpretability and effectiveness in capturing latent periodic patterns.
Authors: Hadi Keramati, Patrick Kirchen, Mohammed Hannan, Rajeev K. Jaiman
Abstract: This study presents a generative optimization framework that builds on a fine-tuned diffusion model and reward-directed sampling to generate high-performance engineering designs. The framework adopts a parametric representation of the design geometry and produces new parameter sets corresponding to designs with enhanced performance metrics. A key advantage of the reward-directed approach is its suitability for scenarios in which performance metrics rely on costly engineering simulations or surrogate models (e.g. graph-based, ensemble models, or tree-based) are non-differentiable or prohibitively expensive to differentiate. This work introduces the iterative use of a soft value function within a Markov decision process framework to achieve reward-guided decoding in the diffusion model. By incorporating soft-value guidance during both the training and inference phases, the proposed approach reduces computational and memory costs to achieve high-reward designs, even beyond the training data. Empirical results indicate that this iterative reward-directed method substantially improves the ability of the diffusion models to generate samples with reduced resistance in 3D ship hull design and enhanced hydrodynamic performance in 2D airfoil design tasks. The proposed framework generates samples that extend beyond the training data distribution, resulting in a greater 25 percent reduction in resistance for ship design and over 10 percent improvement in the lift-to-drag ratio for the 2D airfoil design. Successful integration of this model into the engineering design life cycle can enhance both designer productivity and overall design performance.
Authors: S. Parab, A. Lamelas, A. Hassan, P. Bhote
Abstract: Over 22 million Americans participate in paddling-related activities annually, contributing to a global paddlesports market valued at 2.4 billion US dollars in 2020. Despite its popularity, the sport has seen limited integration of machine learning (ML) and remains hindered by the cost of coaching and specialized equipment. This study presents a novel AI-based coaching system that uses ML models trained on motion data and delivers stroke feedback via a large language model (LLM). Participants were recruited through a collaboration with the NYU Concrete Canoe Team. Motion data were collected across two sessions, one with suboptimal form and one with corrected technique, using Apple Watches and smartphones secured in sport straps. The data underwent stroke segmentation and feature extraction. ML models, including Support Vector Classifier, Random Forest, Gradient Boosting, and Extremely Randomized Trees, were trained on both raw and engineered features. A web based interface was developed to visualize stroke quality and deliver LLM-based feedback. Across four participants, eight trials yielded 66 stroke samples. The Extremely Randomized Tree model achieved the highest performance with an F score of 0.9496 under five fold cross validation. The web interface successfully provided both quantitative metrics and qualitative feedback. Sensor placement near the wrists improved data quality. Preliminary results indicate that smartwatches and smartphones can enable low cost, accessible alternatives to traditional paddling instruction. While limited by sample size, the study demonstrates the feasibility of using consumer devices and ML to support stroke refinement and technique improvement.
Authors: Ahmed Jaheen, Sarah Elsamanody, Hamada Rizk, Moustafa Youssef
Abstract: Indoor localization plays a pivotal role in supporting a wide array of location-based services, including navigation, security, and context-aware computing within intricate indoor environments. Despite considerable advancements, deploying indoor localization systems in real-world scenarios remains challenging, largely because of non-independent and identically distributed (non-IID) data and device heterogeneity. In response, we propose SimDeep, a novel Federated Learning (FL) framework explicitly crafted to overcome these obstacles and effectively manage device heterogeneity. SimDeep incorporates a Similarity Aggregation Strategy, which aggregates client model updates based on data similarity, significantly alleviating the issues posed by non-IID data. Our experimental evaluations indicate that SimDeep achieves an impressive accuracy of 92.89%, surpassing traditional federated and centralized techniques, thus underscoring its viability for real-world deployment.
Authors: Colby Fronk, Linda Petzold
Abstract: Gradient-based optimization of neural differential equations and other parameterized dynamical systems fundamentally relies on the ability to differentiate numerical solutions with respect to model parameters. In stiff systems, it has been observed that sensitivities to parameters controlling fast-decaying modes become vanishingly small during training, leading to optimization difficulties. In this paper, we show that this vanishing gradient phenomenon is not an artifact of any particular method, but a universal feature of all A-stable and L-stable stiff numerical integration schemes. We analyze the rational stability function for general stiff integration schemes and demonstrate that the relevant parameter sensitivities, governed by the derivative of the stability function, decay to zero for large stiffness. Explicit formulas for common stiff integration schemes are provided, which illustrate the mechanism in detail. Finally, we rigorously prove that the slowest possible rate of decay for the derivative of the stability function is $O(|z|^{-1})$, revealing a fundamental limitation: all A-stable time-stepping methods inevitably suppress parameter gradients in stiff regimes, posing a significant barrier for training and parameter identification in stiff neural ODEs.
Authors: Sahil Sethi, David Chen, Michael C. Burkhart, Nipun Bhandari, Bashar Ramadan, Brett Beaulieu-Jones
Abstract: Prototype-based neural networks offer interpretable predictions by comparing inputs to learned, representative signal patterns anchored in training data. While such models have shown promise in the classification of physiological data, it remains unclear whether their prototypes capture an underlying structure that aligns with broader clinical phenotypes. We use a prototype-based deep learning model trained for multi-label ECG classification using the PTB-XL dataset. Then without modification we performed inference on the MIMIC-IV clinical database. We assess whether individual prototypes, trained solely for classification, are associated with hospital discharge diagnoses in the form of phecodes in this external population. Individual prototypes demonstrate significantly stronger and more specific associations with clinical outcomes compared to the classifier's class predictions, NLP-extracted concepts, or broader prototype classes across all phecode categories. Prototype classes with mixed significance patterns exhibit significantly greater intra-class distances (p $<$ 0.0001), indicating the model learned to differentiate clinically meaningful variations within diagnostic categories. The prototypes achieve strong predictive performance across diverse conditions, with AUCs ranging from 0.89 for atrial fibrillation to 0.91 for heart failure, while also showing substantial signal for non-cardiac conditions such as sepsis and renal disease. These findings suggest that prototype-based models can support interpretable digital phenotyping from physiologic time-series data, providing transferable intermediate phenotypes that capture clinically meaningful physiologic signatures beyond their original training objectives.
Authors: Jingyi Chen, Xinyuan Zhang, Xinwu Qian
Abstract: The Elementary Shortest-Path Problem(ESPP) seeks a minimum cost path from s to t that visits each vertex at most once. The presence of negative-cost cycles renders the problem NP-hard. We present a probabilistic method for finding near-optimal ESPP, enabled by an unsupervised graph neural network that jointly learns node value estimates and edge-selection probabilities via a surrogate loss function. The loss provides a high probability certificate of finding near-optimal ESPP solutions by simultaneously reducing negative-cost cycles and embedding the desired algorithmic alignment. At inference time, a decoding algorithm transforms the learned edge probabilities into an elementary path. Experiments on graphs of up to 100 nodes show that the proposed method surpasses both unsupervised baselines and classical heuristics, while exhibiting high performance in cross-size and cross-topology generalization on unseen synthetic graphs.
Authors: Lingyu Jiang, Yuping Wang, Yao Su, Shuo Xing, Wenjing Chen, Xin Zhang, Zhengzhong Tu, Ziming Zhang, Fangzhou Lin, Michael Zielewski, Kazunori D Yamada
Abstract: In recent years, multilayer perceptrons (MLP)-based deep learning models have demonstrated remarkable success in long-term time series forecasting (LTSF). Existing approaches typically augment MLP backbones with hand-crafted external modules to address the inherent limitations of their flat architectures. Despite their success, these augmented methods neglect hierarchical locality and sequential inductive biases essential for time-series modeling, and recent studies indicate diminishing performance improvements. To overcome these limitations, we explore Kolmogorov-Arnold Networks (KAN), a recently proposed model featuring adaptive basis functions capable of granular, local modulation of nonlinearities. This raises a fundamental question: Can KAN serve as a new modeling core for LTSF? To answer this, we introduce KANMixer, a concise architecture integrating a multi-scale mixing backbone that fully leverages KAN's adaptive capabilities. Extensive evaluation demonstrates that KANMixer achieves state-of-the-art performance in 16 out of 28 experiments across seven benchmark datasets. To uncover the reasons behind this strong performance, we systematically analyze the strengths and limitations of KANMixer in comparison with traditional MLP architectures. Our findings reveal that the adaptive flexibility of KAN's learnable basis functions significantly transforms the influence of network structural prior on forecasting performance. Furthermore, we identify critical design factors affecting forecasting accuracy and offer practical insights for effectively utilizing KAN in LTSF. Together, these insights constitute the first empirically grounded guidelines for effectively leveraging KAN in LTSF. Code is available in the supplementary file.
Authors: Heting Liu, Junzhe Huang, Fang He, Guohong Cao
Abstract: Federated Learning (FL) enables edge devices to collaboratively learn a global model, but it may not perform well when clients have high data heterogeneity. In this paper, we propose a dynamic clustering algorithm for personalized federated learning system (DC-PFL) to address the problem of data heterogeneity. DC-PFL starts with all clients training a global model and gradually groups the clients into smaller clusters for model personalization based on their data similarities. To address the challenge of estimating data heterogeneity without exposing raw data, we introduce a discrepancy metric called model discrepancy, which approximates data heterogeneity solely based on the model weights received by the server. We demonstrate that model discrepancy is strongly and positively correlated with data heterogeneity and can serve as a reliable indicator of data heterogeneity. To determine when and how to change grouping structures, we propose an algorithm based on the rapid decrease period of the training loss curve. Moreover, we propose a layer-wise aggregation mechanism that aggregates the low-discrepancy layers at a lower frequency to reduce the amount of transmitted data and communication costs. We conduct extensive experiments on various datasets to evaluate our proposed algorithm, and our results show that DC-PFL significantly reduces total training time and improves model accuracy compared to baselines.
Authors: Nguyen Cong Luong, Nguyen Duc Hai, Duc Van Le, Huy T. Nguyen, Thai-Hoc Vu, Thien Huynh-The, Ruichen Zhang, Nguyen Duc Duy Anh, Dusit Niyato, Marco Di Renzo, Dong In Kim, Quoc-Viet Pham
Abstract: The rise of Generative AI (GenAI) in recent years has catalyzed transformative advances in wireless communications and networks. Among the members of the GenAI family, Diffusion Models (DMs) have risen to prominence as a powerful option, capable of handling complex, high-dimensional data distribution, as well as consistent, noise-robust performance. In this survey, we aim to provide a comprehensive overview of the theoretical foundations and practical applications of DMs across future communication systems. We first provide an extensive tutorial of DMs and demonstrate how they can be applied to enhance optimizers, reinforcement learning and incentive mechanisms, which are popular approaches for problems in wireless networks. Then, we review and discuss the DM-based methods proposed for emerging issues in future networks and communications, including channel modeling and estimation, signal detection and data reconstruction, integrated sensing and communication, resource management in edge computing networks, semantic communications and other notable issues. We conclude the survey with highlighting technical limitations of DMs and their applications, as well as discussing future research directions.
Authors: Euihyun Kim, Keun Park, Yeoneung Kim
Abstract: Recent advances in denoising diffusion models have enabled rapid generation of optimized structures for topology optimization. However, these models often rely on surrogate predictors to enforce physical constraints, which may fail to capture subtle yet critical design flaws such as floating components or boundary discontinuities that are obvious to human experts. In this work, we propose a novel human-in-the-loop diffusion framework that steers the generative process using a lightweight reward model trained on minimal human feedback. Inspired by preference alignment techniques in generative modeling, our method learns to suppress unrealistic outputs by modulating the reverse diffusion trajectory using gradients of human-aligned rewards. Specifically, we collect binary human evaluations of generated topologies and train classifiers to detect floating material and boundary violations. These reward models are then integrated into the sampling loop of a pre-trained diffusion generator, guiding it to produce designs that are not only structurally performant but also physically plausible and manufacturable. Our approach is modular and requires no retraining of the diffusion model. Preliminary results show substantial reductions in failure modes and improved design realism across diverse test conditions. This work bridges the gap between automated design generation and expert judgment, offering a scalable solution to trustworthy generative design.
Authors: Juyan Zhang, Rhys Newbury, Xinyang Zhang, Tin Tran, Dana Kulic, Michael Burke
Abstract: Score matching enables the estimation of the gradient of a data distribution, a key component in denoising diffusion models used to recover clean data from corrupted inputs. In prior work, a heuristic weighting function has been used for the denoising score matching loss without formal justification. In this work, we demonstrate that heteroskedasticity is an inherent property of the denoising score matching objective. This insight leads to a principled derivation of optimal weighting functions for generalized, arbitrary-order denoising score matching losses, without requiring assumptions about the noise distribution. Among these, the first-order formulation is especially relevant to diffusion models. We show that the widely used heuristical weighting function arises as a first-order Taylor approximation to the trace of the expected optimal weighting. We further provide theoretical and empirical comparisons, revealing that the heuristical weighting, despite its simplicity, can achieve lower variance than the optimal weighting with respect to parameter gradients, which can facilitate more stable and efficient training.
Authors: En Yu, Jie Lu, Kun Wang, Xiaoyu Yang, Guangquan Zhang
Abstract: Learning from multiple data streams in real-world scenarios is fundamentally challenging due to intrinsic heterogeneity and unpredictable concept drifts. Existing methods typically assume homogeneous streams and employ static architectures with indiscriminate knowledge fusion, limiting generalizability in complex dynamic environments. To tackle this gap, we propose CAMEL, a dynamic \textbf{C}ollaborative \textbf{A}ssistance \textbf{M}ixture of \textbf{E}xperts \textbf{L}earning framework. It addresses heterogeneity by assigning each stream an independent system with a dedicated feature extractor and task-specific head. Meanwhile, a dynamic pool of specialized private experts captures stream-specific idiosyncratic patterns. Crucially, collaboration across these heterogeneous streams is enabled by a dedicated assistance expert. This expert employs a multi-head attention mechanism to distill and integrate relevant context autonomously from all other concurrent streams. It facilitates targeted knowledge transfer while inherently mitigating negative transfer from irrelevant sources. Furthermore, we propose an Autonomous Expert Tuner (AET) strategy, which dynamically manages expert lifecycles in response to drift. It instantiates new experts for emerging concepts (freezing prior ones to prevent catastrophic forgetting) and prunes obsolete ones. This expert-level plasticity provides a robust and efficient mechanism for online model capacity adaptation. Extensive experiments demonstrate CAMEL's superior generalizability across diverse multistreams and exceptional resilience against complex concept drifts.
Authors: Xinhan Di, JoyJiaoW
Abstract: Reinforcement learning scaling enhances the reasoning capabilities of large language models, with reinforcement learning serving as the key technique to draw out complex reasoning. However, key technical details of state-of-the-art reasoning LLMs, such as those in the OpenAI O series, Claude 3 series, DeepMind's Gemini 2.5 series, and Grok 3 series, remain undisclosed, making it difficult for the research community to replicate their reinforcement learning training results. Therefore, we start our study from an Early Preview Reinforcement Learning (EPRLI) algorithm built on the open-source GRPO framework, incorporating difficulty-aware intervention for math problems. Applied to a 1.5B-parameter LLM, our method achieves 50.0% on AIME24, 89.2% on Math500, 77.1% on AMC, 35.3% on Minerva, and 51.9% on OBench, superpass O1-Preview and is comparable to O1-mini within standard school-lab settings.
Authors: Sandesh Kumar Singh
Abstract: This work proposes a novel technique Augmented Reinforcement Learning framework for the improvement of decision-making capabilities of machine learning models. The introduction of agents as external overseers checks on model decisions. The external agent can be anyone, like humans or automated scripts, that helps in decision path correction. It seeks to ascertain the priority of the "Garbage-In, Garbage-Out" problem that caused poor data inputs or incorrect actions in reinforcement learning. The ARL framework incorporates two external agents that aid in course correction and the guarantee of quality data at all points of the training cycle. The External Agent 1 is a real-time evaluator, which will provide feedback light of decisions taken by the model, identify suboptimal actions forming the Rejected Data Pipeline. The External Agent 2 helps in selective curation of the provided feedback with relevance and accuracy in business scenarios creates an approved dataset for future training cycles. The validation of the framework is also applied to a real-world scenario, which is "Document Identification and Information Extraction". This problem originates mainly from banking systems, but can be extended anywhere. The method of classification and extraction of information has to be done correctly here. Experimental results show that including human feedback significantly enhances the ability of the model in order to increase robustness and accuracy in making decisions. The augmented approach, with a combination of machine efficiency and human insight, attains a higher learning standard-mainly in complex or ambiguous environments. The findings of this study show that human-in-the-loop reinforcement learning frameworks such as ARL can provide a scalable approach to improving model performance in data-driven applications.
Authors: Yandong Yan, Chenxi Li, Yu Huang, Dexuan Xu, Jiaqi Zhu, Zhongyan Chai, Huamin Zhang
Abstract: The scarcity of large-scale and high-quality electronic health records (EHRs) remains a major bottleneck in biomedical research, especially as large foundation models become increasingly data-hungry. Synthesizing substantial volumes of de-identified and high-fidelity data from existing datasets has emerged as a promising solution. However, existing methods suffer from a series of limitations: they struggle to model the intrinsic properties of heterogeneous multimodal EHR data (e.g., continuous, discrete, and textual modalities), capture the complex dependencies among them, and robustly handle pervasive data incompleteness. These challenges are particularly acute in Traditional Chinese Medicine (TCM). To this end, we propose TCDiff (Triplex Cascaded Diffusion Network), a novel EHR generation framework that cascades three diffusion networks to learn the features of real-world EHR data, formatting a multi-stage generative process: Reference Modalities Diffusion, Cross-Modal Bridging, and Target Modality Diffusion. Furthermore, to validate our proposed framework, besides two public datasets, we also construct and introduce TCM-SZ1, a novel multimodal EHR dataset for benchmarking. Experimental results show that TCDiff consistently outperforms state-of-the-art baselines by an average of 10% in data fidelity under various missing rate, while maintaining competitive privacy guarantees. This highlights the effectiveness, robustness, and generalizability of our approach in real-world healthcare scenarios.
Authors: Xindi Fan, Jing Wu, Mingyi Zhou, Pengwei Liang, Dinh Phung
Abstract: Recent studies have shown that deep learning models are vulnerable to attacks and tend to memorize training data points, raising significant concerns about privacy leakage. This motivates the development of machine unlearning (MU), i.e., a paradigm that enables models to selectively forget specific data points upon request. However, most existing MU algorithms require partial or full fine-tuning on the retain set. This necessitates continued access to the original training data, which is often impractical due to privacy concerns and storage constraints. A few retain-data-free MU methods have been proposed, but some rely on access to auxiliary data and precomputed statistics of the retain set, while others scale poorly when forgetting larger portions of data. In this paper, we propose Influence-guided Machine Unlearning (IMU), a simple yet effective method that conducts MU using only the forget set. Specifically, IMU employs gradient ascent and innovatively introduces dynamic allocation of unlearning intensities across different data points based on their influences. This adaptive strategy significantly enhances unlearning effectiveness while maintaining model utility. Results across vision and language tasks demonstrate that IMU consistently outperforms existing retain-data-free MU methods.
Authors: Yuanteng Chen, Yuantian Shao, Peisong Wang, Jian Cheng
Abstract: Mixture-of-Experts (MoE) has demonstrated promising potential in scaling LLMs. However, it is hindered by two critical challenges: (1) substantial GPU memory consumption to load all experts; (2) low activated parameters cannot be equivalently translated into inference acceleration effects. In this work, we propose EAC-MoE, an Expert-Selection Aware Compressor for MoE-LLMs, which deeply aligns with the characteristics of MoE from the perspectives of quantization and pruning, and introduces two modules to address these two challenges respectively: (1) The expert selection bias caused by low-bit quantization is a major factor contributing to the performance degradation in MoE-LLMs. Based on this, we propose Quantization with Expert-Selection Calibration (QESC), which mitigates the expert selection bias by calibrating the routers within the MoE; (2) There are always certain experts that are not crucial for the corresponding tasks, yet causing inference latency. Therefore, we propose Pruning based on Expert-Selection Frequency (PESF), which significantly improves inference speed by pruning less frequently used experts for current task. Extensive experiments demonstrate that our approach significantly reduces memory usage and improves inference speed with minimal performance degradation.
Authors: Wenzhuo Qian, Hailiang Zhao, Tianlv Chen, Jiayi Chen, Ziqi Wang, Kingsum Chow, Shuiguang Deng
Abstract: Microservice architectures have become the de facto standard for building scalable cloud-native applications, yet their distributed nature introduces significant challenges in performance monitoring and resource management. Traditional approaches often rely on per-request latency metrics, which are highly sensitive to transient noise and fail to reflect the holistic behavior of complex, concurrent workloads. In contrast, window-level P95 tail latency provides a stable and meaningful signal that captures both system-wide trends and user-perceived performance degradation. We identify two key shortcomings in existing methods: (i) inadequate handling of heterogeneous data, where traffic-side features propagate across service dependencies and resource-side signals reflect localized bottlenecks, and (ii) the lack of principled architectural designs that effectively distinguish and integrate these complementary modalities. To address these challenges, we propose USRFNet, a deep learning network that explicitly separates and models traffic-side and resource-side features. USRFNet employs GNNs to capture service interactions and request propagation patterns, while gMLP modules independently model cluster resource dynamics. These representations are then fused into a unified system embedding to predict window-level P95 latency with high accuracy. We evaluate USRFNet on real-world microservice benchmarks under large-scale stress testing conditions, demonstrating substantial improvements in prediction accuracy over state-of-the-art baselines.
Authors: Tianpei Lu, Bingsheng Zhang, Lekun Peng, Bowen Zheng, Lichun Li, Kui Ren
Abstract: With the increasing deployment of generative machine learning models in privacy-sensitive domains such as healthcare and personalized services, ensuring secure inference has become a critical challenge. Secure multi-party computation (MPC) enables privacy-preserving model inference but suffers from high communication and computation overhead. The main bottleneck lies in the expensive secure evaluation of floating-point operations. Quantization offers a promising solution by converting floating-point operations into lower-precision integer computations, significantly reducing overhead. However, existing MPC-based quantized inference methods either rely on public quantization parameters-posing privacy risks-or suffer from inefficiencies, particularly in handling nonlinear functions such as activations and softmax. In this work, we propose a fine-grained, layer-wise quantization scheme and support 1-bit weight fully connected layers in a secure setting. We design a multi-input lookup table protocol to evaluate softmax efficiently and securely. Furthermore, we use dual secret sharing schemes and perform precision conversions via lookup tables, eliminating truncation overhead entirely. Experimental evaluation on BERT-base models demonstrates that our approach achieves up to $8\times$ speedup compared to Lu \emph{et al}. (NDSS 25), $9\times$ speedup compared to Gupta \emph{et al}. (PETS 24) and $22 \times$ speedup compared to Knott \emph{et al}. (NeurIPS 21).
Authors: Minsuk Jang, Changick Kim
Abstract: Current Spiking Neural Networks (SNNs) underutilize the temporal dynamics inherent in spike-based processing, relying primarily on rate coding while overlooking precise timing information that provides rich computational cues. We propose SPARTA (Spiking Priority Attention with Resource-Adaptive Temporal Allocation), a framework that leverages heterogeneous neuron dynamics and spike-timing information to enable efficient sparse attention. SPARTA prioritizes tokens based on temporal cues, including firing patterns, spike timing, and inter-spike intervals, achieving 65.4% sparsity through competitive gating. By selecting only the most salient tokens, SPARTA reduces attention complexity from O(N^2) to O(K^2) with k << n, while maintaining high accuracy. Our method achieves state-of-the-art performance on DVS-Gesture (98.78%) and competitive results on CIFAR10-DVS (83.06%) and CIFAR-10 (95.3%), demonstrating that exploiting spike timing dynamics improves both computational efficiency and accuracy.
Authors: Ziru Niu, Hai Dong, A. K. Qin
Abstract: Federated learning (FL) is a pioneering machine learning paradigm that enables distributed clients to process local data effectively while ensuring data privacy. However, the efficacy of FL is usually impeded by the data heterogeneity among clients, resulting in local models with low generalization performance. To address this problem, traditional model-homogeneous approaches mainly involve debiasing the local training procedures with regularization or dynamically adjusting client weights in aggregation. Nonetheless, these approaches become incompatible for scenarios where clients exhibit heterogeneous model architectures. In this paper, we propose a model-heterogeneous FL framework that can improve clients' generalization performance over unseen data without model aggregation. Instead of model parameters, clients exchange the feature distributions with the server, including the mean and the covariance. Accordingly, clients train a variational transposed convolutional (VTC) neural network with Gaussian latent variables sampled from the feature distributions, and use the VTC model to generate synthetic data. By fine-tuning local models with the synthetic data, clients significantly increase their generalization performance. Experimental results show that our approach obtains higher generalization accuracy than existing model-heterogeneous FL frameworks, as well as lower communication costs and memory consumption
Authors: Ali Forootani, Raffaele Iervolino
Abstract: Federated Learning (FL) enables collaborative model training across decentralized devices while preserving data privacy. However, traditional FL suffers from communication overhead, system heterogeneity, and straggler effects. Asynchronous Federated Learning (AFL) addresses these by allowing clients to update independently, improving scalability and reducing synchronization delays. This paper extends AFL to handle non-convex objective functions and heterogeneous datasets, common in modern deep learning. We present a rigorous convergence analysis, deriving bounds on the expected gradient norm and studying the effects of staleness, variance, and heterogeneity. To mitigate stale updates, we introduce a staleness aware aggregation that prioritizes fresher updates and a dynamic learning rate schedule that adapts to client staleness and heterogeneity, improving stability and convergence. Our framework accommodates variations in computational power, data distribution, and communication delays, making it practical for real world applications. We also analyze the impact of client selection strategies-sampling with or without replacement-on variance and convergence. Implemented in PyTorch with Python's asyncio, our approach is validated through experiments demonstrating improved performance and scalability for asynchronous, heterogeneous, and non-convex FL scenarios.
Authors: Alberto Maria Metelli, Simone Drago, Marco Mussi
Abstract: We study the regret minimization problem in the novel setting of generalized kernelized bandits (GKBs), where we optimize an unknown function $f^*$ belonging to a reproducing kernel Hilbert space (RKHS) having access to samples generated by an exponential family (EF) noise model whose mean is a non-linear function $\mu(f^*)$. This model extends both kernelized bandits (KBs) and generalized linear bandits (GLBs). We propose an optimistic algorithm, GKB-UCB, and we explain why existing self-normalized concentration inequalities do not allow to provide tight regret guarantees. For this reason, we devise a novel self-normalized Bernstein-like dimension-free inequality resorting to Freedman's inequality and a stitching argument, which represents a contribution of independent interest. Based on it, we conduct a regret analysis of GKB-UCB, deriving a regret bound of order $\widetilde{O}( \gamma_T \sqrt{T/\kappa_*})$, being $T$ the learning horizon, ${\gamma}_T$ the maximal information gain, and $\kappa_*$ a term characterizing the magnitude the reward nonlinearity. Our result matches, up to multiplicative constants and logarithmic terms, the state-of-the-art bounds for both KBs and GLBs and provides a unified view of both settings.
Authors: Kayvan Karim, Hani Ragab Hassen. Hadj Batatia
Abstract: Data representation remains a fundamental challenge in machine learning, particularly when adapting sequence-based architectures like Transformers and Large Language Models (LLMs) for structured tabular data. Existing methods often fail to cohesively encode the mix of numerical and categorical features or preserve the inherent structure of tables. This paper introduces a novel, hybrid tokenisation methodology designed to convert tabular data into a unified, sequential format suitable for LLM training. Our approach combines predefined fixed tokens to represent structural elements and low-cardinality categorical features, with a learned subword vocabulary using Byte-Pair Encoding (BPE) for high-cardinality and continuous values. We demonstrate the efficacy of this technique by applying it to a large-scale NetFlow dataset (CIDDS-001), preparing a corpus for a Network Intrusion Detection System (NIDS) foundation model. The evaluation shows that our method is highly efficient, processing over 31 million network flows in under five hours and achieving a significant data compression ratio of 6.18:1. This process resulted in a computationally manageable corpus of over one billion tokens, establishing a viable and generalisable pathway for training foundation models on structured data.
Authors: Maciej Mozolewski, Szymon Bobek, Grzegorz J. Nalepa
Abstract: Explaining machine learning (ML) models for time series (TS) classification remains challenging due to the difficulty of interpreting raw time series and the high dimensionality of the input space. We introduce PHAR-Post-hoc Attribution Rules-a unified framework that transforms numeric feature attributions from post-hoc, instance-wise explainers (e.g., LIME, SHAP) into structured, human-readable rules. These rules define interpretable intervals that indicate where and when key decision boundaries occur, enhancing model transparency. PHAR performs comparably to native rule-based methods, such as Anchor, while scaling more efficiently to long TS sequences and achieving broader instance coverage. A dedicated rule fusion step consolidates rule sets using strategies like weighted selection and lasso-based refinement, balancing key quality metrics: coverage, confidence, and simplicity. This fusion ensures each instance receives a concise and unambiguous rule, improving both explanation fidelity and consistency. We further introduce visualization techniques to illustrate specificity-generalization trade-offs in the derived rules. PHAR resolves conflicting and overlapping explanations-a common effect of the Rashomon phenomenon-into coherent, domain-adaptable insights. Comprehensive experiments on UCI datasets demonstrate that PHAR improves interpretability, decision transparency, and practical applicability for TS classification tasks.
Authors: Asmit Bandyopadhyay, Rohit Basu, Tanmay Sen, Swagatam Das
Abstract: Human Activity Recognition (HAR) plays a vital role in applications such as fitness tracking, smart homes, and healthcare monitoring. Traditional HAR systems often rely on single modalities, such as motion sensors or cameras, limiting robustness and accuracy in real-world environments. This work presents FedTime-MAGNET, a novel multimodal federated learning framework that advances HAR by combining heterogeneous data sources: depth cameras, pressure mats, and accelerometers. At its core is the Multimodal Adaptive Graph Neural Expert Transformer (MAGNET), a fusion architecture that uses graph attention and a Mixture of Experts to generate unified, discriminative embeddings across modalities. To capture complex temporal dependencies, a lightweight T5 encoder only architecture is customized and adapted within this framework. Extensive experiments show that FedTime-MAGNET significantly improves HAR performance, achieving a centralized F1 Score of 0.934 and a strong federated F1 Score of 0.881. These results demonstrate the effectiveness of combining multimodal fusion, time series LLMs, and federated learning for building accurate and robust HAR systems.
Authors: Yeongjong Kim, Yeoneung Kim, Minseok Kim, Namkyeong Cho
Abstract: We propose a physics-informed neural network policy iteration (PINN-PI) framework for solving stochastic optimal control problems governed by second-order Hamilton--Jacobi--Bellman (HJB) equations. At each iteration, a neural network is trained to approximate the value function by minimizing the residual of a linear PDE induced by a fixed policy. This linear structure enables systematic $L^2$ error control at each policy evaluation step, and allows us to derive explicit Lipschitz-type bounds that quantify how value gradient errors propagate to the policy updates. This interpretability provides a theoretical basis for evaluating policy quality during training. Our method extends recent deterministic PINN-based approaches to stochastic settings, inheriting the global exponential convergence guarantees of classical policy iteration under mild conditions. We demonstrate the effectiveness of our method on several benchmark problems, including stochastic cartpole, pendulum problems and high-dimensional linear quadratic regulation (LQR) problems in up to 10D.
Authors: Xin Ding, Yun Chen, Yongwei Wang, Kao Zhang, Sen Zhang, Peibei Cao, Xiangxue Wang
Abstract: Recent advances in conditional generative modeling have introduced Continuous conditional Generative Adversarial Network (CcGAN) and Continuous Conditional Diffusion Model (CCDM) for estimating high-dimensional data distributions conditioned on scalar, continuous regression labels (e.g., angles, ages, or temperatures). However, these approaches face fundamental limitations: CcGAN suffers from data imbalance due to fixed-size vicinity constraints, while CCDM requires computationally expensive iterative sampling. We present CcGAN-AVAR, an enhanced CcGAN framework that addresses both challenges: (1) leveraging the GAN framework's native one-step generation to overcome CCDMs' sampling bottleneck (achieving 300x-2000x faster inference), while (2) two novel components specifically target data imbalance - an adaptive vicinity mechanism that dynamically adjusts vicinity's size, and a multi-task discriminator that constructs two regularization terms (through auxiliary regression and density ratio estimation) to significantly improve generator training. Extensive experiments on four benchmark datasets (64x64 to 192x192 resolution) across eight challenging imbalanced settings demonstrate that CcGAN-AVAR achieves state-of-the-art generation quality while maintaining sampling efficiency.
Authors: Sisuo Lyu, Siru Zhong, Weilin Ruan, Qingxiang Liu, Qingsong Wen, Hui Xiong, Yuxuan Liang
Abstract: Time series forecasting is fundamental to diverse applications, with recent approaches leverage large vision models (LVMs) to capture temporal patterns through visual representations. We reveal that while vision models enhance forecasting performance, 99% of their parameters are unnecessary for time series tasks. Through cross-modal analysis, we find that time series align with low-level textural features but not high-level semantics, which can impair forecasting accuracy. We propose OccamVTS, a knowledge distillation framework that extracts only the essential 1% of predictive information from LVMs into lightweight networks. Using pre-trained LVMs as privileged teachers, OccamVTS employs pyramid-style feature alignment combined with correlation and feature distillation to transfer beneficial patterns while filtering out semantic noise. Counterintuitively, this aggressive parameter reduction improves accuracy by eliminating overfitting to irrelevant visual features while preserving essential temporal patterns. Extensive experiments across multiple benchmark datasets demonstrate that OccamVTS consistently achieves state-of-the-art performance with only 1% of the original parameters, particularly excelling in few-shot and zero-shot scenarios.
Authors: Zicong Ye, Kunming Zhang, Guoming Tang
Abstract: The explosive growth of interactive Large Language Models (LLMs) has placed unprecedented demands for low latency on cloud GPUs, forcing them into high-power modes and causing escalating energy costs. Real-time inference workloads exhibit significant dynamic volatility, presenting substantial energy-saving opportunities. However, traditional static or rule-based power management strategies struggle to exploit these opportunities without compromising peak performance. To address this challenge, we propose AGFT (An Adaptive GPU Frequency Tuner), a framework that employs online reinforcement learning to autonomously learn an optimal frequency tuning policy. By monitoring real-time features like request load and latency, AGFT utilizes fine-grained frequency control for precise adjustments and intelligent action space pruning for stable, efficient decision-making. This creates a robust, automated energy management solution. We comprehensively evaluated AGFT in an environment simulating realistic, fluctuating inference requests. The experimental results demonstrate that AGFT successfully saves 44.3% of GPU energy consumption while introducing a minimal performance latency overhead of under 10%. This achievement translates into a comprehensive Energy-Delay Product (EDP) optimization of up to 40.3%, clearly showing that our framework can significantly enhance the energy efficiency and economic benefits of existing LLM inference clusters without compromising service quality.
Authors: Xiangwang Hou, Jingjing Wang, Fangming Guan, Jun Du, Chunxiao Jiang, Yong Ren
Abstract: Emerging real-time computer vision (CV) applications on wireless edge devices demand energy-efficient and privacy-preserving learning. Federated learning (FL) enables on-device training without raw data sharing, yet remains challenging in resource-constrained environments due to energy-intensive computation and communication, as well as limited and non-i.i.d. local data. We propose FedDPQ, an ultra energy-efficient FL framework for real-time CV over unreliable wireless networks. FedDPQ integrates diffusion-based data augmentation, model pruning, communication quantization, and transmission power control to enhance training efficiency. It expands local datasets using synthetic data, reduces computation through pruning, compresses updates via quantization, and mitigates transmission outages with adaptive power control. We further derive a closed-form energy-convergence model capturing the coupled impact of these components, and develop a Bayesian optimization(BO)-based algorithm to jointly tune data augmentation strategy, pruning ratio, quantization level, and power control. To the best of our knowledge, this is the first work to jointly optimize FL performance from the perspectives of data, computation, and communication under unreliable wireless conditions. Experiments on representative CV tasks show that FedDPQ achieves superior convergence speed and energy efficiency.
Authors: Rui Ding, Hanyang Meng, Zeyang Zhang, Jielong Yang
Abstract: Diffusion models have demonstrated strong performance in time series forecasting, yet often suffer from semantic misalignment between generated trajectories and conditioning covariates, especially under complex or multimodal conditions. To address this issue, we propose SemGuide, a plug-and-play, inference-time method that enhances covariate consistency in conditional diffusion models. Our approach introduces a scoring network to assess the semantic alignment between intermediate diffusion states and future covariates. These scores serve as proxy likelihoods in a stepwise importance reweighting procedure, which progressively adjusts the sampling path without altering the original training process. The method is model-agnostic and compatible with any conditional diffusion framework. Experiments on real-world forecasting tasks show consistent gains in both predictive accuracy and covariate alignment, with especially strong performance under complex conditioning scenarios.
Authors: Ruiqi Wang, Diego Klabjan
Abstract: The concept of learning to optimize involves utilizing a trainable optimization strategy rather than relying on manually defined full gradient estimations such as ADAM. We present a framework that jointly trains the full gradient estimator and the trainable weights of the model. Specifically, we prove that pseudo-linear TO (Trainable Optimizer), a linear approximation of the full gradient, matches SGD's convergence rate while effectively reducing variance. Pseudo-linear TO incurs negligible computational overhead, requiring only minimal additional tensor multiplications. To further improve computational efficiency, we introduce two simplified variants of Pseudo-linear TO. Experiments demonstrate that TO methods converge faster than benchmark algorithms (e.g., ADAM) in both strongly convex and non-convex settings, and fine tuning of an LLM.
Authors: Shiyan Liu, Bohan Tan, Yan Jin
Abstract: The routing problems such as the Traveling Salesman Problem (TSP) and the Capacitated Vehicle Routing Problem (CVRP) are well-known combinatorial optimization challenges with broad practical relevance. Recent data-driven optimization methods have made significant progress, yet they often face limitations in training efficiency and generalization to large-scale instances. In this paper, we propose a novel Vision-Augmented Asymmetric Group Preference Optimization (VAGPO) approach for solving the routing problems. By leveraging ResNet-based visual encoding and Transformer-based sequential modeling, VAGPO captures both spatial structure and temporal dependencies. Furthermore, we introduce an asymmetric group preference optimization strategy that significantly accelerates convergence compared to commonly used policy gradient methods. Experimental results on TSP and CVRP benchmarks show that the proposed VAGPO not only achieves highly competitive solution quality but also exhibits strong generalization to larger instances (up to 1000 nodes) without re-training, highlighting its effectiveness in both learning efficiency and scalability.
Authors: Ignacy St\k{e}pka, Nicholas Gisolfi, Kacper Tr\k{e}bacz, Artur Dubrawski
Abstract: We consider the problem of persistent client dropout in asynchronous Decentralized Federated Learning (DFL). Asynchronicity and decentralization obfuscate information about model updates among federation peers, making recovery from a client dropout difficult. Access to the number of learning epochs, data distributions, and all the information necessary to precisely reconstruct the missing neighbor's loss functions is limited. We show that obvious mitigations do not adequately address the problem and introduce adaptive strategies based on client reconstruction. We show that these strategies can effectively recover some performance loss caused by dropout. Our work focuses on asynchronous DFL with local regularization and differs substantially from that in the existing literature. We evaluate the proposed methods on tabular and image datasets, involve three DFL algorithms, and three data heterogeneity scenarios (iid, non-iid, class-focused non-iid). Our experiments show that the proposed adaptive strategies can be effective in maintaining robustness of federated learning, even if they do not reconstruct the missing client's data precisely. We also discuss the limitations and identify future avenues for tackling the problem of client dropout.
Authors: Haoran Li, Muhao Guo, Yang Weng, Hanghang Tong
Abstract: Deep sequence models have achieved notable success in time-series analysis, such as interpolation and forecasting. Recent advances move beyond discrete-time architectures like Recurrent Neural Networks (RNNs) toward continuous-time formulations such as the family of Neural Ordinary Differential Equations (Neural ODEs). Generally, they have shown that capturing the underlying dynamics is beneficial for generic tasks like interpolation, extrapolation, and classification. However, existing methods approximate the dynamics using unconstrained neural networks, which struggle to adapt reliably under distributional shifts. In this paper, we recast time-series problems as the continuous ODE-based optimal control problem. Rather than learning dynamics solely from data, we optimize control actions that steer ODE trajectories toward task objectives, bringing control-theoretical performance guarantees. To achieve this goal, we need to (1) design the appropriate control actions and (2) apply effective optimal control algorithms. As the actions should contain rich context information, we propose to employ the discrete-time model to process past sequences and generate actions, leading to a coordinate model to extract long-term temporal features to modulate short-term continuous dynamics. During training, we apply model predictive control to plan multi-step future trajectories, minimize a task-specific cost, and greedily select the optimal current action. We show that, under mild assumptions, this multi-horizon optimization leads to exponential convergence to infinite-horizon solutions, indicating that the coordinate model can gain robust and generalizable performance. Extensive experiments on diverse time-series datasets validate our method's superior generalization and adaptability compared to state-of-the-art baselines.
Authors: Gian Marco Paldino, Gianluca Bontempi
Abstract: Discovering causal relationships in complex multivariate time series is a fundamental scientific challenge. Traditional methods often falter, either by relying on restrictive linear assumptions or on conditional independence tests that become uninformative in the presence of intricate, non-linear dynamics. This paper proposes a new paradigm, shifting from statistical testing to pattern recognition. We hypothesize that a causal link creates a persistent and learnable asymmetry in the flow of information through a system's temporal graph, even when clear conditional independencies are obscured. We introduce Temporal Dependency to Causality (TD2C), a supervised learning framework that operationalizes this hypothesis. TD2C learns to recognize these complex causal signatures from a rich set of information-theoretic and statistical descriptors. Trained exclusively on a diverse collection of synthetic time series, TD2C demonstrates remarkable zero-shot generalization to unseen dynamics and established, realistic benchmarks. Our results show that TD2C achieves state-of-the-art performance, consistently outperforming established methods, particularly in high-dimensional and non-linear settings. By reframing the discovery problem, our work provides a robust and scalable new tool for uncovering causal structures in complex systems.
Authors: Ning Yang, Pengyu Wang, Guoqing Liu, Haifeng Zhang, Pin Lyu, Jun Wang
Abstract: Safe Reinforcement Learning (RL) often faces significant issues such as constraint violations and instability, necessitating the use of constrained policy optimization, which seeks optimal policies while ensuring adherence to specific constraints like safety. Typically, constrained optimization problems are addressed by the Lagrangian method, a post-violation remedial approach that may result in oscillations and overshoots. Motivated by this, we propose a novel method named Proactive Constrained Policy Optimization (PCPO) that incorporates a preemptive penalty mechanism. This mechanism integrates barrier items into the objective function as the policy nears the boundary, imposing a cost. Meanwhile, we introduce a constraint-aware intrinsic reward to guide boundary-aware exploration, which is activated only when the policy approaches the constraint boundary. We establish theoretical upper and lower bounds for the duality gap and the performance of the PCPO update, shedding light on the method's convergence characteristics. Additionally, to enhance the optimization performance, we adopt a policy iteration approach. An interesting finding is that PCPO demonstrates significant stability in experiments. Experimental results indicate that the PCPO framework provides a robust solution for policy optimization under constraints, with important implications for future research and practical applications.
Authors: Navneet Verma, Ying Xie
Abstract: The increasing penetration of renewable energy sources in day-ahead energy markets introduces challenges in balancing supply and demand, ensuring grid resilience, and maintaining trust in decentralized trading systems. This paper proposes a novel framework that integrates the Proximal Policy Optimization (PPO) algorithm, a state-of-the-art reinforcement learning method, with blockchain technology to optimize automated trading strategies for prosumers in day-ahead energy markets. We introduce a comprehensive framework that employs RL agent for multi-objective energy optimization and blockchain for tamper-proof data and transaction management. Simulations using real-world data from the Electricity Reliability Council of Texas (ERCOT) demonstrate the effectiveness of our approach. The RL agent achieves demand-supply balancing within 2\% and maintains near-optimal supply costs for the majority of the operating hours. Moreover, it generates robust battery storage policies capable of handling variability in solar and wind generation. All decisions are recorded on an Algorand-based blockchain, ensuring transparency, auditability, and security - key enablers for trustworthy multi-agent energy trading. Our contributions include a novel system architecture, curriculum learning for robust agent development, and actionable policy insights for practical deployment.
Authors: Jianshu She, Xinyue Li, Eric Xing, Zhengzhong Liu, Qirong Ho
Abstract: Language models can be steered by modifying their internal representations to control concepts such as emotion, style, or truthfulness in generation. However, the conditions for an effective intervention remain unclear and are often validated through heuristics and trial-and-error. To fill this gap, we demonstrate that intervention efficacy, measured by linear steerability (i.e., the ability to adjust output via linear transformations of hidden states), emerges during intermediate stages of training. Moreover, even closely related concepts (e.g., anger and sadness) exhibit steerability emergence at distinct stages of training. To better interpret the dynamics of steerability during training, we adapt existing intervention techniques into a unified framework, referred to as the "Intervention Detector" (ID), which is designed to reveal how linear steerability evolves over the course of training through hidden state and representation analysis. ID reveals that concepts become increasingly linearly separable in the hidden space as training progresses, which strongly correlates with the emergence of linear steerability. We further introduce ID-based metrics, such as heatmaps, entropy trends, and cosine similarity, to help interpret how linear steerability evolves throughout training. In addition, we apply ID across different model families to ensure the generality of our findings on steerability dynamics.
Authors: Istabrak Abbes, Gopeshh Subbaraj, Matthew Riemer, Nizar Islah, Benjamin Therien, Tsuguchika Tabaru, Hiroaki Kingetsu, Sarath Chandar, Irina Rish
Abstract: Training large language models (LLMs) typically involves pre-training on massive corpora, only to restart the process entirely when new data becomes available. A more efficient and resource-conserving approach would be continual pre-training, where models are updated with new data rather than retraining from scratch. However, the introduction of new data often causes distribution shifts, leading to performance degradation on previously learned tasks. In this paper, we take a deeper look at two popular proposals for addressing this distribution shift within the continual learning literature: experience replay and gradient alignment. We consider continual pre-training of models within the Llama family of architectures at a large scale across languages with 100 billion tokens of training data in each language, finding that both replay and gradient alignment lead to more stable learning without forgetting. This conclusion holds both as we vary the model scale and as we vary the number and diversity of tasks. Moreover, we are the first to demonstrate the effectiveness of gradient alignment techniques in the context of LLM pre-training and propose an efficient implementation of meta-experience replay (MER) that imbues experience replay with the benefits of gradient alignment despite negligible compute and memory overhead. Our scaling analysis across model sizes and replay rates indicates that small rates of replaying old examples are definitely a more valuable use of compute than investing in model size, but that it is more compute efficient to scale the size of the model than invest in high rates of replaying old examples.
Authors: Xinting Huang, Michael Hahn
Abstract: Understanding internal representations of neural models is a core interest of mechanistic interpretability. Due to its large dimensionality, the representation space can encode various aspects about inputs. To what extent are different aspects organized and encoded in separate subspaces? Is it possible to find these ``natural'' subspaces in a purely unsupervised way? Somewhat surprisingly, we can indeed achieve this and find interpretable subspaces by a seemingly unrelated training objective. Our method, neighbor distance minimization (NDM), learns non-basis-aligned subspaces in an unsupervised manner. Qualitative analysis shows subspaces are interpretable in many cases, and encoded information in obtained subspaces tends to share the same abstract concept across different inputs, making such subspaces similar to ``variables'' used by the model. We also conduct quantitative experiments using known circuits in GPT-2; results show a strong connection between subspaces and circuit variables. We also provide evidence showing scalability to 2B models by finding separate subspaces mediating context and parametric knowledge routing. Viewed more broadly, our findings offer a new perspective on understanding model internals and building circuits.
Authors: Zhuomin Chen, Jingchao Ni, Hojat Allah Salehi, Xu Zheng, Dongsheng Luo
Abstract: Graph Neural Networks (GNNs) have achieved remarkable performance in a wide range of graph-related learning tasks. However, explaining their predictions remains a challenging problem, especially due to the mismatch between the graphs used during training and those encountered during explanation. Most existing methods optimize soft edge masks on weighted graphs to highlight important substructures, but these graphs differ from the unweighted graphs on which GNNs are trained. This distributional shift leads to unreliable gradients and degraded explanation quality, especially when generating small, sparse subgraphs. To address this issue, we propose a novel iterative explanation framework which improves explanation robustness by aligning the model's training data distribution with the weighted graph distribution appeared during explanation. Our method alternates between two phases: explanation subgraph identification and model adaptation. It begins with a relatively large explanation subgraph where soft mask optimization is reliable. Based on this subgraph, we assign importance-aware edge weights to explanatory and non-explanatory edges, and retrain the GNN on these weighted graphs. This process is repeated with progressively smaller subgraphs, forming an iterative refinement procedure. We evaluate our method on multiple benchmark datasets using different GNN backbones and explanation methods. Experimental results show that our method consistently improves explanation quality and can be flexibly integrated with different architectures.
Authors: Yuly Wu, Jiamou Liu, Libo Zhang
Abstract: Partially Observable Markov Decision Processes (POMDPs) are fundamental to many real-world applications. Although reinforcement learning (RL) has shown success in fully observable domains, learning policies from traces in partially observable environments remains challenging due to non-Markovian observations. Inferring an automaton to handle the non-Markovianity is a proven effective approach, but faces two limitations: 1) existing automaton representations focus only on reward-based non-Markovianity, leading to unnatural problem formulations; 2) inference algorithms face enormous computational costs. For the first limitation, we introduce Transition Machines (TMs) to complement existing Reward Machines (RMs). To develop a unified inference algorithm for both automata types, we propose the Dual Behavior Mealy Machine (DBMM) that subsumes both TMs and RMs. We then introduce DB-RPNI, a passive automata learning algorithm that efficiently infers DBMMs while avoiding the costly reductions required by prior work. We further develop optimization techniques and identify sufficient conditions for inferring the minimal correct automata. Experimentally, our inference method achieves speedups of up to three orders of magnitude over SOTA baselines.
Authors: Max Gupta
Abstract: Rapidly learning abstract concepts from limited examples is a hallmark of human intelligence. This work investigates whether gradient-based meta-learning can equip neural networks with inductive biases for efficient few-shot acquisition of discrete concepts. We compare meta-learning methods against a supervised learning baseline on Boolean tasks generated by a probabilistic context-free grammar (PCFG). By systematically varying concept dimensionality (number of features) and compositionality (depth of grammar recursion), we identify regimes in which meta-learning robustly improves few-shot concept learning. We find improved performance and sample efficiency by training a multilayer perceptron (MLP) across concept spaces increasing in dimensional and compositional complexity. We are able to show that meta-learners are much better able to handle compositional complexity than featural complexity and establish an empirical analysis demonstrating how featural complexity shapes 'concept basins' of the loss landscape, allowing curvature-aware optimization to be more effective than first order methods. We see that we can robustly increase generalization on complex concepts by increasing the number of adaptation steps in meta-SGD, encouraging exploration of rougher loss basins. Overall, this work highlights the intricacies of learning compositional versus featural complexity in high dimensional concept spaces and provides a road to understanding the role of 2nd order methods and extended gradient adaptation in few-shot concept learning.
Authors: Dekang Meng, Rabab Haider, Pascal van Hentenryck
Abstract: This paper introduces OptiGridML, a machine learning framework for discrete topology optimization in power grids. The task involves selecting substation breaker configurations that maximize cross-region power exports, a problem typically formulated as a mixed-integer program (MIP) that is NP-hard and computationally intractable for large networks. OptiGridML replaces repeated MIP solves with a two-stage neural architecture: a line-graph neural network (LGNN) that approximates DC power flows for a given network topology, and a heterogeneous GNN (HeteroGNN) that predicts breaker states under structural and physical constraints. A physics-informed consistency loss connects these components by enforcing Kirchhoff's law on predicted flows. Experiments on synthetic networks with up to 1,000 breakers show that OptiGridML achieves power export improvements of up to 18% over baseline topologies, while reducing inference time from hours to milliseconds. These results demonstrate the potential of structured, flow-aware GNNs for accelerating combinatorial optimization in physical networked systems.
Authors: Alexander Norcliffe, Changhee Lee, Fergus Imrie, Mihaela van der Schaar, Pietro Lio
Abstract: Active Feature Acquisition is an instance-wise, sequential decision making problem. The aim is to dynamically select which feature to measure based on current observations, independently for each test instance. Common approaches either use Reinforcement Learning, which experiences training difficulties, or greedily maximize the conditional mutual information of the label and unobserved features, which makes myopic acquisitions. To address these shortcomings, we introduce a latent variable model, trained in a supervised manner. Acquisitions are made by reasoning about the features across many possible unobserved realizations in a stochastic latent space. Extensive evaluation on a large range of synthetic and real datasets demonstrates that our approach reliably outperforms a diverse set of baselines.
Authors: Yixin Shen
Abstract: Fine-tuning massive pre-trained language models across many tasks demands adapters that are both parameter-efficient and highly expressive. We introduce \textbf{Kron-LoRA}, a two-stage adapter that first factorizes each frozen linear update as a Kronecker product \[ \Delta W = A \otimes B \] and then compresses \[ B \in \mathbb{R}^{d_{B2}\times d_{B1}} \] via an \(r\)-rank LoRA decomposition \(B \approx B_{1}B_{2}\). By leveraging \[ \mathrm{rank}(A \otimes B) \;=\; \mathrm{rank}(A)\,\mathrm{rank}(B), \] Kron-LoRA retains the expressivity of the update while using up to $4\!\times\!$ fewer parameters than a standard rank-8 LoRA adapter. Its compact adapter matrices also quantize to 8- or 4-bit with less accuracy degradation than LoRA, enabling further memory and storage savings for on-device deployment. We benchmark on DistilBERT and Mistral-7B across five tasks (PIQA, HellaSwag, WinoGrande, ARC-Easy, ARC-Challenge) over multiple epochs of adapter-only tuning: on DistilBERT, an 840 K-parameter Kron-LoRA matches LoRA-16's performance, and on Mistral-7B, a 5.7 M-parameter Kron-LoRA rivals LoRA-8 with modest memory savings and only a 3-8\% speed overhead. In sequential fine-tuning from ARC-Challenge to ARC-Easy, Kron-LoRA retains 55.18\% accuracy versus 53.17\% for LoRA-8-despite using only one-quarter of the adapter parameters-underscoring its competitive cross-task transfer performance. By uniting Kronecker structure, low-rank compression, quantization-friendliness, and by providing transparent trade-off analysis, Kron-LoRA offers a scalable, sustainable, and continual-learning-ready solution for multi-task adaptation of large language models.
Authors: Seyyed Saeid Cheshmi, Azal Ahmad Khan, Xinran Wang, Zirui Liu, Ali Anwar
Abstract: Large Language Models (LLMs) are increasingly relied upon for solving complex reasoning tasks in domains such as mathematics, logic, and multi-step question answering. A growing line of work seeks to improve reasoning quality by scaling inference time compute particularly through Process Reward Models (PRMs), used to reward the reasoning at intermediate steps. While effective, these methods introduce substantial computational overhead, especially when generating large numbers of solutions in parallel. In this paper, we investigate whether PRMs can be used mid-generation to provide early signals that enable the rejection of suboptimal candidates before full generation of step is complete. We introduce the hypothesis that PRMs are also Partial Reward Models, meaning that the scores they assign to partially completed reasoning step are predictive of final output quality. This allows for principled early rejection based on intermediate token-level signals. We support this hypothesis both theoretically, by proving that the risk of discarding optimal beams decreases exponentially with generation length and empirically, by demonstrating a strong correlation between partial and final rewards across multiple reward models. On math reasoning benchmarks, our method achieves up to 1.4$\times$-9$\times$ reduction in inference FLOPs without degrading final performance. These results suggest that early rejection is a powerful mechanism for improving the compute-efficiency of reasoning in LLMs.
Authors: Rituparna Datta, Jiaming Cui, Zihan Guan, Rupesh Silwal, Joshua C Eby, Gregory Madden, Anil Vullikanti
Abstract: Accurate prediction of clinical outcomes using Electronic Health Records (EHRs) is critical for early intervention, efficient resource allocation, and improved patient care. EHRs contain multimodal data, including both structured data and unstructured clinical notes that provide rich, context-specific information. In this work, we introduce a unified framework that seamlessly integrates these diverse modalities, leveraging all relevant available information through a two-stage architecture for clinical risk prediction. In the first stage, a fine-tuned Large Language Model (LLM) extracts crucial, task-relevant information from clinical notes, which is enhanced by graph-based retrieval of external domain knowledge from sources such as a medical corpus like PubMed, grounding the LLM's understanding. The second stage combines both unstructured representations and features derived from the structured data to generate the final predictions. This approach supports a wide range of clinical tasks. Here, we demonstrate its effectiveness on 30-day readmission and in-hospital mortality prediction. Experimental results show that our framework achieves strong performance, with AUC scores of $0.84$ and $0.92$, respectively, despite these tasks involving severely imbalanced datasets, with positive rates ranging from approximately $4\%$ to $13\%$. Moreover, it outperforms all existing baselines and clinical practices, including established risk scoring systems. To the best of our knowledge, this is one of the first frameworks for healthcare prediction which enhances the power of an LLM-based graph-guided knowledge retrieval method by combining it with structured data for improved clinical outcome prediction.
Authors: Ziyu Zhou, Yiming Huang, Yanyun Wang, Yuankai Wu, James Kwok, Yuxuan Liang
Abstract: Irregular multivariate time series (IMTS), characterized by uneven sampling and inter-variate asynchrony, fuel many forecasting applications yet remain challenging to model efficiently. Canonical Pre-Alignment (CPA) has been widely adopted in IMTS modeling by padding zeros at every global timestamp, thereby alleviating inter-variate asynchrony and unifying the series length, but its dense zero-padding inflates the pre-aligned series length, especially when numerous variates are present, causing prohibitive compute overhead. Recent graph-based models with patching strategies sidestep CPA, but their local message passing struggles to capture global inter-variate correlations. Therefore, we posit that CPA should be retained, with the pre-aligned series properly handled by the model, enabling it to outperform state-of-the-art graph-based baselines that sidestep CPA. Technically, we propose KAFNet, a compact architecture grounded in CPA for IMTS forecasting that couples (1) Pre-Convolution module for sequence smoothing and sparsity mitigation, (2) Temporal Kernel Aggregation module for learnable compression and modeling of intra-series irregularity, and (3) Frequency Linear Attention blocks for the low-cost inter-series correlations modeling in the frequency domain. Experiments on multiple IMTS datasets show that KAFNet achieves state-of-the-art forecasting performance, with a 7.2$\times$ parameter reduction and a 8.4$\times$ training-inference acceleration.
Authors: Hyungjin Chung, Jeongsol Kim, Jong Chul Ye
Abstract: Using diffusion priors to solve inverse problems in imaging have significantly matured over the years. In this chapter, we review the various different approaches that were proposed over the years. We categorize the approaches into the more classic explicit approximation approaches and others, which include variational inference, sequential monte carlo, and decoupled data consistency. We cover the extension to more challenging situations, including blind cases, high-dimensional data, and problems under data scarcity and distribution mismatch. More recent approaches that aim to leverage multimodal information through texts are covered. Through this chapter, we aim to (i) distill the common mathematical threads that connect these algorithms, (ii) systematically contrast their assumptions and performance trade-offs across representative inverse problems, and (iii) spotlight the open theoretical and practical challenges by clarifying the landscape of diffusion model based inverse problem solvers.
Authors: Shutong Qiao, Wei Yuan, Junliang Yu, Tong Chen, Quoc Viet Hung Nguyen, Hongzhi Yin
Abstract: Recommender systems (RSs) are now fundamental to various online platforms, but their dependence on user-contributed data leaves them vulnerable to shilling attacks that can manipulate item rankings by injecting fake users. Although widely studied, most existing attack models fail to meet two critical objectives simultaneously: achieving strong adversarial promotion of target items while maintaining realistic behavior to evade detection. As a result, the true severity of shilling threats that manage to reconcile the two objectives remains underappreciated. To expose this overlooked vulnerability, we present DLDA, a diffusion-based attack framework that can generate highly effective yet indistinguishable fake users by enabling fine-grained control over target promotion. Specifically, DLDA operates in a pre-aligned collaborative embedding space, where it employs a conditional latent diffusion process to iteratively synthesize fake user profiles with precise target item control. To evade detection, DLDA introduces a dispersive regularization mechanism that promotes variability and realism in generated behavioral patterns. Extensive experiments on three real-world datasets and five popular RS models demonstrate that, compared to prior attacks, DLDA consistently achieves stronger item promotion while remaining harder to detect. These results highlight that modern RSs are more vulnerable than previously recognized, underscoring the urgent need for more robust defenses.
Authors: Hongze Sun, Wuque Cai, Duo Chen, Shifeng Mao, Jiayi He, Zhenxing Wang, Dezhong Yao, Daqing Guo
Abstract: As a foundational architecture of artificial intelligence models, Transformer has been recently adapted to spiking neural networks with promising performance across various tasks. However, existing spiking Transformer (ST)-based models require a substantial number of parameters and incur high computational costs, thus limiting their deployment in resource-constrained environments. To address these challenges, we propose combining synapse pruning with a synergistic learning-based compensation strategy to derive lightweight ST-based models. Specifically, two types of tailored pruning strategies are introduced to reduce redundancy in the weight matrices of ST blocks: an unstructured $\mathrm{L_{1}P}$ method to induce sparse representations, and a structured DSP method to induce low-rank representations. In addition, we propose an enhanced spiking neuron model, termed the synergistic leaky integrate-and-fire (sLIF) neuron, to effectively compensate for model pruning through synergistic learning between synaptic and intrinsic plasticity mechanisms. Extensive experiments on benchmark datasets demonstrate that the proposed methods significantly reduce model size and computational overhead while maintaining competitive performance. These results validate the effectiveness of the proposed pruning and compensation strategies in constructing efficient and high-performing ST-based models.
Authors: Yu Lei, Jiayang Zhao, Yilei Zhao, Zhaoqi Zhang, Linyou Cai, Qianlong Xie, Xingxing Wang
Abstract: Modern auto-bidding systems are required to balance overall performance with diverse advertiser goals and real-world constraints, reflecting the dynamic and evolving needs of the industry. Recent advances in conditional generative models, such as transformers and diffusers, have enabled direct trajectory generation tailored to advertiser preferences, offering a promising alternative to traditional Markov Decision Process-based methods. However, these generative methods face significant challenges, such as the distribution shift between offline and online environments, limited exploration of the action space, and the necessity to meet constraints like marginal Cost-per-Mille (CPM) and Return on Investment (ROI). To tackle these challenges, we propose GRAD (Generative Reward-driven Ad-bidding with Mixture-of-Experts), a scalable foundation model for auto-bidding that combines an Action-Mixture-of-Experts module for diverse bidding action exploration with the Value Estimator of Causal Transformer for constraint-aware optimization. Extensive offline and online experiments demonstrate that GRAD significantly enhances platform revenue, highlighting its effectiveness in addressing the evolving and diverse requirements of modern advertisers. Furthermore, GRAD has been implemented in multiple marketing scenarios at Meituan, one of the world's largest online food delivery platforms, leading to a 2.18% increase in Gross Merchandise Value (GMV) and 10.68% increase in ROI.
Authors: Xinzheng Wu, Junyi Chen, Shaolingfeng Ye, Wei Jiang, Yong Shen
Abstract: In the autonomous driving testing methods based on evolving scenarios, the construction method of the driver model, which determines the driving maneuvers of background vehicles (BVs) in the scenario, plays a critical role in generating safety-critical scenarios. In particular, the cooperative adversarial driving characteristics between BVs can contribute to the efficient generation of safety-critical scenarios with high testing value. In this paper, a multi-agent reinforcement learning (MARL) method is used to train and generate a dual-modal driver model (Dual-DM) with non-adversarial and adversarial driving modalities. The model is then connected to a continuous simulated traffic environment to generate complex, diverse and strong interactive safety-critical scenarios through evolving scenario generation method. After that, the generated evolving scenarios are evaluated in terms of fidelity, test efficiency, complexity and diversity. Results show that without performance degradation in scenario fidelity (>85% similarity to real-world scenarios) and complexity (complexity metric: 0.45, +32.35% and +12.5% over two baselines), Dual-DM achieves a substantial enhancement in the efficiency of generating safety-critical scenarios (efficiency metric: 0.86, +195% over two baselines). Furthermore, statistical analysis and case studies demonstrate the diversity of safety-critical evolving scenarios generated by Dual-DM in terms of the adversarial interaction patterns. Therefore, Dual-DM can greatly improve the performance of the generation of safety-critical scenarios through evolving scenario generation method.
Authors: Zhilong Zhao, Yindi Liu
Abstract: LLMs enable qualitative coding at large scale, but assessing the reliability of their output remains challenging in domains where human experts seldom agree. Analysing 5,680 coding decisions from eight state-of-the-art LLMs across ten thematic categories, we confirm that a model's mean self-confidence already tracks inter-model agreement closely (Pearson r=0.82). Adding model diversity-quantified as the normalised Shannon entropy of the panel's votes-turns this single cue into a dual signal that explains agreement almost completely (R^2=0.979). The confidence-diversity duo enables a three-tier workflow that auto-accepts 35% of segments with <5% audit-detected error and routes the remainder for targeted human review, cutting manual effort by up to 65%. Cross-domain replication on six public datasets spanning finance, medicine, law and multilingual tasks confirms these gains (kappa improvements of 0.20-0.78). Our results establish a generalisable, evidence-based criterion for calibrating AI judgement in qualitative research.
Authors: Sijia Wang, Ricardo Henao
Abstract: Increasing concerns for data privacy and other difficulties associated with retrieving source data for model training have created the need for source-free transfer learning, in which one only has access to pre-trained models instead of data from the original source domains. This setting introduces many challenges, as many existing transfer learning methods typically rely on access to source data, which limits their direct applicability to scenarios where source data is unavailable. Further, practical concerns make it more difficult, for instance efficiently selecting models for transfer without information on source data, and transferring without full access to the source models. So motivated, we propose a model recycling framework for parameter-efficient training of models that identifies subsets of related source models to reuse in both white-box and black-box settings. Consequently, our framework makes it possible for Model as a Service (MaaS) providers to build libraries of efficient pre-trained models, thus creating an opportunity for multi-source data-free supervised transfer learning.
Authors: Hang Yin, Zipeng Liu, Xiaoyong Peng, Liyao Xiang
Abstract: Graph unlearning is tailored for GNNs to handle widespread and various graph structure unlearning requests, which remain largely unexplored. The GIF (graph influence function) achieves validity under partial edge unlearning, but faces challenges in dealing with more disturbing node unlearning. To avoid the overhead of retraining and realize the model utility of unlearning, we proposed a novel node unlearning method to reverse the process of aggregation in GNN by embedding reconstruction and to adopt Range-Null Space Decomposition for the nodes' interaction learning. Experimental results on multiple representative datasets demonstrate the SOTA performance of our proposed approach.
Authors: Rui Sun, Chenghua Gong, Tianjun Gu, Yuhao Zheng, Jie Ding, Juyuan Zhang, Liming Pan, Linyuan L\"u
Abstract: Advancing epidemic dynamics forecasting is vital for targeted interventions and safeguarding public health. Current approaches mainly fall into two categories: mechanism-based and data-driven models. Mechanism-based models are constrained by predefined compartmental structures and oversimplified system assumptions, limiting their ability to model complex real-world dynamics, while data-driven models focus solely on intrinsic data dependencies without physical or epidemiological constraints, risking biased or misleading representations. Although recent studies have attempted to integrate epidemiological knowledge into neural architectures, most of them fail to reconcile explicit physical priors with neural representations. To overcome these obstacles, we introduce Epi$^2$-Net, a Epidemic Forecasting Framework built upon Physics-Inspired Neural Networks. Specifically, we propose reconceptualizing epidemic transmission from the physical transport perspective, introducing the concept of neural epidemic transport. Further, we present a physic-inspired deep learning framework, and integrate physical constraints with neural modules to model spatio-temporal patterns of epidemic dynamics. Experiments on real-world datasets have demonstrated that Epi$^2$-Net outperforms state-of-the-art methods in epidemic forecasting, providing a promising solution for future epidemic containment. The code is available at: https://anonymous.4open.science/r/Epi-2-Net-48CE.
Authors: Guojiang Zhao, Sihang Li, Zixiang Lu, Zheng Cheng, Haitao Lin, Lirong Wu, Hanchen Xia, Hengxing Cai, Wentao Guo, Hongshuai Wang, Mingjun Xu, Siyu Zhu, Guolin Ke, Linfeng Zhang, Zhifeng Gao
Abstract: Large Language Models(LLMs) have demonstrated remarkable performance across various domains, yet their capabilities in molecular reasoning remain insufficiently explored. Current approaches tend to rely heavily on general-purpose prompting, which lacks domain-specific molecular semantics, while those that use fine-tuning strategies often face challenges with interpretability and reasoning depth. To address these issues, we introduce MolReasoner, a two-stage framework designed to transition LLMs from memorization towards chemical reasoning. First, we propose Mol-SFT, which initializes the model's reasoning abilities via synthetic Chain-of-Thought(CoT) samples generated by GPT-4o and verified for chemical accuracy. Subsequently, Mol-RL applies reinforcement learning with specialized reward functions designed explicitly to align chemical structures with linguistic descriptions, thereby enhancing molecular reasoning capabilities. Our approach notably enhances interpretability, improving the model 's molecular understanding and enabling better generalization. Extensive experiments demonstrate that MolReasoner outperforms existing methods, and marking a significant shift from memorization-based outputs to robust chemical reasoning.
Authors: Bang Hu, Changze Lv, Mingjie Li, Yunpeng Liu, Xiaoqing Zheng, Fengzhe Zhang, Wei cao, Fan Zhang
Abstract: Spiking neural networks (SNNs), inspired by the spiking behavior of biological neurons, offer a distinctive approach for capturing the complexities of temporal data. However, their potential for spatial modeling in multivariate time-series forecasting remains largely unexplored. To bridge this gap, we introduce a brand new SNN architecture, which is among the first to seamlessly integrate graph structural learning with spike-based temporal processing for multivariate time-series forecasting. Specifically, we first embed time features and an adaptive matrix, eliminating the need for predefined graph structures. We then further learn sequence features through the Observation (OBS) Block. Building upon this, our Multi-Scale Spike Aggregation (MSSA) hierarchically aggregates neighborhood information through spiking SAGE layers, enabling multi-hop feature extraction while eliminating the need for floating-point operations. Finally, we propose a Dual-Path Spike Fusion (DSF) Block to integrate spatial graph features and temporal dynamics via a spike-gated mechanism, combining LSTM-processed sequences with spiking self-attention outputs, effectively improve the model accuracy of long sequence datasets. Experiments show that our model surpasses the state-of-the-art SNN-based iSpikformer on all datasets and outperforms traditional temporal models at long horizons, thereby establishing a new paradigm for efficient spatial-temporal modeling.
Authors: Amitava Das, Abhilekh Borah, Vinija Jain, Aman Chadha
Abstract: Low-rank adaptation (LoRA) has become a standard tool for efficiently fine-tuning large language models (LLMs). Yet, even minor LoRA updates can induce alignment drift, weakening safety and behavioral constraints through entangled parameter changes. To address this, we propose AlignGuard-LoRA (AGL), a principled framework for preserving alignment during finetuning. AGL introduces several key components: a primary task loss for supervision, Fisher Information Matrix-based regularization to restrict updates in alignment-sensitive subspaces, and task-specific regularization to stabilize the integration of new knowledge. We further introduce collision-aware regularization, blending Riemannian overlap -- which penalizes coordinate-wise interference -- and geodesic separation -- which encourages disjoint update geometry. We curate DriftCaps, a targeted diagnostic benchmark of safe and unsafe prompts designed to quantify alignment drift and safety degradation. Empirical evaluations show that AGL mitigates alignment drift by up to 50% on safety-critical benchmarks without degrading downstream task performance. Comprehensive ablation confirms that each component contributes distinctly to preserving latent safety behaviors. Finally, we derive and validate a scaling law for catastrophic forgetting, revealing that AGL flattens post-finetuning loss escalation while preserving adaptation dynamics. AGL is a structurally grounded refinement of LoRA, ensuring alignment preservation with minimal trade-offs. To encourage further exploration and development, we open-source our implementation.
Authors: Pawel Gajer, Jacques Ravel
Abstract: This paper presents a mathematical framework for analyzing machine learning models through the geometry of their induced partitions. By representing partitions as Riemannian simplicial complexes, we capture not only adjacency relationships but also geometric properties including cell volumes, volumes of faces where cells meet, and dihedral angles between adjacent cells. For neural networks, we introduce a differential forms approach that tracks geometric structure through layers via pullback operations, making computations tractable by focusing on data-containing cells. The framework enables geometric regularization that directly penalizes problematic spatial configurations and provides new tools for model refinement through extended Laplacians and simplicial splines. We also explore how data distribution induces effective geometric curvature in model partitions, developing discrete curvature measures for vertices that quantify local geometric complexity and statistical Ricci curvature for edges that captures pairwise relationships between cells. While focused on mathematical foundations, this geometric perspective offers new approaches to model interpretation, regularization, and diagnostic tools for understanding learning dynamics.
Authors: Xiaoya Li, Xiaofei Sun, Albert Wang, Chris Shum, Jiwei Li
Abstract: Approximate nearest-neighbor search (ANNS) algorithms have become increasingly critical for recent AI applications, particularly in retrieval-augmented generation (RAG) and agent-based LLM applications. In this paper, we present CRINN, a new paradigm for ANNS algorithms. CRINN treats ANNS optimization as a reinforcement learning problem where execution speed serves as the reward signal. This approach enables the automatic generation of progressively faster ANNS implementations while maintaining accuracy constraints. Our experimental evaluation demonstrates CRINN's effectiveness across six widely-used NNS benchmark datasets. When compared against state-of-the-art open-source ANNS algorithms, CRINN achieves best performance on three of them (GIST-960-Euclidean, MNIST-784-Euclidean, and GloVe-25-angular), and tied for first place on two of them (SIFT-128-Euclidean and GloVe-25-angular). The implications of CRINN's success reach well beyond ANNS optimization: It validates that LLMs augmented with reinforcement learning can function as an effective tool for automating sophisticated algorithmic optimizations that demand specialized knowledge and labor-intensive manual refinement.Code can be found at https://github.com/deepreinforce-ai/CRINN
Authors: Runze Zhao, Yue Yu, Ruhan Wang, Chunfeng Huang, Dongruo Zhou
Abstract: Continuous-time reinforcement learning (CTRL) provides a natural framework for sequential decision-making in dynamic environments where interactions evolve continuously over time. While CTRL has shown growing empirical success, its ability to adapt to varying levels of problem difficulty remains poorly understood. In this work, we investigate the instance-dependent behavior of CTRL and introduce a simple, model-based algorithm built on maximum likelihood estimation (MLE) with a general function approximator. Unlike existing approaches that estimate system dynamics directly, our method estimates the state marginal density to guide learning. We establish instance-dependent performance guarantees by deriving a regret bound that scales with the total reward variance and measurement resolution. Notably, the regret becomes independent of the specific measurement strategy when the observation frequency adapts appropriately to the problem's complexity. To further improve performance, our algorithm incorporates a randomized measurement schedule that enhances sample efficiency without increasing measurement cost. These results highlight a new direction for designing CTRL algorithms that automatically adjust their learning behavior based on the underlying difficulty of the environment.
Authors: Abyad Enan, Abdullah Al Mamun, Gurcan Comert, Debbie Aisiana Indah, Judith Mwakalonge, Amy W. Apon, Mashrur Chowdhury
Abstract: Large trucks substantially contribute to work zone-related crashes, primarily due to their large size and blind spots. When approaching a work zone, large trucks often need to merge into an adjacent lane because of lane closures caused by construction activities. This study aims to enhance the safety of large truck merging maneuvers in work zones by evaluating the risk associated with merging conflicts and establishing a decision-making strategy for merging based on this risk assessment. To predict the risk of large trucks merging into a mixed traffic stream within a work zone, a Long Short-Term Memory (LSTM) neural network is employed. For a large truck intending to merge, it is critical that the immediate downstream vehicle in the target lane maintains a minimum safe gap to facilitate a safe merging process. Once a conflict-free merging opportunity is predicted, large trucks are instructed to merge in response to the lane closure. Our LSTM-based conflict prediction method is compared against baseline approaches, which include probabilistic risk-based merging, 50th percentile gap-based merging, and 85th percentile gap-based merging strategies. The results demonstrate that our method yields a lower conflict risk, as indicated by reduced Time Exposed Time-to-Collision (TET) and Time Integrated Time-to-Collision (TIT) values relative to the baseline models. Furthermore, the findings indicate that large trucks that use our method can perform early merging while still in motion, as opposed to coming to a complete stop at the end of the current lane prior to closure, which is commonly observed with the baseline approaches.
Authors: Ju Chen, Jun Feng, Shenyu Zhang
Abstract: Multi-class classification annotations have significantly advanced AI applications, with truth inference serving as a critical technique for aggregating noisy and biased annotations. Existing state-of-the-art methods typically model each annotator's expertise using a confusion matrix. However, these methods suffer from two widely recognized issues: 1) when most annotators label only a few tasks, or when classes are imbalanced, the estimated confusion matrices are unreliable, and 2) a single confusion matrix often remains inadequate for capturing each annotator's full expertise patterns across all tasks. To address these issues, we propose a novel confusion-matrix-based method, PTBCC (ProtoType learning-driven Bayesian Classifier Combination), to introduce a reliable and richer annotator estimation by prototype learning. Specifically, we assume that there exists a set $S$ of prototype confusion matrices, which capture the inherent expertise patterns of all annotators. Rather than a single confusion matrix, the expertise per annotator is extended as a Dirichlet prior distribution over these prototypes. This prototype learning-driven mechanism circumvents the data sparsity and class imbalance issues, ensuring a richer and more flexible characterization of annotators. Extensive experiments on 11 real-world datasets demonstrate that PTBCC achieves up to a 15% accuracy improvement in the best case, and a 3% higher average accuracy while reducing computational cost by over 90%.
Authors: Saleh Nikooroo, Thomas Engel
Abstract: While modern deep networks have demonstrated remarkable versatility, their training dynamics remain poorly understood--often driven more by empirical tweaks than architectural insight. This paper investigates how internal structural choices shape the behavior of learning systems. Building on prior efforts that introduced simple architectural constraints, we explore the broader implications of structure for convergence, generalization, and adaptation. Our approach centers on a family of enriched transformation layers that incorporate constrained pathways and adaptive corrections. We analyze how these structures influence gradient flow, spectral sensitivity, and fixed-point behavior--uncovering mechanisms that contribute to training stability and representational regularity. Theoretical analysis is paired with empirical studies on synthetic and structured tasks, demonstrating improved robustness, smoother optimization, and scalable depth behavior. Rather than prescribing fixed templates, we emphasize principles of tractable design that can steer learning behavior in interpretable ways. Our findings support a growing view that architectural design is not merely a matter of performance tuning, but a critical axis for shaping learning dynamics in scalable and trustworthy neural systems.
Authors: Tai An, Ruwu Cai, Yanzhe Zhang, Yang Liu, Hao Chen, Pengcheng Xie, Sheng Chang, Yiwu Yao, Gongyi Wang
Abstract: In the era of large language models (LLMs), N:M sparsity has emerged as a structured compression technique critical for accelerating inference. While prior work has primarily focused on weight sparsity, it often suffers from significant accuracy degradation. Activation sparsity, though promising, is typically training-dependent and faces challenges in generalization. To address these limitations, we introduce Amber Pruner, a training-free N:M activation sparsity method designed specifically for the prefill stage, targeting the acceleration of linear projection layers in LLMs. Extensive experiments across multiple models and sparsity ratios (2:4, 4:8, and 8:16) demonstrate that Amber Pruner can effectively sparsify and accelerate more than 55% of linear computations without requiring model retraining. To further enhance generality and efficiency, we propose Outstanding-sparse, a unified framework that integrates Amber Pruner with post-training W8A8 quantization. Our approach preserves strong performance across a range of downstream tasks, with notable advantages in generative tasks. This work pioneers a new frontier in activation sparsity, providing foundational insights that are poised to guide the co-evolution of algorithms and architectures in the design of next-generation AI systems.
Authors: Boyuan Zheng, Victor W. Chu, Zhidong Li, Evan Webster, Ashley Rootsey
Abstract: Climate change has intensified the frequency and severity of extreme weather events, presenting unprecedented challenges to the agricultural industry worldwide. In this investigation, we focus on kiwifruit farming in New Zealand. We propose to examine the impacts of climate-induced extreme events, specifically frost, drought, extreme rainfall, and heatwave, on kiwifruit harvest yields. These four events were selected due to their significant impacts on crop productivity and their prevalence as recorded by climate monitoring institutions in the country. We employed Isolation Forest, an unsupervised anomaly detection method, to analyse climate history and recorded extreme events, alongside with kiwifruit yields. Our analysis reveals considerable variability in how different types of extreme event affect kiwifruit yields underscoring notable discrepancies between climatic extremes and individual farm's yield outcomes. Additionally, our study highlights critical limitations of current anomaly detection approaches, particularly in accurately identifying events such as frost. These findings emphasise the need for integrating supplementary features like farm management strategies with climate adaptation practices. Our further investigation will employ ensemble methods that consolidate nearby farms' yield data and regional climate station features to reduce variance, thereby enhancing the accuracy and reliability of extreme event detection and the formulation of response strategies.
Authors: Qi Xiong, Hai Dong, Nasrin Sohrabi, Zahir Tari
Abstract: Sybil attacks pose a significant threat to federated learning, as malicious nodes can collaborate and gain a majority, thereby overwhelming the system. Therefore, it is essential to develop countermeasures that ensure the security of federated learning environments. We present a novel defence method against targeted data poisoning, which is one of the types of Sybil attacks, called Linear Algebra-based Detection (FedLAD). Unlike existing approaches, such as clustering and robust training, which struggle in situations where malicious nodes dominate, FedLAD models the federated learning aggregation process as a linear problem, transforming it into a linear algebra optimisation challenge. This method identifies potential attacks by extracting the independent linear combinations from the original linear combinations, effectively filtering out redundant and malicious elements. Extensive experimental evaluations demonstrate the effectiveness of FedLAD compared to five well-established defence methods: Sherpa, CONTRA, Median, Trimmed Mean, and Krum. Using tasks from both image classification and natural language processing, our experiments confirm that FedLAD is robust and not dependent on specific application settings. The results indicate that FedLAD effectively protects federated learning systems across a broad spectrum of malicious node ratios. Compared to baseline defence methods, FedLAD maintains a low attack success rate for malicious nodes when their ratio ranges from 0.2 to 0.8. Additionally, it preserves high model accuracy when the malicious node ratio is between 0.2 and 0.5. These findings underscore FedLAD's potential to enhance both the reliability and performance of federated learning systems in the face of data poisoning attacks.
Authors: Zhongyue Zhang, Jiahua Rao, Jie Zhong, Weiqiang Bai, Dongxue Wang, Shaobo Ning, Lifeng Qiao, Sheng Xu, Runze Ma, Will Hua, Jack Xiaoyu Chen, Odin Zhang, Wei Lu, Hanyi Feng, He Yang, Xinchao Shi, Rui Li, Wanli Ouyang, Xinzhu Ma, Jiahao Wang, Jixian Zhang, Jia Duan, Siqi Sun, Jian Zhang, Shuangjia Zheng
Abstract: Most human proteins remain undrugged, over 96% of human proteins remain unexploited by approved therapeutics. While structure-based virtual screening promises to expand the druggable proteome, existing methods lack atomic-level precision and fail to predict binding fitness, limiting translational impact. We present AuroBind, a scalable virtual screening framework that fine-tunes a custom atomic-level structural model on million-scale chemogenomic data. AuroBind integrates direct preference optimization, self-distillation from high-confidence complexes, and a teacher-student acceleration strategy to jointly predict ligand-bound structures and binding fitness. The proposed models outperform state-of-the-art models on structural and functional benchmarks while enabling 100,000-fold faster screening across ultra-large compound libraries. In a prospective screen across ten disease-relevant targets, AuroBind achieved experimental hit rates of 7-69%, with top compounds reaching sub-nanomolar to picomolar potency. For the orphan GPCRs GPR151 and GPR160, AuroBind identified both agonists and antagonists with success rates of 16-30%, and functional assays confirmed GPR160 modulation in liver and prostate cancer models. AuroBind offers a generalizable framework for structure-function learning and high-throughput molecular screening, bridging the gap between structure prediction and therapeutic discovery.
Authors: Kuiyuan DIng, Caili Guo, Yang Yang, Zhongtian Du, Walid Saad
Abstract: Large-scale models (LSMs) can be an effective framework for semantic representation and understanding, thereby providing a suitable tool for designing semantic communication (SC) systems. However, their direct deployment is often hindered by high computational complexity and resource requirements. In this paper, a novel robust knowledge distillation based semantic communication (RKD-SC) framework is proposed to enable efficient and \textcolor{black}{channel-noise-robust} LSM-powered SC. The framework addresses two key challenges: determining optimal compact model architectures and effectively transferring knowledge while maintaining robustness against channel noise. First, a knowledge distillation-based lightweight differentiable architecture search (KDL-DARTS) algorithm is proposed. This algorithm integrates knowledge distillation loss and a complexity penalty into the neural architecture search process to identify high-performance, lightweight semantic encoder architectures. Second, a novel two-stage robust knowledge distillation (RKD) algorithm is developed to transfer semantic capabilities from an LSM (teacher) to a compact encoder (student) and subsequently enhance system robustness. To further improve resilience to channel impairments, a channel-aware transformer (CAT) block is introduced as the channel codec, trained under diverse channel conditions with variable-length outputs. Extensive simulations on image classification tasks demonstrate that the RKD-SC framework significantly reduces model parameters while preserving a high degree of the teacher model's performance and exhibiting superior robustness compared to existing methods.
Authors: Dongchi Huang, Jiaqi Wang, Yang Li, Chunhe Xia, Tianle Zhang, Kaige Zhang
Abstract: Partial observability presents a significant challenge for safe reinforcement learning, as it impedes the identification of potential risks and rewards. Leveraging specific types of privileged information during training to mitigate the effects of partial observability has yielded notable empirical successes. In this paper, we propose Asymmetric Constrained Partially Observable Markov Decision Processes (ACPOMDPs) to theoretically examine the advantages of incorporating privileged information. Building upon ACPOMDPs, we propose the Privileged Information Guided Dreamer, a model-based safe reinforcement learning approach that leverages privileged information to enhance the agent's safety and performance through privileged representation alignment and an asymmetric actor-critic structure. Our empirical results demonstrate that our approach significantly outperforms existing methods in terms of safety and task-centric performance. Meanwhile, compared to alternative privileged model-based reinforcement learning methods, our approach exhibits superior performance and ease of training.
Authors: Wei Hao, Bin Chong, Ronghua Ji, Chen Hou
Abstract: Trajectory prediction is essential for formulating proactive strategies that anticipate user mobility and support advance preparation. Therefore, how to reduce the forecasting error in user trajectory prediction within an acceptable inference time arises as an interesting issue. However, trajectory data contains both global and local temporal information, complicating the extraction of the complete temporal pattern. Moreover, user behavior occurs over different time scales, increasing the difficulty of capturing behavioral patterns. To address these challenges, a trajectory prediction model based on multilayer perceptron (MLP), multi-scale convolutional neural network (MSCNN), and cross-attention (CA) is proposed. Specifically, MLP is used to extract the global temporal information of each feature. In parallel, MSCNN is employed to extract the local temporal information by modeling interactions among features within a local temporal range. Convolutional kernels with different sizes are used in MSCNN to capture temporal information at multiple resolutions, enhancing the model's adaptability to different behavioral patterns. Finally, CA is applied to fuse the global and local temporal information. Experimental results show that our model reduces mean squared error (MSE) by 5.04% and mean absolute error (MAE) by 4.35% compared with ModernTCN in 12-step prediction, while maintaining similar inference time.
Authors: Kexin Zhao, Bo Wang, Cuiying Zhao, Tongyao Wan
Abstract: Optimizing credit limits, interest rates, and loan terms is crucial for managing borrower risk and lifetime value (LTV) in personal loan platform. However, counterfactual estimation of these continuous, multi-dimensional treatments faces significant challenges: randomized trials are often prohibited by risk controls and long repayment cycles, forcing reliance on biased observational data. Existing causal methods primarily handle binary/discrete treatments and struggle with continuous, multi-dimensional settings. Furthermore, financial domain knowledge mandates provably monotonic treatment-outcome relationships (e.g., risk increases with credit limit).To address these gaps, we propose Multi-Treatment-DML, a novel framework leveraging Double Machine Learning (DML) to: (i) debias observational data for causal effect estimation; (ii) handle arbitrary-dimensional continuous treatments; and (iii) enforce monotonic constraints between treatments and outcomes, guaranteeing adherence to domain requirements.Extensive experiments on public benchmarks and real-world industrial datasets demonstrate the effectiveness of our approach. Furthermore, online A/B testing conducted on a realworld personal loan platform, confirms the practical superiority of Multi-Treatment-DML in real-world loan operations.
Authors: Manh Nguyen, Sunil Gupta, Hung Le
Abstract: Ensuring truthfulness in large language models remains a critical challenge for reliable text generation. While supervised fine-tuning and reinforcement learning with human feedback have shown promise, they require substantial amount of annotated data and computational resources, limiting scalability. In contrast, decoding-time interventions offer lightweight alternatives without model retraining. However, existing decoding strategies often face issues like prompt sensitivity, limited generalization, or dependence on internal model states. We propose a context-aware adaptive decoding method that leverages a compact reference grounding space, built from as few as 10 annotated examples and comprising pairs of context embeddings and next token logits from truthful responses, to enable retrieval-based logit shaping during inference. At each decoding step, our method retrieves top-N semantically similar contexts and aggregates their associated next token logits to modify the LLM's logits. Across three open-ended question-answering benchmarks, our approach achieves a 2.8 percent average improvement on TruthfulQA and further outperforms existing baselines on both Biographies and WikiQA. Experimental results also demonstrate cross-task generalization, with TruthfulQA-derived grounding enhancing biography generation. Our model-agnostic, scalable, and efficient method requires only a single generation pass, highlighting the potential of context-aware decoding for factual reliability in LLMs.
Authors: Yigit Turkmen, Baturalp Buyukates, Melih Bastopcu
Abstract: Recent advancements in Large Language Models (LLMs) have transformed many fields including scientific discovery, content generation, biomedical text mining, and educational technology. However, the substantial requirements for training data, computational resources, and energy consumption pose significant challenges for their practical deployment. A promising alternative is to leverage smaller, specialized language models and aggregate their outputs to improve overall response quality. In this work, we investigate a networked LLM system composed of multiple users, a central task processor, and clusters of topic-specialized LLMs. Each user submits categorical binary (true/false) queries, which are routed by the task processor to a selected cluster of $m$ LLMs. After gathering individual responses, the processor returns a final aggregated answer to the user. We characterize both the information accuracy and response timeliness in this setting, and formulate a joint optimization problem to balance these two competing objectives. Our extensive simulations demonstrate that the aggregated responses consistently achieve higher accuracy than those of individual LLMs. Notably, this improvement is more significant when the participating LLMs exhibit similar standalone performance.
Authors: Yike Zhang, Zhiyuan He, Huiqiang Jiang, Chengruidong Zhang, Yuqing Yang, Jianyong Wang, Lili Qiu
Abstract: Large language models (LLMs) enable long-context tasks but face efficiency challenges due to the growing key-value (KV) cache. We propose LeanK, a learning-based method that prunes unimportant key (K) cache channels by leveraging static channel sparsity. With a novel two-stage training process, LeanK learns channel-wise static mask that could satisfy specific sparsity ratio and hardware alignment requirement. LeanK reduces GPU memory and accelerates decoding without sacrificing accuracy. Experiments demonstrate up to 70% K cache and 16%-18% V cache memory reduction. Custom decoding kernel enables 1.3x speedup for attention computation. We also provide insights into model channels and attention heads during long-context inference by analyzing the learned importance distribution. Our code is available at https://aka.ms/LeanK.
URLs: https://aka.ms/LeanK.
Authors: Zeyu Zhao, Yueling Che, Kaichen Liu, Jian Li, Junmei Yao
Abstract: Multi-objective reinforcement learning (MORL) plays a pivotal role in addressing multi-criteria decision-making problems in the real world. The multi-policy (MP) based methods are widely used to obtain high-quality Pareto front approximation for the MORL problems. However, traditional MP methods only rely on the online reinforcement learning (RL) and adopt the evolutionary framework with a large policy population. This may lead to sample inefficiency and/or overwhelmed agent-environment interactions in practice. By forsaking the evolutionary framework, we propose the novel Multi-policy Pareto Front Tracking (MPFT) framework without maintaining any policy population, where both online and offline MORL algorithms can be applied. The proposed MPFT framework includes four stages: Stage 1 approximates all the Pareto-vertex policies, whose mapping to the objective space fall on the vertices of the Pareto front. Stage 2 designs the new Pareto tracking mechanism to track the Pareto front, starting from each of the Pareto-vertex policies. Stage 3 identifies the sparse regions in the tracked Pareto front, and introduces a new objective weight adjustment method to fill the sparse regions. Finally, by combining all the policies tracked in Stages 2 and 3, Stage 4 approximates the Pareto front. Experiments are conducted on seven different continuous-action robotic control tasks with both online and offline MORL algorithms, and demonstrate the superior hypervolume performance of our proposed MPFT approach over the state-of-the-art benchmarks, with significantly reduced agent-environment interactions and hardware requirements.
Authors: Sangjun Park, Tony Q. S. Quek, Hyowoon Seo
Abstract: Recent advances in split learning (SL) have established it as a promising framework for privacy-preserving, communication-efficient distributed learning at the network edge. However, SL's sequential update process is vulnerable to even a single malicious client, which can significantly degrade model accuracy. To address this, we introduce Pigeon-SL, a novel scheme grounded in the pigeonhole principle that guarantees at least one entirely honest cluster among M clients, even when up to N of them are adversarial. In each global round, the access point partitions the clients into N+1 clusters, trains each cluster independently via vanilla SL, and evaluates their validation losses on a shared dataset. Only the cluster with the lowest loss advances, thereby isolating and discarding malicious updates. We further enhance training and communication efficiency with Pigeon-SL+, which repeats training on the selected cluster to match the update throughput of standard SL. We validate the robustness and effectiveness of our approach under three representative attack models -- label flipping, activation and gradient manipulation -- demonstrating significant improvements in accuracy and resilience over baseline SL methods in future intelligent wireless networks.
Authors: Tiantian Liu, Xiao Li, Huan Li, Hua Lu, Christian S. Jensen, Jianliang Xu
Abstract: Shortest path search is a core operation in graph-based applications, yet existing methods face important limitations. Classical algorithms such as Dijkstra's and A* become inefficient as graphs grow more complex, while index-based techniques often require substantial preprocessing and storage. Recent learning-based approaches typically focus on spatial graphs and rely on context-specific features like geographic coordinates, limiting their general applicability. We propose a versatile learning-based framework for shortest path search on generic graphs, without requiring domain-specific features. At the core of our approach is the construction of a skeleton graph that captures multi-level distance and hop information in a compact form. A Skeleton Graph Neural Network (SGNN) operates on this structure to learn node embeddings and predict distances and hop lengths between node pairs. These predictions support LSearch, a guided search algorithm that uses model-driven pruning to reduce the search space while preserving accuracy. To handle larger graphs, we introduce a hierarchical training strategy that partitions the graph into subgraphs with individually trained SGNNs. This structure enables HLSearch, an extension of our method for efficient path search across graph partitions. Experiments on five diverse real-world graphs demonstrate that our framework achieves strong performance across graph types, offering a flexible and effective solution for learning-based shortest path search.
Authors: Xiangru Tang, Zhuoyun Yu, Jiapeng Chen, Yan Cui, Daniel Shao, Weixu Wang, Fang Wu, Yuchen Zhuang, Wenqi Shi, Zhi Huang, Arman Cohan, Xihong Lin, Fabian Theis, Smita Krishnaswamy, Mark Gerstein
Abstract: Virtual cell modeling represents an emerging frontier at the intersection of artificial intelligence and biology, aiming to predict quantities such as responses to diverse perturbations quantitatively. However, autonomously building computational models for virtual cells is challenging due to the complexity of biological systems, the heterogeneity of data modalities, and the need for domain-specific expertise across multiple disciplines. Here, we introduce CellForge, an agentic system that leverages a multi-agent framework that transforms presented biological datasets and research objectives directly into optimized computational models for virtual cells. More specifically, given only raw single-cell multi-omics data and task descriptions as input, CellForge outputs both an optimized model architecture and executable code for training virtual cell models and inference. The framework integrates three core modules: Task Analysis for presented dataset characterization and relevant literature retrieval, Method Design, where specialized agents collaboratively develop optimized modeling strategies, and Experiment Execution for automated generation of code. The agents in the Design module are separated into experts with differing perspectives and a central moderator, and have to collaboratively exchange solutions until they achieve a reasonable consensus. We demonstrate CellForge's capabilities in single-cell perturbation prediction, using six diverse datasets that encompass gene knockouts, drug treatments, and cytokine stimulations across multiple modalities. CellForge consistently outperforms task-specific state-of-the-art methods. Overall, CellForge demonstrates how iterative interaction between LLM agents with differing perspectives provides better solutions than directly addressing a modeling challenge. Our code is publicly available at https://github.com/gersteinlab/CellForge.
Authors: Francis Boabang, Samuel Asante Gyamerah
Abstract: In insurance fraud prediction, handling class imbalance remains a critical challenge. This paper presents a novel multistage focal loss function designed to enhance the performance of machine learning models in such imbalanced settings by helping to escape local minima and converge to a good solution. Building upon the foundation of the standard focal loss, our proposed approach introduces a dynamic, multi-stage convex and nonconvex mechanism that progressively adjusts the focus on hard-to-classify samples across training epochs. This strategic refinement facilitates more stable learning and improved discrimination between fraudulent and legitimate cases. Through extensive experimentation on a real-world insurance dataset, our method achieved better performance than the traditional focal loss, as measured by accuracy, precision, F1-score, recall and Area Under the Curve (AUC) metrics on the auto insurance dataset. These results demonstrate the efficacy of the multistage focal loss in boosting model robustness and predictive accuracy in highly skewed classification tasks, offering significant implications for fraud detection systems in the insurance industry. An explainable model is included to interpret the results.
Authors: Chenqing Lin, Mostafa Hussien, Chengyao Yu, Mohamed Cheriet, Osama Abdelrahman, Ruixing Ming
Abstract: Neural network pruning is a critical compression technique that facilitates the deployment of large-scale neural networks on resource-constrained edge devices, typically by identifying and eliminating redundant or insignificant parameters to reduce computational and memory overhead. This paper proposes the Flexible Automatic Identification and Removal (FAIR)-Pruner, a novel method for neural network structured pruning. Specifically, FAIR-Pruner first evaluates the importance of each unit (e.g., neuron or channel) through the Utilization Score quantified by the Wasserstein distance. To reflect the performance degradation after unit removal, it then introduces the Reconstruction Error, which is computed via the Taylor expansion of the loss function. Finally, FAIR-Pruner identifies superfluous units with negligible impact on model performance by controlling the proposed Tolerance of Difference, which measures differences between unimportant units and those that cause performance degradation. A major advantage of FAIR-Pruner lies in its capacity to automatically determine the layer-wise pruning rates, which yields a more efficient subnetwork structure compared to applying a uniform pruning rate. Another advantage of the FAIR-Pruner is its great one-shot performance without post-pruning fine-tuning. Furthermore, with utilization scores and reconstruction errors, users can flexibly obtain pruned models under different pruning ratios. Comprehensive experimental validation on diverse benchmark datasets (e.g., ImageNet) and various neural network architectures (e.g., VGG) demonstrates that FAIR-Pruner achieves significant model compression while maintaining high accuracy.
Authors: Abdulmajid Murad, Massimiliano Ruocco
Abstract: Access to comprehensive flight operations data remains severely restricted in aviation due to commercial sensitivity and competitive considerations, hindering the development of predictive models for operational planning. This paper investigates whether synthetic data can effectively replace real operational data for training machine learning models in pre-tactical aviation scenarios-predictions made hours to days before operations using only scheduled flight information. We evaluate four state-of-the-art synthetic data generators on three prediction tasks: aircraft turnaround time, departure delays, and arrival delays. Using a Train on Synthetic, Test on Real (TSTR) methodology on over 1.7 million European flight records, we first validate synthetic data quality through fidelity assessments, then assess both predictive performance and the preservation of operational relationships. Our results show that advanced neural network architectures, specifically transformer-based generators, can retain 94-97% of real-data predictive performance while maintaining feature importance patterns informative for operational decision-making. Our analysis reveals that even with real data, prediction accuracy is inherently limited when only scheduled information is available-establishing realistic baselines for pre-tactical forecasting. These findings suggest that high-quality synthetic data can enable broader access to aviation analytics capabilities while preserving commercial confidentiality, though stakeholders must maintain realistic expectations about pre-tactical prediction accuracy given the stochastic nature of flight operations.
Authors: Guofu Xie, Yunsheng Shi, Hongtao Tian, Ting Yao, Xiao Zhang
Abstract: Reinforcement Learning with Verifiable Rewards (RLVR) has improved the reasoning abilities of Large Language Models (LLMs) by using rule-based binary feedback, helping to mitigate reward hacking. However, current RLVR methods typically treat whole responses as single actions, assigning the same reward to every token. This coarse-grained feedback hampers precise credit assignment, making it hard for models to identify which reasoning steps lead to success or failure, and often results in suboptimal policies and inefficient learning. Methods like PPO provide credit assignment through value estimation, but often yield inaccurate and unverifiable signals due to limited sampling. On the other hand, methods using Process Reward Models can provide step-by-step judgments for each reasoning step, but they require high-quality process supervision labels and are time-consuming when applied in online reinforcement learning (RL). To overcome these limitations, we introduce a simple but efficient method Credit Assignment Policy Optimization (CAPO). Given a reasoning response rollout from the policy model, CAPO directly leverages an off-the-shelf, general-purpose LLM as a Generative Process Reward Model (LLM-as-GenPRM) to generate all step-wise critique by one pass, thereby providing verifiable token-level rewards to refine the tokens that were originally assigned identical rule-based rewards. This enables more fine-grained credit assignment in an effective way. Furthermore, to enhance the accuracy and robustness of CAPO, we employ voting mechanisms that scale with the number of generated critiques. Extensive experiments using different backbones like Llama and Qwen models and in different sizes show that CAPO consistently outperforms supervised learning-based and RL-based fine-tuning methods across six challenging mathematical benchmarks and three out-of-domain benchmarks.
Authors: Boran Zhao, Haiduo Huang, Qiwei Dang, Wenzhe Zhao, Tian Xia, Pengju Ren
Abstract: Training deep neural networks (DNNs) on edge devices has attracted increasing attention due to its potential to address challenges related to domain adaptation and privacy preservation. However, DNNs typically rely on large datasets for training, which results in substantial energy consumption, making the training in edge devices impractical. Some dataset compression methods have been proposed to solve this challenge. For instance, the coreset selection and dataset distillation reduce the training cost by selecting and generating representative samples respectively. Nevertheless, these methods have two significant defects: (1) The necessary of leveraging a DNN model to evaluate the quality of representative samples, which inevitably introduces inductive bias of DNN, resulting in a severe generalization issue; (2) All training images require multiple accesses to the DDR via long-distance PCB connections, leading to substantial energy overhead. To address these issues, inspired by the nonlinear manifold stationary of the human brain, we firstly propose a DNN-free sample-selecting algorithm, called DE-SNE, to improve the generalization issue. Secondly, we innovatively utilize the near-memory computing technique to implement DE-SNE, thus only a small fraction of images need to access the DDR via long-distance PCB. It significantly reduces DDR energy consumption. As a result, we build a novel expedited DNN training system with a more efficient in-place Near-Memory Sampling characteristic for edge devices, dubbed NMS. As far as we know, our NMS is the first DNN-free near-memory sampling technique that can effectively alleviate generalization issues and significantly reduce DDR energy caused by dataset access. The experimental results show that our NMS outperforms the current state-of-the-art (SOTA) approaches, namely DQ, DQAS, and NeSSA, in model accuracy.
Authors: Parth Naik, Harikrishnan N B
Abstract: We propose a novel classification framework grounded in symbolic dynamics and data compression using chaotic maps. The core idea is to model each class by generating symbolic sequences from thresholded real-valued training data, which are then evolved through a one-dimensional chaotic map. For each class, we compute the transition probabilities of symbolic patterns (e.g., `00', `01', `10', and `11' for the second return map) and aggregate these statistics to form a class-specific probabilistic model. During testing phase, the test data are thresholded and symbolized, and then encoded using the class-wise symbolic statistics via back iteration, a dynamical reconstruction technique. The predicted label corresponds to the class yielding the shortest compressed representation, signifying the most efficient symbolic encoding under its respective chaotic model. This approach fuses concepts from dynamical systems, symbolic representations, and compression-based learning. We evaluate the proposed method: \emph{ChaosComp} on both synthetic and real-world datasets, demonstrating competitive performance compared to traditional machine learning algorithms (e.g., macro F1-scores for the proposed method on Breast Cancer Wisconsin = 0.9531, Seeds = 0.9475, Iris = 0.8469 etc.). Rather than aiming for state-of-the-art performance, the goal of this research is to reinterpret the classification problem through the lens of dynamical systems and compression, which are foundational perspectives in learning theory and information processing.
Authors: Joon-Hyun Park, Mujin Cheon, Dong-Yeun Koh
Abstract: The performance of Bayesian optimization (BO), a highly sample-efficient method for expensive black-box problems, is critically governed by the selection of its hyperparameters, including the kernel and acquisition functions. This presents a challenge: an inappropriate combination of these can lead to poor performance and wasted evaluations. While individual improvements to kernel functions (e.g., tree-based kernels, deep kernel learning) and acquisition functions (e.g., multi-step lookahead, tree-based planning) have been explored, the joint and autonomous selection of the best pair of these fundamental hyperparameters has been overlooked. This forces practitioners to rely on heuristics or costly manual training. We propose a simple yet effective framework, BOOST (Bayesian Optimization with Optimal Kernel and Acquisition Function Selection Technique), that automates this selection. BOOST utilizes a lightweight, offline evaluation stage to predict the performance of various kernel-acquisition function pairs and identify the most suitable configuration before expensive evaluations. BOOST partitions data-in-hand into two subsets: a reference subset and a query subset, and it prepares all possible kernel-acquisition pairs from the user's chosen candidates. For each configuration, BOOST conducts internal BO runs using the reference subset, evaluating how effectively each pair guides the search toward the optimum in the unknown query subset, thereby identifying the configuration with the best retrospective performance for future optimization. Experiments on both synthetic benchmark functions and real-world hyperparameter optimization tasks demonstrate that BOOST consistently outperforms standard BO approaches with fixed hyperparameters, highlighting its effectiveness and robustness in diverse problem landscapes.
Authors: V\"ain\"o Yrj\"an\"ainen, Isac Bostr\"om, M{\aa}ns Magnusson, Johan Jonasson
Abstract: Quantifying uncertainty in word embeddings is crucial for reliable inference from textual data. However, existing Bayesian methods such as Hamiltonian Monte Carlo (HMC) and mean-field variational inference (MFVI) are either computationally infeasible for large data or rely on restrictive assumptions. We propose a scalable Gibbs sampler using Polya-Gamma augmentation as well as Laplace approximation and compare them with MFVI and HMC for word embeddings. In addition, we address non-identifiability in word embeddings. Our Gibbs sampler and HMC correctly estimate uncertainties, while MFVI does not, and Laplace approximation only does so on large sample sizes, as expected. Applying the Gibbs sampler to the US Congress and the Movielens datasets, we demonstrate the feasibility on larger real data. Finally, as a result of having draws from the full posterior, we show that the posterior mean of word embeddings improves over maximum a posteriori (MAP) estimates in terms of hold-out likelihood, especially for smaller sampling sizes, further strengthening the need for posterior sampling of word embeddings.
Authors: Wenyuan Liu, Haoqian Meng, Yilun Luo, Peng Zhang, Xindian Ma
Abstract: Quantization significantly accelerates inference in large language models (LLMs) by replacing original high-precision matrices with low-precision counterparts. Recent advances in weight-activation quantization have primarily focused on mapping both weights and activations to the INT4 format. Although the new FP4 Tensor Cores in NVIDIA's Blackwell architecture offer up to 4x speedup over FP16, existing INT4-based kernels fail to fully exploit this capability due to mismatched data formats. To bridge this gap, we propose MicroMix, a co-designed mixed-precision quantization algorithm and matrix multiplication kernel based on Microscaling (MX) data formats. Tailored for the Blackwell architecture, the MicroMix kernel supports arbitrary combinations of MXFP4, MXFP6, and MXFP8 channels, and produces BFloat16 outputs. To achieve a favorable trade-off between accuracy and efficiency for each linear layer, we introduce quantization thresholds that identify activation elements where lower-precision formats (MXFP4 or MXFP6) incur excessive quantization error. Our algorithm selectively allocates higher-precision channels to preserve accuracy while maintaining compute efficiency. MicroMix achieves competitive or superior performance across diverse downstream tasks, including zero-shot and few-shot learning, language modeling, code generation, and mathematical reasoning. On both consumer-grade (RTX 5070Ti laptop) and server-grade (RTX 5090) GPUs, our kernel delivers at least 20% faster execution than TensorRT-FP8. Furthermore, when applied to various Llama and Qwen models, MicroMix consistently improves prefill latency and memory efficiency across a range of batch sizes compared to TensorRT baselines. Our code is available at https://github.com/lwy2020/MicroMix.
Authors: Moritz Piening, Robert Beinert
Abstract: The Gromov--Wasserstein (GW) distance and its fused extension (FGW) are powerful tools for comparing heterogeneous data. Their computation is, however, challenging since both distances are based on non-convex, quadratic optimal transport (OT) problems. Leveraging 1D OT, a sliced version of GW has been proposed to lower the computational burden. Unfortunately, this sliced version is restricted to Euclidean geometry and loses invariance to isometries, strongly limiting its application in practice. To overcome these issues, we propose a novel slicing technique for GW as well as for FGW that is based on an appropriate lower bound, hierarchical OT, and suitable quadrature rules for the underlying 1D OT problems. Our novel sliced FGW significantly reduces the numerical effort while remaining invariant to isometric transformations and allowing the comparison of arbitrary geometries. We show that our new distance actually defines a pseudo-metric for structured spaces that bounds FGW from below and study its interpolation properties between sliced Wasserstein and GW. Since we avoid the underlying quadratic program, our sliced distance is numerically more robust and reliable than the original GW and FGW distance; especially in the context of shape retrieval and graph isomorphism testing.
Authors: Adam Darmanin, Vince Vella
Abstract: Algorithmic trading requires short-term decisions aligned with long-term financial goals. While reinforcement learning (RL) has been explored for such tactical decisions, its adoption remains limited by myopic behavior and opaque policy rationale. In contrast, large language models (LLMs) have recently demonstrated strategic reasoning and multi-modal financial signal interpretation when guided by well-designed prompts. We propose a hybrid system where LLMs generate high-level trading strategies to guide RL agents in their actions. We evaluate (i) the rationale of LLM-generated strategies via expert review, and (ii) the Sharpe Ratio (SR) and Maximum Drawdown (MDD) of LLM-guided agents versus unguided baselines. Results show improved return and risk metrics over standard RL.
Authors: Zuxin Ma, Yunhe Cui, Yongbin Qin
Abstract: Non-uniform structured network pruning methods can effectively reduce Large Language Model (LLM) size by eliminating redundant channels or layers, offering lower performance degradation than uniform strategies. However, existing non-uniform methods rely heavily on manually designed pruning policies (e.g., layer importance and scaling factors), and therefore cannot efficiently adapt to scenarios with dynamic pruning ratio requirements. Additionly, a critical bottleneck -- the time-consuming evaluation of pruning policies -- further limits the feasibility of iteratively and dynamically finding optimal pruning policies. To address these limitations, we propose PPF (Predictive Pruning Framework), a novel pruning framework for LLMs that eliminates manual design dependencies via second-level performance prediction. PPF not only supports real-time pruning decisions under dynamic pruning ratios but is also applicable to static pruning scenarios. It employs an agent for producing adaptive and real-time pruning actions, while a lightweight performance predictor that can evaluate a pruning policy in seconds, significantly speeding up the iterative optimization process. Experiments on Llama2-7B and Llama3-8B show that PPF can generate dynamic/static pruning policies and it reduces perplexity by up to 33.4% (dynamic pruning) and 84.78% (static pruning) over existing methods, outperforming manually designed pruning policies. The performance predictor achieves second-level performance prediction with high accuracy (prediction error < 0.0011). It reduces the mean evaluation latency from minute-level (1 minute and 38.02 seconds of test-set evaluation methods) to second-level (1.52 second), achieving over 64 times speedup. Our code will be available at https://github.com/Ma-zx/PPF .
Authors: Changjie Sheng, Zhichao Zhang, Wei Yao
Abstract: Spectral graph embedding plays a critical role in graph representation learning by generating low-dimensional vector representations from graph spectral information. However, the embedding space of traditional spectral embedding methods often exhibit limited expressiveness, failing to exhaustively capture latent structural features across alternative transform domains. To address this issue, we use the graph fractional Fourier transform to extend the existing state-of-the-art generalized frequency filtering embedding (GEFFE) into fractional domains, giving birth to the generalized fractional filtering embedding (GEFRFE), which enhances embedding informativeness via the graph fractional domain. The GEFRFE leverages graph fractional domain filtering and a nonlinear composition of eigenvector components derived from a fractionalized graph Laplacian. To dynamically determine the fractional order, two parallel strategies are introduced: search-based optimization and a ResNet18-based adaptive learning. Extensive experiments on six benchmark datasets demonstrate that the GEFRFE captures richer structural features and significantly enhance classification performance. Notably, the proposed method retains computational complexity comparable to GEFFE approaches.
Authors: Jialiang Wang, Xiong Zhou, Deming Zhai, Junjun Jiang, Xiangyang Ji, Xianming Liu
Abstract: Noisy labels pose a common challenge for training accurate deep neural networks. To mitigate label noise, prior studies have proposed various robust loss functions to achieve noise tolerance in the presence of label noise, particularly symmetric losses. However, they usually suffer from the underfitting issue due to the overly strict symmetric condition. In this work, we propose a simple yet effective approach for relaxing the symmetric condition, namely $\epsilon$-softmax, which simply modifies the outputs of the softmax layer to approximate one-hot vectors with a controllable error $\epsilon$. Essentially, $\epsilon$-softmax not only acts as an alternative for the softmax layer, but also implicitly plays the crucial role in modifying the loss function. We prove theoretically that $\epsilon$-softmax can achieve noise-tolerant learning with controllable excess risk bound for almost any loss function. Recognizing that $\epsilon$-softmax-enhanced losses may slightly reduce fitting ability on clean datasets, we further incorporate them with one symmetric loss, thereby achieving a better trade-off between robustness and effective learning. Extensive experiments demonstrate the superiority of our method in mitigating synthetic and real-world label noise. The code is available at https://github.com/cswjl/eps-softmax.
Authors: Mirko Konstantin, Moritz Fuchs, Anirban Mukhopadhyay
Abstract: Federated Learning (FL) allows the training of deep neural networks in a distributed and privacy-preserving manner. However, this concept suffers from malfunctioning updates sent by the attending clients that cause global model performance degradation. Reasons for this malfunctioning might be technical issues, disadvantageous training data, or malicious attacks. Most of the current defense mechanisms are meant to require impractical prerequisites like knowledge about the number of malfunctioning updates, which makes them unsuitable for real-world applications. To counteract these problems, we introduce a novel method called Angular Support for Malfunctioning Client Resilience (ASMR), that dynamically excludes malfunctioning clients based on their angular distance. Our novel method does not require any hyperparameters or knowledge about the number of malfunctioning clients. Our experiments showcase the detection capabilities of ASMR in an image classification task on a histopathological dataset, while also presenting findings on the significance of dynamically adapting decision boundaries.
Authors: Khoa Tuan Nguyen, Gaeun Oh, Ho-min Park, Francesca Tozzi, Wouter Willaert, Joris Vankerschaver, Niki Rashidian, Wesley De Neve
Abstract: While 3D medical shape generative models such as diffusion models have shown promise in synthesizing diverse and anatomically plausible structures, the absence of ground truth makes quality evaluation challenging. Existing evaluation metrics commonly measure distributional distances between training and generated sets, while the medical field requires assessing quality at the individual level for each generated shape, which demands labor-intensive expert review. In this paper, we investigate the use of classical machine learning (ML) methods and PointNet as an alternative, interpretable approach for assessing the quality of generated liver shapes. We sample point clouds from the surfaces of the generated liver shapes, extract handcrafted geometric features, and train a group of supervised ML and PointNet models to classify liver shapes as good or bad. These trained models are then used as proxy discriminators to assess the quality of synthetic liver shapes produced by generative models. Our results show that ML-based shape classifiers provide not only interpretable feedback but also complementary insights compared to expert evaluation. This suggests that ML classifiers can serve as lightweight, task-relevant quality metrics in 3D organ shape generation, supporting more transparent and clinically aligned evaluation protocols in medical shape modeling.
Authors: Yuming Ai, Xunkai Li, Jiaqi Chao, Bowen Fan, Zhengyu Wu, Yinlin Zhu, Rong-Hua Li, Guoren Wang
Abstract: The demand for data privacy has led to the development of frameworks like Federated Graph Learning (FGL), which facilitate decentralized model training. However, a significant operational challenge in such systems is adhering to the right to be forgotten. This principle necessitates robust mechanisms for two distinct types of data removal: the selective erasure of specific entities and their associated knowledge from local subgraphs and the wholesale removal of a user's entire dataset and influence. Existing methods often struggle to fully address both unlearning requirements, frequently resulting in incomplete data removal or the persistence of residual knowledge within the system. This work introduces a unified framework, conceived to provide a comprehensive solution to these challenges. The proposed framework employs a bifurcated strategy tailored to the specific unlearning request. For fine-grained Meta Unlearning, it uses prototype gradients to direct the initial local forgetting process, which is then refined by generating adversarial graphs to eliminate any remaining data traces among affected clients. In the case of complete client unlearning, the framework utilizes adversarial graph generation exclusively to purge the departed client's contributions from the remaining network. Extensive experiments on multiple benchmark datasets validate the proposed approach. The framework achieves substantial improvements in model prediction accuracy across both client and meta-unlearning scenarios when compared to existing methods. Furthermore, additional studies confirm its utility as a plug-in module, where it materially enhances the predictive capabilities and unlearning effectiveness of other established methods.
Authors: Kunyu Zhang, Lin Gu, Liangchen Liu, Yingke Chen, Binyang Wang, Jin Yan, Yingying Zhu
Abstract: Many previous studies have proposed extracting image labels from clinical notes to create large-scale medical image datasets at a low cost. However, these approaches inherently suffer from label noise due to uncertainty from the clinical experts. When radiologists and physicians analyze medical images to make diagnoses, they often include uncertainty-aware notes such as ``maybe'' or ``not excluded''. Unfortunately, current text-mining methods overlook these nuances, resulting in the creation of noisy labels. Existing methods for handling noisy labels in medical image analysis, which typically address the problem through post-processing techniques, have largely ignored the important issue of expert-driven uncertainty contributing to label noise. To better incorporate the expert-written uncertainty in clinical notes into medical image analysis and address the label noise issue, we first examine the impact of clinical expert uncertainty on label noise. We then propose a clinical expert uncertainty-aware benchmark, along with a label smoothing method, which significantly improves performance compared to current state-of-the-art approaches.
Authors: Daniela Thyssens, Tim Dernedde, Wilson Sentanoe, Lars Schmidt-Thieme
Abstract: Neural Combinatorial Optimization aims to learn to solve a class of combinatorial problems through data-driven methods and notably through employing neural networks by learning the underlying distribution of problem instances. While, so far neural methods struggle to outperform highly engineered problem specific meta-heuristics, this work explores a novel approach to formulate the distribution of problem instances to learn from and, more importantly, plant a structure in the sampled problem instances. In application to routing problems, we generate large problem instances that represent custom base problem instance distributions from which training instances are sampled. The test instances to evaluate the methods on the routing task consist of unseen problems sampled from the underlying large problem instance. We evaluate representative NCO methods and specialized Operation Research meta heuristics on this novel task and demonstrate that the performance gap between neural routing solvers and highly specialized meta-heuristics decreases when learning from sub-samples drawn from a fixed base node distribution.
Authors: Yao Lai, Souradip Poddar, Sungyoung Lee, Guojin Chen, Mengkang Hu, Bei Yu, Ping Luo, David Z. Pan
Abstract: Despite advances in analog design automation, analog front-end design still heavily depends on expert intuition and iterative simulations, underscoring critical gaps in fully automated optimization for performance-critical applications. Recently, the rapid development of Large Language Models (LLMs) has brought new promise to analog design automation. However, existing work remains in its early stages, and holistic joint optimization for practical end-to-end solutions remains largely unexplored. We propose AnalogCoder-Pro, a unified multimodal LLM-based framework that integrates generative capabilities and optimization techniques to jointly explore circuit topologies and optimize device sizing, automatically generating performance-specific, fully sized schematic netlists. AnalogCoder-Pro employs rejection sampling for fine-tuning LLMs on high-quality synthesized circuit data and introduces a multimodal diagnosis and repair workflow based on functional specifications and waveform images. By leveraging LLMs to interpret generated circuit netlists, AnalogCoder-Pro automates the extraction of critical design parameters and the formulation of parameter spaces, establishing an end-to-end workflow for simultaneous topology generation and device sizing optimization. Extensive experiments demonstrate that these orthogonal approaches significantly improve the success rate of analog circuit design and enhance circuit performance.
Authors: Shunxian Gu, Chaoqun You, Bangbang Ren, Deke Guo
Abstract: The hierarchical architecture of Open Radio Access Network (O-RAN) has enabled a new Federated Learning (FL) paradigm that trains models using data from non- and near-real-time (near-RT) Radio Intelligent Controllers (RICs). However, the ever-increasing model size leads to longer training time, jeopardizing the deadline requirements for both non-RT and near-RT RICs. To address this issue, split federated learning (SFL) offers an approach by offloading partial model layers from near-RT-RIC to high-performance non-RT-RIC. Nonetheless, its deployment presents two challenges: (i) Frequent data/gradient transfers between near-RT-RIC and non-RT-RIC in SFL incur significant communication cost in O-RAN. (ii) Proper allocation of computational and communication resources in O-RAN is vital to satisfying the deadline and affects SFL convergence. Therefore, we propose SplitMe, an SFL framework that exploits mutual learning to alternately and independently train the near-RT-RIC's model and the non-RT-RIC's inverse model, eliminating frequent transfers. The ''inverse'' of the inverse model is derived via a zeroth-order technique to integrate the final model. Then, we solve a joint optimization problem for SplitMe to minimize overall resource costs with deadline-aware selection of near-RT-RICs and adaptive local updates. Our numerical results demonstrate that SplitMe remarkably outperforms FL frameworks like SFL, FedAvg and O-RANFed regarding costs and convergence.
Authors: Xi Chen, Jianchuan Yang, Junjie Zhang, Runnan Yang, Xu Liu, Hong Wang, Ziyu Ren, Wenqi Hu
Abstract: Physics-Informed Neural Networks (PINNs) are effective for solving PDEs by incorporating physical laws into the learning process. However, they face challenges with irregular boundaries, leading to instability and slow convergence due to inconsistent normalization, inaccurate boundary enforcement, and imbalanced loss terms. A common solution is to map the domain to a regular space, but traditional methods rely on case-specific meshes and simple geometries, limiting their compatibility with modern frameworks. To overcome these limitations, we introduce JacobiNet, a neural network-based coordinate transformation method that learns continuous, differentiable mappings from supervised point pairs. Utilizing lightweight MLPs, JacobiNet allows for direct Jacobian computation via autograd and integrates seamlessly with downstream PINNs, enabling end-to-end differentiable PDE solving without the need for meshing or explicit Jacobian computation. JacobiNet effectively addresses normalization challenges, facilitates hard constraints of boundary conditions, and mitigates the long-standing imbalance among loss terms. It demonstrates significant improvements, reducing the relative L2 error from 0.287-0.637 to 0.013-0.039, achieving an average accuracy improvement of 18.3*. In vessel-like domains, it enables rapid mapping for unseen geometries, improving prediction accuracy by 3.65* and achieving over 10* speedup, showcasing its generalization, accuracy, and efficiency.
Authors: Valeria Ruscio, Umberto Nanni, Fabrizio Silvestri
Abstract: Attention sink (AS) is a consistent pattern in transformer attention maps where certain tokens (often special tokens or positional anchors) disproportionately attract attention from other tokens. We show that in transformers, AS is not an architectural artifact, but it is the manifestation of a fundamental geometric principle: the establishment of reference frames that anchor representational spaces. We analyze several architectures and identify three distinct reference frame types, centralized, distributed, and bidirectional, that correlate with the attention sink phenomenon. We show that they emerge during the earliest stages of training as optimal solutions to the problem of establishing stable coordinate systems in high-dimensional spaces. We show the influence of architecture components, particularly position encoding implementations, on the specific type of reference frame. This perspective transforms our understanding of transformer attention mechanisms and provides insights for both architecture design and the relationship with AS.
Authors: Nys Tjade Siegel, James H. Cole, Mohamad Habes, Stefan Haufe, Kerstin Ritter, Marc-Andr\'e Schulz
Abstract: Trustworthy interpretation of deep learning models is critical for neuroimaging applications, yet commonly used Explainable AI (XAI) methods lack rigorous validation, risking misinterpretation. We performed the first large-scale, systematic comparison of XAI methods on ~45,000 structural brain MRIs using a novel XAI validation framework. This framework establishes verifiable ground truth by constructing prediction tasks with known signal sources - from localized anatomical features to subject-specific clinical lesions - without artificially altering input images. Our analysis reveals systematic failures in two of the most widely used methods: GradCAM consistently failed to localize predictive features, while Layer-wise Relevance Propagation generated extensive, artifactual explanations that suggest incompatibility with neuroimaging data characteristics. Our results indicate that these failures stem from a domain mismatch, where methods with design principles tailored to natural images require substantial adaptation for neuroimaging data. In contrast, the simpler, gradient-based method SmoothGrad, which makes fewer assumptions about data structure, proved consistently accurate, suggesting its conceptual simplicity makes it more robust to this domain shift. These findings highlight the need for domain-specific adaptation and validation of XAI methods, suggest that interpretations from prior neuroimaging studies using standard XAI methodology warrant re-evaluation, and provide urgent guidance for practical application of XAI in neuroimaging.
Authors: Javier Fumanal-Idocin, Raquel Fernandez-Peralta, Javier Andreu-Perez
Abstract: Dynamic feature selection (DFS) offers a compelling alternative to traditional, static feature selection by adapting the selected features to each individual sample. Unlike classical methods that apply a uniform feature set, DFS customizes feature selection per sample, providing insight into the decision-making process for each case. DFS is especially significant in settings where decision transparency is key, i.e., clinical decisions; however, existing methods use opaque models, which hinder their applicability in real-life scenarios. This paper introduces a novel approach leveraging a rule-based system as a base classifier for the DFS process, which enhances decision interpretability compared to neural estimators. We also show how this method provides a quantitative measure of uncertainty for each feature query and can make the feature selection process computationally lighter by constraining the feature search space. We also discuss when greedy selection of conditional mutual information is equivalent to selecting features that minimize the difference with respect to the global model predictions. Finally, we demonstrate the competitive performance of our rule-based DFS approach against established and state-of-the-art greedy and RL methods, which are mostly considered opaque, compared to our explainable rule-based system.
Authors: Yilun Liu, Yunpu Ma, Yuetian Lu, Shuo Chen, Zifeng Ding, Volker Tresp
Abstract: Mixture-of-Experts (MoE) benefits from a dynamic routing mechanism among their specialized experts, which existing Parameter- Efficient Fine-Tuning (PEFT) strategies fail to leverage. This motivates us to investigate whether adaptation modules themselves should incorporate routing mechanisms to align with MoE's multi-expert architecture. We analyze dynamics of core components when applying PEFT to MoE language models and examine how different routing strategies affect adaptation effectiveness. Extensive experiments adapting OLMoE-1B-7B and Mixtral-8x7B on various commonsense and math reasoning tasks validate the performance and efficiency of our routed approach. We identify the optimal configurations for different scenarios and provide empirical analyses with practical insights to facilitate better PEFT and MoE applications.
Authors: Xudong Wang, Tongxin Li, Chris Ding, Jicong Fan
Abstract: Graph data often exhibits complex geometric heterogeneity, where structures with varying local curvature, such as tree-like hierarchies and dense communities, coexist within a single network. Existing geometric GNNs, which embed graphs into single fixed-curvature manifolds or discrete product spaces, struggle to capture this diversity. We introduce Adaptive Riemannian Graph Neural Networks (ARGNN), a novel framework that learns a continuous and anisotropic Riemannian metric tensor field over the graph. It allows each node to determine its optimal local geometry, enabling the model to fluidly adapt to the graph's structural landscape. Our core innovation is an efficient parameterization of the node-wise metric tensor, specializing to a learnable diagonal form that captures directional geometric information while maintaining computational tractability. To ensure geometric regularity and stable training, we integrate a Ricci flow-inspired regularization that smooths the learned manifold. Theoretically, we establish the rigorous geometric evolution convergence guarantee for ARGNN and provide a continuous generalization that unifies prior fixed or mixed-curvature GNNs. Empirically, our method demonstrates superior performance on both homophilic and heterophilic benchmark datasets with the ability to capture diverse structures adaptively. Moreover, the learned geometries both offer interpretable insights into the underlying graph structure and empirically corroborate our theoretical analysis.
Authors: Siyi Liu, Yujia Zheng, Yongqi Zhang
Abstract: The application of machine learning on tabular data in specialized domains is severely limited by data scarcity. While generative models offer a solution, traditional methods falter in low-data regimes, and recent Large Language Models (LLMs) often ignore the explicit dependency structure of tabular data, leading to low-fidelity synthetics. To address these limitations, we introduce StructSynth, a novel framework that integrates the generative power of LLMs with robust structural control. StructSynth employs a two-stage architecture. First, it performs explicit structure discovery to learn a Directed Acyclic Graph (DAG) from the available data. Second, this learned structure serves as a high-fidelity blueprint to steer the LLM's generation process, forcing it to adhere to the learned feature dependencies and thereby ensuring the generated data respects the underlying structure by design. Our extensive experiments demonstrate that StructSynth produces synthetic data with significantly higher structural integrity and downstream utility than state-of-the-art methods. It proves especially effective in challenging low-data scenarios, successfully navigating the trade-off between privacy preservation and statistical fidelity.
Authors: Jiayin Jin, Zhimeng Pan, Yang Tang, Jiarui Feng, Kungang Li, Chongyuan Xiang, Jiacheng Li, Runze Su, Siping Ji, Han Sun, Ling Leng, Prathibha Deshikachar
Abstract: Graph Neural Networks (GNN) have been extensively applied to industry recommendation systems, as seen in models like GraphSage\cite{GraphSage}, TwHIM\cite{TwHIM}, LiGNN\cite{LiGNN} etc. In these works, graphs were constructed based on users' activities on the platforms, and various graph models were developed to effectively learn node embeddings. In addition to users' onsite activities, their offsite conversions are crucial for Ads models to capture their shopping interest. To better leverage offsite conversion data and explore the connection between onsite and offsite activities, we constructed a large-scale heterogeneous graph based on users' onsite ad interactions and opt-in offsite conversion activities. Furthermore, we introduced TransRA (TransR\cite{TransR} with Anchors), a novel Knowledge Graph Embedding (KGE) model, to more efficiently integrate graph embeddings into Ads ranking models. However, our Ads ranking models initially struggled to directly incorporate Knowledge Graph Embeddings (KGE), and only modest gains were observed during offline experiments. To address this challenge, we employed the Large ID Embedding Table technique and innovated an attention based KGE finetuning approach within the Ads ranking models. As a result, we observed a significant AUC lift in Click-Through Rate (CTR) and Conversion Rate (CVR) prediction models. Moreover, this framework has been deployed in Pinterest's Ads Engagement Model and contributed to $2.69\%$ CTR lift and $1.34\%$ CPC reduction. We believe the techniques presented in this paper can be leveraged by other large-scale industrial models.
Authors: Ali Forootani, Mohammad Khosravi, Masoud Barati
Abstract: Time series forecasting plays a vital role across scientific, industrial, and environmental domains, especially when dealing with high-dimensional and nonlinear systems. While Transformer-based models have recently achieved state-of-the-art performance in long-range forecasting, they often suffer from interpretability issues and instability in the presence of noise or dynamical uncertainty. In this work, we propose DeepKoopFormer, a principled forecasting framework that combines the representational power of Transformers with the theoretical rigor of Koopman operator theory. Our model features a modular encoder-propagator-decoder structure, where temporal dynamics are learned via a spectrally constrained, linear Koopman operator in a latent space. We impose structural guarantees-such as bounded spectral radius, Lyapunov based energy regularization, and orthogonal parameterization to ensure stability and interpretability. Comprehensive evaluations are conducted on both synthetic dynamical systems, real-world climate dataset (wind speed and surface pressure), financial time series (cryptocurrency), and electricity generation dataset using the Python package that is prepared for this purpose. Across all experiments, DeepKoopFormer consistently outperforms standard LSTM and baseline Transformer models in terms of accuracy, robustness to noise, and long-term forecasting stability. These results establish DeepKoopFormer as a flexible, interpretable, and robust framework for forecasting in high dimensional and dynamical settings.
Authors: Riccardo Francia, Maurizio Leone, Giorgio Leonardi, Stefania Montani, Marzio Pennisi, Manuel Striani, Sandra D'Alfonso
Abstract: Medical datasets are typically affected by issues such as missing values, class imbalance, a heterogeneous feature types, and a high number of features versus a relatively small number of samples, preventing machine learning models from obtaining proper results in classification and regression tasks. This paper introduces AutoML-Med, an Automated Machine Learning tool specifically designed to address these challenges, minimizing user intervention and identifying the optimal combination of preprocessing techniques and predictive models. AutoML-Med's architecture incorporates Latin Hypercube Sampling (LHS) for exploring preprocessing methods, trains models using selected metrics, and utilizes Partial Rank Correlation Coefficient (PRCC) for fine-tuned optimization of the most influential preprocessing steps. Experimental results demonstrate AutoML-Med's effectiveness in two different clinical settings, achieving higher balanced accuracy and sensitivity, which are crucial for identifying at-risk patients, compared to other state-of-the-art tools. AutoML-Med's ability to improve prediction results, especially in medical datasets with sparse data and class imbalance, highlights its potential to streamline Machine Learning applications in healthcare.
Authors: Austin Rockman
Abstract: We demonstrate that a single 3x3 convolutional kernel can produce emergent audio effects when trained on 200 samples from a personalized corpus. We achieve this through two key techniques: (1) Conditioning Aware Kernels (CAK), where output = input + (learned_pattern x control), with a soft-gate mechanism supporting identity preservation at zero control; and (2) AuGAN (Audit GAN), which reframes adversarial training from "is this real?" to "did you apply the requested value?" Rather than learning to generate or detect forgeries, our networks cooperate to verify control application, discovering unique transformations. The learned kernel exhibits a diagonal structure creating frequency-dependent temporal shifts that are capable of producing musical effects based on input characteristics. Our results show the potential of adversarial training to discover audio transformations from minimal data, enabling new approaches to effect design.
Authors: Jiaxi Li, Lu Yin, Li Shen, Jinjin Xu, Liwu Xu, Tianjin Huang, Wenwu Wang, Shiwei Liu, Xilu Wang
Abstract: While large language models (LLMs) have achieved remarkable performance across a wide range of tasks, their massive scale incurs prohibitive computational and memory costs for pre-training from scratch. Recent studies have investigated the use of low-rank parameterization as a means of reducing model size and training cost. In this context, sparsity is often employed as a complementary technique to recover important information lost in low-rank compression by capturing salient features in the residual space. However, existing approaches typically combine low-rank and sparse components in a simplistic or ad hoc manner, often resulting in undesirable performance degradation compared to full-rank training. In this paper, we propose \textbf{LO}w-rank and \textbf{S}parse pre-\textbf{T}raining (\textbf{LOST}) for LLMs, a novel method that ingeniously integrates low-rank and sparse structures to enable effective training of LLMs from scratch under strict efficiency constraints. LOST applies singular value decomposition to weight matrices, preserving the dominant low-rank components, while allocating the remaining singular values to construct channel-wise sparse components to complement the expressiveness of low-rank training. We evaluate LOST on LLM pretraining ranging from 60M to 7B parameters. Our experiments show that LOST achieves competitive or superior performance compared to full-rank models, while significantly reducing both memory and compute overhead. Moreover, Code is available at \href{https://github.com/JiaxiLi1/LOST-Low-rank-and-Sparse-Training-for-Large-Language-Models}{LOST Repo}
URLs: https://github.com/JiaxiLi1/LOST-Low-rank-and-Sparse-Training-for-Large-Language-Models
Authors: Jie Cao, Michael Tanana, Zac E. Imel, Eric Poitras, David C. Atkins, Vivek Srikumar
Abstract: Automatically analyzing dialogue can help understand and guide behavior in domains such as counseling, where interactions are largely mediated by conversation. In this paper, we study modeling behavioral codes used to asses a psychotherapy treatment style called Motivational Interviewing (MI), which is effective for addressing substance abuse and related problems. Specifically, we address the problem of providing real-time guidance to therapists with a dialogue observer that (1) categorizes therapist and client MI behavioral codes and, (2) forecasts codes for upcoming utterances to help guide the conversation and potentially alert the therapist. For both tasks, we define neural network models that build upon recent successes in dialogue modeling. Our experiments demonstrate that our models can outperform several baselines for both tasks. We also report the results of a careful analysis that reveals the impact of the various network design tradeoffs for modeling therapy dialogue.
Authors: Jie Cao, Abhijit Suresh, Jennifer Jacobs, Charis Clevenger, Amanda Howard, Chelsea Brown, Brent Milne, Tom Fischaber, Tamara Sumner, James H. Martin
Abstract: Human tutoring interventions play a crucial role in supporting student learning, improving academic performance, and promoting personal growth. This paper focuses on analyzing mathematics tutoring discourse using talk moves - a framework of dialogue acts grounded in Accountable Talk theory. However, scaling the collection, annotation, and analysis of extensive tutoring dialogues to develop machine learning models is a challenging and resource-intensive task. To address this, we present SAGA22, a compact dataset, and explore various modeling strategies, including dialogue context, speaker information, pretraining datasets, and further fine-tuning. By leveraging existing datasets and models designed for classroom teaching, our results demonstrate that supplementary pretraining on classroom data enhances model performance in tutoring settings, particularly when incorporating longer context and speaker information. Additionally, we conduct extensive ablation studies to underscore the challenges in talk move modeling.
Authors: Jannatun Naim, Jie Cao, Fareen Tasneem, Jennifer Jacobs, Brent Milne, James Martin, Tamara Sumner
Abstract: Effective feedback is essential for refining instructional practices in mathematics education, and researchers often turn to advanced natural language processing (NLP) models to analyze classroom dialogues from multiple perspectives. However, utterance-level discourse analysis encounters two primary challenges: (1) multifunctionality, where a single utterance may serve multiple purposes that a single tag cannot capture, and (2) the exclusion of many utterances from domain-specific discourse move classifications, leading to their omission in feedback. To address these challenges, we proposed a multi-perspective discourse analysis that integrates domain-specific talk moves with dialogue act (using the flattened multi-functional SWBD-MASL schema with 43 tags) and discourse relation (applying Segmented Discourse Representation Theory with 16 relations). Our top-down analysis framework enables a comprehensive understanding of utterances that contain talk moves, as well as utterances that do not contain talk moves. This is applied to two mathematics education datasets: TalkMoves (teaching) and SAGA22 (tutoring). Through distributional unigram analysis, sequential talk move analysis, and multi-view deep dive, we discovered meaningful discourse patterns, and revealed the vital role of utterances without talk moves, demonstrating that these utterances, far from being mere fillers, serve crucial functions in guiding, acknowledging, and structuring classroom discourse. These insights underscore the importance of incorporating discourse relations and dialogue acts into AI-assisted education systems to enhance feedback and create more responsive learning environments. Our framework may prove helpful for providing human educator feedback, but also aiding in the development of AI agents that can effectively emulate the roles of both educators and students.
Authors: Lyle Regenwetter, Yazan Abu Obaideh, Fabien Chiotti, Ioanna Lykourentzou, Faez Ahmed
Abstract: We introduce Bike-Bench, an engineering design benchmark for evaluating generative models on problems with multiple real-world objectives and constraints. As generative AI's reach continues to grow, evaluating its capability to understand physical laws, human guidelines, and hard constraints grows increasingly important. Engineering product design lies at the intersection of these difficult tasks, providing new challenges for AI capabilities. Bike-Bench evaluates AI models' capability to generate designs that not only resemble the dataset, but meet specific performance objectives and constraints. To do so, Bike-Bench quantifies a variety of human-centered and multiphysics performance characteristics, such as aerodynamics, ergonomics, structural mechanics, human-rated usability, and similarity to subjective text or image prompts. Supporting the benchmark are several datasets of simulation results, a dataset of 10K human-rated bicycle assessments, and a synthetically-generated dataset of 1.4M designs, each with a parametric, CAD/XML, SVG, and PNG representation. Bike-Bench is uniquely configured to evaluate tabular generative models, LLMs, design optimization, and hybrid algorithms side-by-side. Our experiments indicate that LLMs and tabular generative models fall short of optimization and optimization-augmented generative models in both validity and optimality scores, suggesting significant room for improvement. We hope Bike-Bench, a first-of-its-kind benchmark, will help catalyze progress in generative AI for constrained multi-objective engineering design problems. Code, data, and other resources are published at decode.mit.edu/projects/bikebench/.
Authors: Florian Felten, Gabriel Apaza, Gerhard Br\"aunlich, Cashen Diniz, Xuliang Dong, Arthur Drake, Milad Habibi, Nathaniel J. Hoffman, Matthew Keeler, Soheyl Massoudi, Francis G. VanGessel, Mark Fuge
Abstract: Engineering design optimization seeks to automatically determine the shapes, topologies, or parameters of components that maximize performance under given conditions. This process often depends on physics-based simulations, which are difficult to install, computationally expensive, and require domain-specific expertise. To mitigate these challenges, we introduce EngiBench, the first open-source library and datasets spanning diverse domains for data-driven engineering design. EngiBench provides a unified API and a curated set of benchmarks -- covering aeronautics, heat conduction, photonics, and more -- that enable fair, reproducible comparisons of optimization and machine learning algorithms, such as generative or surrogate models. We also release EngiOpt, a companion library offering a collection of such algorithms compatible with the EngiBench interface. Both libraries are modular, letting users plug in novel algorithms or problems, automate end-to-end experiment workflows, and leverage built-in utilities for visualization, dataset generation, feasibility checks, and performance analysis. We demonstrate their versatility through experiments comparing state-of-the-art techniques across multiple engineering design problems, an undertaking that was previously prohibitively time-consuming to perform. Finally, we show that these problems pose significant challenges for standard machine learning methods due to highly sensitive and constrained design manifolds.
Authors: Andrea Gayon-Lombardo, Ehecatl A. del Rio-Chanona, Catalina A. Pino-Munoz, Nigel P. Brandon
Abstract: The generation of multiphase porous electrode microstructures with optimum morphological and transport properties is essential in the design of improved electrochemical energy storage devices, such as lithium-ion batteries. Electrode characteristics directly influence battery performance by acting as the main sites where the electrochemical reactions coupled with transport processes occur. This work presents a generation-optimisation closed-loop algorithm for the design of microstructures with tailored properties. A deep convolutional Generative Adversarial Network is used as a deep kernel and employed to generate synthetic three-phase three-dimensional images of a porous lithium-ion battery cathode material. A Gaussian Process Regression uses the latent space of the generator and serves as a surrogate model to correlate the morphological and transport properties of the synthetic microstructures. This surrogate model is integrated into a deep kernel Bayesian optimisation framework, which optimises cathode properties as a function of the latent space of the generator. A set of objective functions were defined to perform the maximisation of morphological properties (e.g., volume fraction, specific surface area) and transport properties (relative diffusivity). We demonstrate the ability to perform simultaneous maximisation of correlated properties (specific surface area and relative diffusivity), as well as constrained optimisation of these properties. This is the maximisation of morphological or transport properties constrained by constant values of the volume fraction of the phase of interest. Visualising the optimised latent space reveals its correlation with morphological properties, enabling the fast generation of visually realistic microstructures with customised properties.
Authors: Songlin Xu, Xinyu Zhang
Abstract: In this paper, we introduce an AI-mediated framework that can provide intelligent feedback to augment human cognition. Specifically, we leverage deep reinforcement learning (DRL) to provide adaptive time pressure feedback to improve user performance in a math arithmetic task. Time pressure feedback could either improve or deteriorate user performance by regulating user attention and anxiety. Adaptive time pressure feedback controlled by a DRL policy according to users' real-time performance could potentially solve this trade-off problem. However, the DRL training and hyperparameter tuning may require large amounts of data and iterative user studies. Therefore, we propose a dual-DRL framework that trains a regulation DRL agent to regulate user performance by interacting with another simulation DRL agent that mimics user cognition behaviors from an existing dataset. Our user study demonstrates the feasibility and effectiveness of the dual-DRL framework in augmenting user performance, in comparison to the baseline group.
Authors: Yuemin Ma, Uksang Yoo, Yunchao Yao, Shahram Najam Syed, Luca Bondi, Jonathan Francis, Jean Oh, Jeffrey Ichnowski
Abstract: Accurately estimating hand pose and hand-object contact events is essential for robot data-collection, immersive virtual environments, and biomechanical analysis, yet remains challenging due to visual occlusion, subtle contact cues, limitations in vision-only sensing, and the lack of accessible and flexible tactile sensing. We therefore introduce VibeMesh, a novel wearable system that fuses vision with active acoustic sensing for dense, per-vertex hand contact and pose estimation. VibeMesh integrates a bone-conduction speaker and sparse piezoelectric microphones, distributed on a human hand, emitting structured acoustic signals and capturing their propagation to infer changes induced by contact. To interpret these cross-modal signals, we propose a graph-based attention network that processes synchronized audio spectra and RGB-D-derived hand meshes to predict contact with high spatial resolution. We contribute: (i) a lightweight, non-intrusive visuo-acoustic sensing platform; (ii) a cross-modal graph network for joint pose and contact inference; (iii) a dataset of synchronized RGB-D, acoustic, and ground-truth contact annotations across diverse manipulation scenarios; and (iv) empirical results showing that VibeMesh outperforms vision-only baselines in accuracy and robustness, particularly in occluded or static-contact settings.
Authors: Minghan Li, Congcong Wen, Yu Tian, Min Shi, Yan Luo, Hao Huang, Yi Fang, Mengyu Wang
Abstract: Fairness remains a critical concern in healthcare, where unequal access to services and treatment outcomes can adversely affect patient health. While Federated Learning (FL) presents a collaborative and privacy-preserving approach to model training, ensuring fairness is challenging due to heterogeneous data across institutions, and current research primarily addresses non-medical applications. To fill this gap, we establish the first experimental benchmark for fairness in medical FL, evaluating six representative FL methods across diverse demographic attributes and imaging modalities. We introduce FairFedMed, the first medical FL dataset specifically designed to study group fairness (i.e., demographics). It comprises two parts: FairFedMed-Oph, featuring 2D fundus and 3D OCT ophthalmology samples with six demographic attributes; and FairFedMed-Chest, which simulates real cross-institutional FL using subsets of CheXpert and MIMIC-CXR. Together, they support both simulated and real-world FL across diverse medical modalities and demographic groups. Existing FL models often underperform on medical images and overlook fairness across demographic groups. To address this, we propose FairLoRA, a fairness-aware FL framework based on SVD-based low-rank approximation. It customizes singular value matrices per demographic group while sharing singular vectors, ensuring both fairness and efficiency. Experimental results on the FairFedMed dataset demonstrate that FairLoRA not only achieves state-of-the-art performance in medical image classification but also significantly improves fairness across diverse populations. Our code and dataset can be accessible via link: https://wang.hms.harvard.edu/fairfedmed/.
Authors: Nicholas Fidalgo, Puyuan Ye
Abstract: Modern key-value stores rely heavily on Log-Structured Merge (LSM) trees for write optimization, but this design introduces significant read amplification. Auxiliary structures like Bloom filters help, but impose memory costs that scale with tree depth and dataset size. Recent advances in learned data structures suggest that machine learning models can augment or replace these components, trading handcrafted heuristics for data-adaptive behavior. In this work, we explore two approaches for integrating learned predictions into the LSM-tree lookup path. The first uses a classifier to selectively bypass Bloom filter probes for irrelevant levels, aiming to reduce average-case query latency. The second replaces traditional Bloom filters with compact learned models and small backup filters, targeting memory footprint reduction without compromising correctness. We implement both methods atop a Monkey-style LSM-tree with leveled compaction, per-level Bloom filters, and realistic workloads. Our experiments show that the classifier reduces GET latency by up to 2.28x by skipping over 30% of Bloom filter checks with high precision, though it incurs a modest false-negative rate. The learned Bloom filter design achieves zero false negatives and retains baseline latency while cutting memory usage per level by 70-80%. Together, these designs illustrate complementary trade-offs between latency, memory, and correctness, and highlight the potential of learned index components in write-optimized storage systems.
Authors: Hagyeong Shin, Binoy Robin Dalal, Iwona Bialynicka-Birula, Navjot Matharu, Ryan Muir, Xingwei Yang, Samuel W. K. Wong
Abstract: Large language models (LLMs) are known to hallucinate, producing natural language outputs that are not grounded in the input, reference materials, or real-world knowledge. In enterprise applications where AI features support business decisions, such hallucinations can be particularly detrimental. LLMs that analyze and summarize contact center conversations introduce a unique set of challenges for factuality evaluation, because ground-truth labels often do not exist for analytical interpretations about sentiments captured in the conversation and root causes of the business problems. To remedy this, we first introduce a \textbf{3D} -- \textbf{Decompose, Decouple, Detach} -- paradigm in the human annotation guideline and the LLM-judges' prompt to ground the factuality labels in linguistically-informed evaluation criteria. We then introduce \textbf{FECT}, a novel benchmark dataset for \textbf{F}actuality \textbf{E}valuation of Interpretive AI-Generated \textbf{C}laims in Contact Center Conversation \textbf{T}ranscripts, labeled under our 3D paradigm. Lastly, we report our findings from aligning LLM-judges on the 3D paradigm. Overall, our findings contribute a new approach for automatically evaluating the factuality of outputs generated by an AI system for analyzing contact center conversations.
Authors: Fali Wang, Hui Liu, Zhenwei Dai, Jingying Zeng, Zhiwei Zhang, Zongyu Wu, Chen Luo, Zhen Li, Xianfeng Tang, Qi He, Suhang Wang
Abstract: Test-time scaling (TTS) enhances the performance of large language models (LLMs) by allocating additional compute resources during inference. However, existing research primarily investigates TTS in single-stage tasks; while many real-world problems are multi-stage complex tasks, composed of a sequence of heterogeneous subtasks with each subtask requires LLM of specific capability. Therefore, we study a novel problem: the test-time compute-optimal scaling in multi-stage complex tasks, aiming to select suitable models and allocate budgets per subtask to maximize overall performance. TTS in multi-stage tasks introduces two fundamental challenges: (i) The combinatorial search space of model and budget allocations, combined with the high cost of inference, makes brute-force search impractical. (ii) The optimal model and budget allocations across subtasks are interdependent, increasing the complexity of the compute-optimal search. To address this gap, we conduct extensive pilot experiments on four tasks across six datasets, deriving three empirical insights characterizing the behavior of LLMs in multi-stage complex tasks. Informed by these insights, we propose AgentTTS, an LLM-agent-based framework that autonomously searches for compute-optimal allocations through iterative feedback-driven interactions with the execution environment. Experimental results demonstrate that AgentTTS significantly outperforms traditional and other LLM-based baselines in search efficiency, and shows improved robustness to varying training set sizes and enhanced interpretability.
Authors: Luiz Emilio Allem, Konstantin Avrachenkov, Carlos Hoppen, Hariprasad Manjunath, Lucas Siviero Sibemberg
Abstract: In this paper, we consider the soft geometric block model (SGBM) with a fixed number $k \geq 2$ of homogeneous communities in the dense regime, and we introduce a spectral clustering algorithm for community recovery on graphs generated by this model. Given such a graph, the algorithm produces an embedding into $\mathbb{R}^{k-1}$ using the eigenvectors associated with the $k-1$ eigenvalues of the adjacency matrix of the graph that are closest to a value determined by the parameters of the model. It then applies $k$-means clustering to the embedding. We prove weak consistency and show that a simple local refinement step ensures strong consistency. A key ingredient is an application of a non-standard version of Davis-Kahan theorem to control eigenspace perturbations when eigenvalues are not simple. We also analyze the limiting spectrum of the adjacency matrix, using a combination of combinatorial and matrix techniques.
Authors: Tsuyoshi Id\'e, Kohei Miyaguchi
Abstract: Cross-process root-cause analysis of wafer defects is among the most critical yet challenging tasks in semiconductor manufacturing due to the heterogeneity and combinatorial nature of processes along the processing route. This paper presents a new framework for wafer defect root cause analysis, called Potential Loss Analysis (PLA), as a significant enhancement of the previously proposed partial trajectory regression approach. The PLA framework attributes observed high wafer defect densities to upstream processes by comparing the best possible outcomes generated by partial processing trajectories. We show that the task of identifying the best possible outcome can be reduced to solving a Bellman equation. Remarkably, the proposed framework can simultaneously solve the prediction problem for defect density as well as the attribution problem for defect scores. We demonstrate the effectiveness of the proposed framework using real wafer history data.
Authors: Abeer Dyoub, Ivan Letteri, Francesca A. Lisi
Abstract: The emergence of Symbiotic AI (SAI) introduces new challenges to ethical decision-making as it deepens human-AI collaboration. As symbiosis grows, AI systems pose greater ethical risks, including harm to human rights and trust. Ethical Risk Assessment (ERA) thus becomes crucial for guiding decisions that minimize such risks. However, ERA is hindered by uncertainty, vagueness, and incomplete information, and morality itself is context-dependent and imprecise. This motivates the need for a flexible, transparent, yet robust framework for ERA. Our work supports ethical decision-making by quantitatively assessing and prioritizing multiple ethical risks so that artificial agents can select actions aligned with human values and acceptable risk levels. We introduce ff4ERA, a fuzzy framework that integrates Fuzzy Logic, the Fuzzy Analytic Hierarchy Process (FAHP), and Certainty Factors (CF) to quantify ethical risks via an Ethical Risk Score (ERS) for each risk type. The final ERS combines the FAHP-derived weight, propagated CF, and risk level. The framework offers a robust mathematical approach for collaborative ERA modeling and systematic, step-by-step analysis. A case study confirms that ff4ERA yields context-sensitive, ethically meaningful risk scores reflecting both expert input and sensor-based evidence. Risk scores vary consistently with relevant factors while remaining robust to unrelated inputs. Local sensitivity analysis shows predictable, mostly monotonic behavior across perturbations, and global Sobol analysis highlights the dominant influence of expert-defined weights and certainty factors, validating the model design. Overall, the results demonstrate ff4ERA ability to produce interpretable, traceable, and risk-aware ethical assessments, enabling what-if analyses and guiding designers in calibrating membership functions and expert judgments for reliable ethical decision support.
Authors: Rajeev Patwari, Ashish Sirasao, Devleena Das
Abstract: Large language models (LLMs) have been increasingly deployed as local agents on personal devices with CPUs, NPUs and integrated GPUs. However, forecasting inference performance on devices with such heterogeneity remains challenging due to the dynamic compute and memory demands. Existing approaches rely on GPU benchmarking or machine learning-based latency predictors, which are often hardware-specific and lack generalizability. To this end, we introduce LIFE, a lightweight and modular analytical framework that is comprised of modular analytical model of operators, configurable to characterize LLM inference workloads in a hardware and dataset-agnostic manner. LIFE characterizes the influence of software and model optimizations, such as quantization, KV cache compression, LoRA adapters, chunked prefill, different attentions, and operator fusion, on performance metrics such as time-to-first-token (TTFT), time-per-output-token (TPOT) and tokens-per-second (TPS). LIFE enables performance forecasting using only hardware specifications, such as TOPS and memory bandwidth, without requiring extensive dataset benchmarking. We validate LIFE's forecasting with inference on AMD Ryzen CPUs, NPUs, iGPUs and NVIDIA V100 GPUs, with Llama2-7B variants, demonstrating the utility of LIFE in forecasting LLM performance through lens of system efficiency to enable efficient LLM deployment across different hardware platforms.
Authors: Terry Yue Zhuo, Dingmin Wang, Hantian Ding, Varun Kumar, Zijian Wang
Abstract: Large Language Models (LLMs) have achieved remarkable success in software engineering tasks when trained with executable runtime environments, particularly in resolving GitHub issues. However, such runtime environments are often unavailable in other domains, especially cybersecurity, where challenge configurations and execution contexts are ephemeral or restricted. We present Cyber-Zero, the first runtime-free framework for synthesizing high-quality agent trajectories to train cybersecurity LLMs. Cyber-Zero leverages publicly available CTF writeups and employs persona-driven LLM simulation to reverse-engineer runtime behaviors and generate realistic, long-horizon interaction sequences without actual environments. Using trajectories synthesized by Cyber-Zero, we train LLM-based agents that achieve up to 13.1% absolute performance gains over baseline models on three prominent CTF benchmarks: InterCode-CTF, NYU CTF Bench, and Cybench. Our best model, Cyber-Zero-32B, establishes new state-of-the-art performance among open-weight models, matching the capabilities of proprietary systems like DeepSeek-V3-0324 and Claude-3.5-Sonnet while offering superior cost-effectiveness, and demonstrating that runtime-free trajectory synthesis can effectively democratize the development of state-of-the-art cybersecurity agents.
Authors: Mohammad Mohammadi, Ziyi Wu, Igor Gilitschenski
Abstract: Long-term temporal information is crucial for event-based perception tasks, as raw events only encode pixel brightness changes. Recent works show that when trained from scratch, recurrent models achieve better results than feedforward models in these tasks. However, when leveraging self-supervised pre-trained weights, feedforward models can outperform their recurrent counterparts. Current self-supervised learning (SSL) methods for event-based pre-training largely mimic RGB image-based approaches. They pre-train feedforward models on raw events within a short time interval, ignoring the temporal information of events. In this work, we introduce TESPEC, a self-supervised pre-training framework tailored for learning spatio-temporal information. TESPEC is well-suited for recurrent models, as it is the first framework to leverage long event sequences during pre-training. TESPEC employs the masked image modeling paradigm with a new reconstruction target. We design a novel method to accumulate events into pseudo grayscale videos containing high-level semantic information about the underlying scene, which is robust to sensor noise and reduces motion blur. Reconstructing this target thus requires the model to reason about long-term history of events. Extensive experiments demonstrate our state-of-the-art results in downstream tasks, including object detection, semantic segmentation, and monocular depth estimation. Project webpage: https://mhdmohammadi.github.io/TESPEC_webpage.
Authors: Dominic Simon, Rickard Ewetz
Abstract: Large Language Models (LLMs) require lightweight avenues of updating stored information that has fallen out of date. Knowledge Editing (KE) approaches have been successful in updating model knowledge for simple factual queries but struggle with handling tasks that require compositional reasoning such as multi-hop question answering (MQA). We observe that existing knowledge editors leverage decompositional techniques that result in illogical reasoning processes. In this paper, we propose a knowledge editor for MQA based on semantic analysis called CHECK. Our framework is based on insights from an analogy between compilers and reasoning using LLMs. Similar to how source code is first compiled before being executed, we propose to semantically analyze reasoning chains before executing the chains to answer questions. Reasoning chains with semantic errors are revised to ensure consistency through logic optimization and re-prompting the LLM model at a higher temperature. We evaluate the effectiveness of CHECK against five state-of-the-art frameworks on four datasets and achieve an average 22.8% improved MQA accuracy.
Authors: Kenrick Howin Chai, Stefan Hildebrand, Tobias Lachnit, Martin Benfer, Gisela Lanza, Sandra Klinge
Abstract: Rental-based business models and increasing sustainability requirements intensify the need for efficient strategies to manage large machine and vehicle fleet renewal and upgrades. Optimized fleet upgrade strategies maximize overall utility, cost, and sustainability. However, conventional fleet optimization does not account for upgrade options and is based on integer programming with exponential runtime scaling, which leads to substantial computational cost when dealing with large fleets and repeated decision-making processes. This contribution firstly suggests an extended integer programming approach that determines optimal renewal and upgrade decisions. The computational burden is addressed by a second, alternative machine learning-based method that transforms the task to a mixed discrete-continuous optimization problem. Both approaches are evaluated in a real-world automotive industry case study, which shows that the machine learning approach achieves near-optimal solutions with significant improvements in the scalability and overall computational performance, thus making it a practical alternative for large-scale fleet management.
Authors: Jiayuan Wang, Farhad Pourpanah, Q. M. Jonathan Wu, Ning Zhang
Abstract: Connected autonomous vehicles (CAVs) must simultaneously perform multiple tasks, such as object detection, semantic segmentation, depth estimation, trajectory prediction, motion prediction, and behaviour prediction, to ensure safe and reliable navigation in complex environments. Vehicle-to-everything (V2X) communication enables cooperative driving among CAVs, thereby mitigating the limitations of individual sensors, reducing occlusions, and improving perception over long distances. Traditionally, these tasks are addressed using distinct models, which leads to high deployment costs, increased computational overhead, and challenges in achieving real-time performance. Multi-task learning (MTL) has recently emerged as a promising solution that enables the joint learning of multiple tasks within a single unified model. This offers improved efficiency and resource utilization. To the best of our knowledge, this survey is the first comprehensive review focused on MTL in the context of CAVs. We begin with an overview of CAVs and MTL to provide foundational background. We then explore the application of MTL across key functional modules, including perception, prediction, planning, control, and multi-agent collaboration. Finally, we discuss the strengths and limitations of existing methods, identify key research gaps, and provide directions for future research aimed at advancing MTL methodologies for CAV systems.
Authors: Lirong Wu, Junjie Wang, Zhifeng Gao, Xiaohong Ji, Rong Zhu, Xinyu Li, Linfeng Zhang, Guolin Ke, Weinan E
Abstract: Organic reaction, the foundation of modern chemical industry, is crucial for new material development and drug discovery. However, deciphering reaction mechanisms and modeling multi-molecular relationships remain formidable challenges due to the complexity of molecular dynamics. While several state-of-the-art models like Uni-Mol2 have revolutionized single-molecular representation learning, their extension to multi-molecular systems, where chemical reactions inherently occur, has been underexplored. This paper introduces Uni-Mol3, a novel deep learning framework that employs a hierarchical pipeline for multi-molecular reaction modeling. At its core, Uni-Mol3 adopts a multi-scale molecular tokenizer (Mol-Tokenizer) that encodes 3D structures of molecules and other features into discrete tokens, creating a 3D-aware molecular language. The framework innovatively combines two pre-training stages: molecular pre-training to learn the molecular grammars and reaction pre-training to capture fundamental reaction principles, forming a progressive learning paradigm from single- to multi-molecular systems. With prompt-aware downstream fine-tuning, Uni-Mol3 demonstrates exceptional performance in diverse organic reaction tasks and supports multi-task prediction with strong generalizability. Experimental results across 10 datasets spanning 4 downstream tasks show that Uni-Mol3 outperforms existing methods, validating its effectiveness in modeling complex organic reactions. This work not only ushers in an alternative paradigm for multi-molecular computational modeling but also charts a course for intelligent organic reaction by bridging molecular representation with reaction mechanism understanding.
Authors: Bernarda Petek, David Nabergoj, Erik \v{S}trumbelj
Abstract: Visualizing uncertainty is integral to data analysis, yet its application is often hindered by the need for specialized methods for quantifying and representing uncertainty for different types of graphics. We introduce a general approach that simplifies this process. The core idea is to treat the statistical graphic as a function of the underlying distribution. Instead of first calculating uncertainty metrics and then plotting them, the method propagates uncertainty through to the visualization. By repeatedly sampling from the data distribution and generating a complete statistical graphic for each sample, a distribution over graphics is produced. These graphics are aggregated pixel-by-pixel to create a single, static image. This approach is versatile, requires no specific knowledge from the user beyond how to create the basic statistical graphic, and comes with theoretical coverage guarantees for standard cases such as confidence intervals and bands. We provide a reference implementation as a Python library to demonstrate the method's utility. Our approach not only reproduces conventional uncertainty visualizations for point estimates and regression lines but also seamlessly extends to non-standard cases, including pie charts, stacked bar charts, and tables. This approach makes uncertainty visualization more accessible to practitioners and can be a valuable tool for teaching uncertainty.
Authors: Daniel Andr\'es L\'opez, Vincent Weber, Severin Zentgraf, Barlo Hillen, Perikles Simon, Elmar Sch\"omer
Abstract: Infrared thermography is emerging as a powerful tool in sports medicine, allowing assessment of thermal radiation during exercise and analysis of anatomical regions of interest, such as the well-exposed calves. Building on our previous advanced automatic annotation method, we aimed to transfer the stereo- and multimodal-based labeling approach from treadmill running to ergometer cycling. Therefore, the training of the semantic segmentation network with automatic labels and fine-tuning on high-quality manually annotated images has been examined and compared in different data set combinations. The results indicate that fine-tuning with a small fraction of manual data is sufficient to improve the overall performance of the deep neural network. Finally, combining automatically generated labels with small manually annotated data sets accelerates the adaptation of deep neural networks to new use cases, such as the transition from treadmill to bicycle.
Authors: Beata E. Kowal, Krzysztof M. Graczyk, Artur M. Ankowski, Rwik Dharmapal Banerjee, Jose L. Bonilla, Hemant Prasad, Jan T. Sobczyk
Abstract: We present an updated deep neural network model for inclusive electron-carbon scattering. Using the bootstrap model [Phys.Rev.C 110 (2024) 2, 025501] as a prior, we incorporate recent experimental data, as well as older measurements in the deep inelastic scattering region, to derive a re-optimized posterior model. We examine the impact of these new inputs on model predictions and associated uncertainties. Finally, we evaluate the resulting cross-section predictions in the kinematic range relevant to the Hyper-Kamiokande and DUNE experiments.
Authors: Byron Dowling, Jozef Probcin, Adam Czajka
Abstract: Can we teach machines to assess the expertise of humans solving visual tasks automatically based on eye tracking features? This paper proposes AutoSIGHT, Automatic System for Immediate Grading of Human experTise, that classifies expert and non-expert performers, and builds upon an ensemble of features extracted from eye tracking data while the performers were solving a visual task. Results on the task of iris Presentation Attack Detection (PAD) used for this study show that with a small evaluation window of just 5 seconds, AutoSIGHT achieves an average average Area Under the ROC curve performance of 0.751 in subject-disjoint train-test regime, indicating that such detection is viable. Furthermore, when a larger evaluation window of up to 30 seconds is available, the Area Under the ROC curve (AUROC) increases to 0.8306, indicating the model is effectively leveraging more information at a cost of slightly delayed decisions. This work opens new areas of research on how to incorporate the automatic weighing of human and machine expertise into human-AI pairing setups, which need to react dynamically to nonstationary expertise distribution between the human and AI players (e.g. when the experts need to be replaced, or the task at hand changes rapidly). Along with this paper, we offer the eye tracking data used in this study collected from 6 experts and 53 non-experts solving iris PAD visual task.
Authors: Omar Elgohary, Nathan Jorgenson, Trenton Marple
Abstract: This replication study modifies ALMM, the Adaptive Linear Mapping Model constructed for the next song recommendation, to the news recommendation problem on the MIND dataset. The original version of ALMM computes latent representations for users, last-time items, and current items in a tensor factorization structure and learns a linear mapping from content features to latent item vectors. Our replication aims to improve recommendation performance in cold-start scenarios by restructuring this model to sequential news click behavior, viewing consecutively read articles as (last news, next news) tuples. Instead of the original audio features, we apply BERT and a TF-IDF (Term Frequency-Inverse Document Frequency) to news titles and abstracts to extract token contextualized representations and align them with triplet-based user reading patterns. We also propose a reproducibly thorough pre-processing pipeline combining news filtering and feature integrity validation. Our implementation of ALMM with TF-IDF shows relatively improved recommendation accuracy and robustness over Forbes and Oord baseline models in the cold-start scenario. We demonstrate that ALMM in a minimally modified state is not suitable for next news recommendation.
Authors: Theo Di Piazza, Carole Lazarus, Olivier Nempont, Loic Boussel
Abstract: With the increasing number of CT scan examinations, there is a need for automated methods such as organ segmentation, anomaly detection and report generation to assist radiologists in managing their increasing workload. Multi-label classification of 3D CT scans remains a critical yet challenging task due to the complex spatial relationships within volumetric data and the variety of observed anomalies. Existing approaches based on 3D convolutional networks have limited abilities to model long-range dependencies while Vision Transformers suffer from high computational costs and often require extensive pre-training on large-scale datasets from the same domain to achieve competitive performance. In this work, we propose an alternative by introducing a new graph-based approach that models CT scans as structured graphs, leveraging axial slice triplets nodes processed through spectral domain convolution to enhance multi-label anomaly classification performance. Our method exhibits strong cross-dataset generalization, and competitive performance while achieving robustness to z-axis translation. An ablation study evaluates the contribution of each proposed component.
Authors: Paul N. Patrone, Anthony J. Kearsley
Abstract: Motivated by canonical problems in medical diagnostics, we propose and study properties of an objective function that uniformly bounds uncertainties in quantities of interest extracted from classifiers and related data analysis tools. We begin by adopting a set-theoretic perspective to show how two main tasks in diagnostics -- classification and prevalence estimation -- can be recast in terms of a variation on the confusion (or error) matrix ${\boldsymbol {\rm P}}$ typically considered in supervised learning. We then combine arguments from conditional probability with the Gershgorin circle theorem to demonstrate that the largest Gershgorin radius $\boldsymbol \rho_m$ of the matrix $\mathbb I-\boldsymbol {\rm P}$ (where $\mathbb I$ is the identity) yields uniform error bounds for both classification and prevalence estimation. In a two-class setting, $\boldsymbol \rho_m$ is minimized via a measure-theoretic ``water-leveling'' argument that optimizes an appropriately defined partition $U$ generating the matrix ${\boldsymbol {\rm P}}$. We also consider an example that illustrates the difficulty of generalizing the binary solution to a multi-class setting and deduce relevant properties of the confusion matrix.
Authors: Santiago Diaz, Xinghui Hu, Josiane Uwumukiza, Giovanni Lavezzi, Victor Rodriguez-Fernandez, Richard Linares
Abstract: To enhance asteroid exploration and autonomous spacecraft navigation, we introduce DreamSat-2.0, a pipeline that benchmarks three state-of-the-art 3D reconstruction models-Hunyuan-3D, Trellis-3D, and Ouroboros-3D-on custom spacecraft and asteroid datasets. Our systematic analysis, using 2D perceptual (image quality) and 3D geometric (shape accuracy) metrics, reveals that model performance is domain-dependent. While models produce higher-quality images of complex spacecraft, they achieve better geometric reconstructions for the simpler forms of asteroids. New benchmarks are established, with Hunyuan-3D achieving top perceptual scores on spacecraft but its best geometric accuracy on asteroids, marking a significant advance over our prior work.
Authors: Yuki Shirai, Kei Ota, Devesh K. Jha, Diego Romeres
Abstract: Non-prehensile manipulation is challenging due to complex contact interactions between objects, the environment, and robots. Model-based approaches can efficiently generate complex trajectories of robots and objects under contact constraints. However, they tend to be sensitive to model inaccuracies and require access to privileged information (e.g., object mass, size, pose), making them less suitable for novel objects. In contrast, learning-based approaches are typically more robust to modeling errors but require large amounts of data. In this paper, we bridge these two approaches to propose a framework for learning closed-loop pivoting manipulation. By leveraging computationally efficient Contact-Implicit Trajectory Optimization (CITO), we design demonstration-guided deep Reinforcement Learning (RL), leading to sample-efficient learning. We also present a sim-to-real transfer approach using a privileged training strategy, enabling the robot to perform pivoting manipulation using only proprioception, vision, and force sensing without access to privileged information. Our method is evaluated on several pivoting tasks, demonstrating that it can successfully perform sim-to-real transfer.
Authors: Jun Qi, Chao-Han Yang, Pin-Yu Chen, Min-Hsiu Hsieh
Abstract: Variational Quantum Computing (VQC) faces fundamental barriers in scalability, primarily due to barren plateaus and quantum noise sensitivity. To address these challenges, we introduce TensoMeta-VQC, a novel tensor-train (TT)-guided meta-learning framework designed to improve the robustness and scalability of VQC significantly. Our framework fully delegates the generation of quantum circuit parameters to a classical TT network, effectively decoupling optimization from quantum hardware. This innovative parameterization mitigates gradient vanishing, enhances noise resilience through structured low-rank representations, and facilitates efficient gradient propagation. Based on Neural Tangent Kernel and statistical learning theory, our rigorous theoretical analyses establish strong guarantees on approximation capability, optimization stability, and generalization performance. Extensive empirical results across quantum dot classification, Max-Cut optimization, and molecular quantum simulation tasks demonstrate that TensoMeta-VQC consistently achieves superior performance and robust noise tolerance, establishing it as a principled pathway toward practical and scalable VQC on near-term quantum devices.
Authors: Saba Ahmadi, Rabiul Awal, Ankur Sikarwar, Amirhossein Kazemnejad, Ge Ya Luo, Juan A. Rodriguez, Sai Rajeswar, Siva Reddy, Christopher Pal, Benno Krojer, Aishwarya Agrawal
Abstract: We explore three strategies to enhance performance on a wide range of image editing tasks: supervised fine-tuning (SFT), reinforcement learning (RL), and Chain-of-Thought (CoT) reasoning. In order to study all these components in one consistent framework, we adopt an autoregressive multimodal model that processes textual and visual tokens in a unified manner. We find RL combined with a large multi-modal LLM verifier to be the most effective of these strategies. As a result, we release EARL: Editing with Autoregression and RL, a strong RL-based image editing model that performs competitively on a diverse range of edits compared to strong baselines, despite using much less training data. Thus, EARL pushes the frontier of autoregressive multimodal models on image editing. We release our code, training data, and trained models at https://github.com/mair-lab/EARL.
Authors: Leyao Wang, Xutao Mao, Xuhui Zhan, Yuying Zhao, Bo Ni, Ryan A. Rossi, Nesreen K. Ahmed, Tyler Derr
Abstract: Textual reviews enrich recommender systems with fine-grained preference signals and enhanced explainability. However, in real-world scenarios, users rarely leave reviews, resulting in severe sparsity that undermines the effectiveness of existing models. A natural solution is to impute or generate missing reviews to enrich the data. However, conventional imputation techniques -- such as matrix completion and LLM-based augmentation -- either lose contextualized semantics by embedding texts into vectors, or overlook structural dependencies among user-item interactions. To address these shortcomings, we propose TWISTER (ToWards Imputation on Sparsity with Textual Edge Graph Representation), a unified framework that imputes missing reviews by jointly modeling semantic and structural signals. Specifically, we represent user-item interactions as a Textual-Edge Graph (TEG), treating reviews as edge attributes. To capture relational context, we construct line-graph views and employ a large language model as a graph-aware aggregator. For each interaction lacking a textual review, our model aggregates the neighborhood's natural-language representations to generate a coherent and personalized review. Experiments on the Amazon and Goodreads datasets show that TWISTER consistently outperforms traditional numeric, graph-based, and LLM baselines, delivering higher-quality imputed reviews and, more importantly, enhanced recommendation performance. In summary, TWISTER generates reviews that are more helpful, authentic, and specific, while smoothing structural signals for improved recommendations.
Authors: Sateesh Kumar, Shivin Dass, Georgios Pavlakos, Roberto Mart\'in-Mart\'in
Abstract: In this work, we study the problem of data retrieval for few-shot imitation learning: selecting data from a large dataset to train a performant policy for a specific task, given only a few target demonstrations. Prior methods retrieve data using a single-feature distance heuristic, assuming that the best demonstrations are those that most closely resemble the target examples in visual, semantic, or motion space. However, this approach captures only a subset of the relevant information and can introduce detrimental demonstrations, e.g., retrieving data from unrelated tasks due to similar scene layouts, or selecting similar motions from tasks with divergent goals. We present COLLAGE, a method for COLLective data AGgrEgation in few-shot imitation learning that uses an adaptive late fusion mechanism to guide the selection of relevant demonstrations based on a task-specific combination of multiple cues. COLLAGE follows a simple, flexible, and efficient recipe: it assigns weights to subsets of the dataset that are pre-selected using a single feature (e.g., appearance, shape, or language similarity), based on how well a policy trained on each subset predicts actions in the target demonstrations. These weights are then used to perform importance sampling during policy training, sampling data more densely or sparsely according to estimated relevance. COLLAGE is general and feature-agnostic, allowing it to combine any number of subsets selected by any retrieval heuristic, and to identify which subsets provide the greatest benefit for the target task. In extensive experiments, COLLAGE outperforms state-of-the-art retrieval and multi-task learning approaches by 5.1% in simulation across 10 tasks, and by 16.6% in the real world across 6 tasks, where we perform retrieval from the large-scale DROID dataset. More information at https://robin-lab.cs.utexas.edu/COLLAGE .
Authors: Wei Zhou, Peng Sun, Xuanhe Zhou, Qianglei Zang, Ji Xu, Tieying Zhang, Guoliang Li, Fan Wu
Abstract: The operation and maintenance (O&M) of database systems is critical to ensuring system availability and performance, typically requiring expert experience (e.g., identifying metric-to-anomaly relations) for effective diagnosis and recovery. However, existing automatic database O&M methods, including commercial products, cannot effectively utilize expert experience. On the one hand, rule-based methods only support basic O&M tasks (e.g., metric-based anomaly detection), which are mostly numerical equations and cannot effectively incorporate literal O&M experience (e.g., troubleshooting guidance in manuals). On the other hand, LLM-based methods, which retrieve fragmented information (e.g., standard documents + RAG), often generate inaccurate or generic results. To address these limitations, we present DBAIOps, a novel hybrid database O&M system that combines reasoning LLMs with knowledge graphs to achieve DBA-style diagnosis. First, DBAIOps introduces a heterogeneous graph model for representing the diagnosis experience, and proposes a semi-automatic graph construction algorithm to build that graph from thousands of documents. Second, DBAIOps develops a collection of (800+) reusable anomaly models that identify both directly alerted metrics and implicitly correlated experience and metrics. Third, for each anomaly, DBAIOps proposes a two-stage graph evolution mechanism to explore relevant diagnosis paths and identify missing relations automatically. It then leverages a reasoning LLM (e.g., DeepSeek-R1) to infer root causes and generate clear diagnosis reports for both DBAs and common users. Our evaluation over four mainstream database systems (Oracle, MySQL, PostgreSQL, and DM8) demonstrates that DBAIOps outperforms state-of-the-art baselines, 34.85% and 47.22% higher in root cause and human evaluation accuracy, respectively.
Authors: Huyu Wu, Duo Su, Junjie Hou, Guang Li
Abstract: Dataset condensation always faces a constitutive trade-off: balancing performance and fidelity under extreme compression. Existing methods struggle with two bottlenecks: image-level selection methods (Coreset Selection, Dataset Quantization) suffer from inefficiency condensation, while pixel-level optimization (Dataset Distillation) introduces semantic distortion due to over-parameterization. With empirical observations, we find that a critical problem in dataset condensation is the oversight of color's dual role as an information carrier and a basic semantic representation unit. We argue that improving the colorfulness of condensed images is beneficial for representation learning. Motivated by this, we propose DC3: a Dataset Condensation framework with Color Compensation. After a calibrated selection strategy, DC3 utilizes the latent diffusion model to enhance the color diversity of an image rather than creating a brand-new one. Extensive experiments demonstrate the superior performance and generalization of DC3 that outperforms SOTA methods across multiple benchmarks. To the best of our knowledge, besides focusing on downstream tasks, DC3 is the first research to fine-tune pre-trained diffusion models with condensed datasets. The FID results prove that training networks with our high-quality datasets is feasible without model collapse or other degradation issues. Code and generated data will be released soon.
Authors: Chengshuai Zhao, Zhen Tan, Pingchuan Ma, Dawei Li, Bohan Jiang, Yancheng Wang, Yingzhen Yang, Huan Liu
Abstract: Chain-of-Thought (CoT) prompting has been shown to improve Large Language Model (LLM) performance on various tasks. With this approach, LLMs appear to produce human-like reasoning steps before providing answers (a.k.a., CoT reasoning), which often leads to the perception that they engage in deliberate inferential processes. However, some initial findings suggest that CoT reasoning may be more superficial than it appears, motivating us to explore further. In this paper, we study CoT reasoning via a data distribution lens and investigate if CoT reasoning reflects a structured inductive bias learned from in-distribution data, allowing the model to conditionally generate reasoning paths that approximate those seen during training. Thus, its effectiveness is fundamentally bounded by the degree of distribution discrepancy between the training data and the test queries. With this lens, we dissect CoT reasoning via three dimensions: task, length, and format. To investigate each dimension, we design DataAlchemy, an isolated and controlled environment to train LLMs from scratch and systematically probe them under various distribution conditions. Our results reveal that CoT reasoning is a brittle mirage that vanishes when it is pushed beyond training distributions. This work offers a deeper understanding of why and when CoT reasoning fails, emphasizing the ongoing challenge of achieving genuine and generalizable reasoning.
Authors: Yan Sun, Faming Liang
Abstract: Deep learning has revolutionized modern data science. However, how to accurately quantify the uncertainty of predictions from large-scale deep neural networks (DNNs) remains an unresolved issue. To address this issue, we introduce a novel post-processing approach. This approach feeds the output from the last hidden layer of a pre-trained large-scale DNN model into a stochastic neural network (StoNet), then trains the StoNet with a sparse penalty on a validation dataset and constructs prediction intervals for future observations. We establish a theoretical guarantee for the validity of this approach; in particular, the parameter estimation consistency for the sparse StoNet is essential for the success of this approach. Comprehensive experiments demonstrate that the proposed approach can construct honest confidence intervals with shorter interval lengths compared to conformal methods and achieves better calibration compared to other post-hoc calibration techniques. Additionally, we show that the StoNet formulation provides us with a platform to adapt sparse learning theory and methods from linear models to DNNs.
Authors: Anzhe Cheng, Chenzhong Yin, Mingxi Cheng, Shukai Duan, Shahin Nazarian, Paul Bogdan
Abstract: The remarkable success of Deep Neural Networks(DNN) is driven by gradient-based optimization, yet this process is often undermined by its tendency to produce disordered weight structures, which harms feature clarity and degrades learning dynamics. To address this fundamental representational flaw, we introduced the Eigen Neural Network (ENN), a novel architecture that reparameterizes each layer's weights in a layer-shared, learned orthonormal eigenbasis. This design enforces decorrelated, well-aligned weight dynamics axiomatically, rather than through regularization, leading to more structured and discriminative feature representations. When integrated with standard BP, ENN consistently outperforms state-of-the-art methods on large-scale image classification benchmarks, including ImageNet, and its superior representations generalize to set a new benchmark in cross-modal image-text retrieval. Furthermore, ENN's principled structure enables a highly efficient, backpropagation-free(BP-free) local learning variant, ENN-$\ell$. This variant not only resolves BP's procedural bottlenecks to achieve over 2$\times$ training speedup via parallelism, but also, remarkably, surpasses the accuracy of end-to-end backpropagation. ENN thus presents a new architectural paradigm that directly remedies the representational deficiencies of BP, leading to enhanced performance and enabling a more efficient, parallelizable training regime.
Authors: Zihan Fang, Zhiyong Xu, Lan Du, Shide Du, Zhiling Cai, Shiping Wang
Abstract: Existing multi-view learning models struggle in open-set scenarios due to their implicit assumption of class completeness. Moreover, static view-induced biases, which arise from spurious view-label associations formed during training, further degrade their ability to recognize unknown categories. In this paper, we propose a multi-view open-set learning framework via ambiguity uncertainty calibration and view-wise debiasing. To simulate ambiguous samples, we design O-Mix, a novel synthesis strategy to generate virtual samples with calibrated open-set ambiguity uncertainty. These samples are further processed by an auxiliary ambiguity perception network that captures atypical patterns for improved open-set adaptation. Furthermore, we incorporate an HSIC-based contrastive debiasing module that enforces independence between view-specific ambiguous and view-consistent representations, encouraging the model to learn generalizable features. Extensive experiments on diverse multi-view benchmarks demonstrate that the proposed framework consistently enhances unknown-class recognition while preserving strong closed-set performance.
Authors: Maximilian Pichler, Yannek K\"aber
Abstract: Modeling forest dynamics under novel climatic conditions requires a careful balance between process-based understanding and empirical flexibility. Dynamic Vegetation Models (DVM) represent ecological processes mechanistically, but their performance is prone to misspecified assumptions about functional forms. Inferring the structure of these processes and their functional forms correctly from data remains a major challenge because current approaches, such as plug-in estimators, have proven ineffective. We introduce Forest Informed Neural Networks (FINN), a hybrid modeling approach that combines a forest gap model with deep neural networks (DNN). FINN replaces processes with DNNs, which are then calibrated alongside the other mechanistic components in one unified step. In a case study on the Barro Colorado Island 50-ha plot we demonstrate that replacing the growth process with a DNN improves predictive performance and succession trajectories compared to a fully mechanistic version of FINN. Furthermore, we discovered that the DNN learned an ecologically plausible, improved functional form of growth, which we extracted from the DNN using explainable AI. In conclusion, our new hybrid modeling approach offers a versatile opportunity to infer forest dynamics from data and to improve forecasts of ecosystem trajectories under unprecedented environmental change.
Authors: Peiran Wang, Yang Liu, Yunfei Lu, Yifeng Cai, Hongbo Chen, Qingyou Yang, Jie Zhang, Jue Hong, Ye Wu
Abstract: Large Language Model (LLM) agents offer a powerful new paradigm for solving various problems by combining natural language reasoning with the execution of external tools. However, their dynamic and non-transparent behavior introduces critical security risks, particularly in the presence of prompt injection attacks. In this work, we propose a novel insight that treats the agent runtime traces as structured programs with analyzable semantics. Thus, we present AgentArmor, a program analysis framework that converts agent traces into graph intermediate representation-based structured program dependency representations (e.g., CFG, DFG, and PDG) and enforces security policies via a type system. AgentArmor consists of three key components: (1) a graph constructor that reconstructs the agent's working traces as graph-based intermediate representations with control flow and data flow described within; (2) a property registry that attaches security-relevant metadata of interacted tools & data, and (3) a type system that performs static inference and checking over the intermediate representation. By representing agent behavior as structured programs, AgentArmor enables program analysis over sensitive data flow, trust boundaries, and policy violations. We evaluate AgentArmor on the AgentDojo benchmark, the results show that AgentArmor can achieve 95.75% of TPR, with only 3.66% of FPR. Our results demonstrate AgentArmor's ability to detect prompt injection vulnerabilities and enforce fine-grained security constraints.
Authors: Raphael Schwinger, Paria Vali Zadeh, Lukas Rauch, Mats Kurz, Tom Hauschild, Sam Lapp, Sven Tomforde
Abstract: Automated bioacoustic analysis is essential for biodiversity monitoring and conservation, requiring advanced deep learning models that can adapt to diverse bioacoustic tasks. This article presents a comprehensive review of large-scale pretrained bioacoustic foundation models and systematically investigates their transferability across multiple bioacoustic classification tasks. We overview bioacoustic representation learning including major pretraining data sources and benchmarks. On this basis, we review bioacoustic foundation models by thoroughly analysing design decisions such as model architecture, pretraining scheme, and training paradigm. Additionally, we evaluate selected foundation models on classification tasks from the BEANS and BirdSet benchmarks, comparing the generalisability of learned representations under both linear and attentive probing strategies. Our comprehensive experimental analysis reveals that BirdMAE, trained on large-scale bird song data with a self-supervised objective, achieves the best performance on the BirdSet benchmark. On BEANS, BEATs$_{NLM}$, the extracted encoder of the NatureLM-audio large audio model, is slightly better. Both transformer-based models require attentive probing to extract the full performance of their representations. ConvNext$_{BS}$ and Perch models trained with supervision on large-scale bird song data remain competitive for passive acoustic monitoring classification tasks of BirdSet in linear probing settings. Training a new linear classifier has clear advantages over evaluating these models without further training. While on BEANS, the baseline model BEATs trained with self-supervision on AudioSet outperforms bird-specific models when evaluated with attentive probing. These findings provide valuable guidance for practitioners selecting appropriate models to adapt them to new bioacoustic classification tasks via probing.
Authors: Yanmei Hu, Siyuan Yin, Yihang Wu, Xue Yue, Yue Liu
Abstract: Identifying influential nodes in complex networks is of great importance, and has many applications in practice. For example, finding influential nodes in e-commerce network can provide merchants with customers with strong purchase intent; identifying influential nodes in computer information system can help locating the components that cause the system break down and identifying influential nodes in these networks can accelerate the flow of information in networks. Thus, a lot of efforts have been made on the problem of indentifying influential nodes. However, previous efforts either consider only one aspect of the network structure, or using global centralities with high time consuming as node features to identify influential nodes, and the existing methods do not consider the relationships between different centralities. To solve these problems, we propose a Graph Convolutional Network Framework based on Feature Network, abbreviated as FNGCN (graph convolutional network is abbreviated as GCN in the following text). Further, to exclude noises and reduce redundency, FNGCN utilizes feature network to represent the complicated relationships among the local centralities, based on which the most suitable local centralities are determined. By taking a shallow GCN and a deep GCN into the FNGCN framework, two FNGCNs are developed. With ground truth obtained from the widely used Susceptible Infected Recovered (SIR) model, the two FNGCNs are compared with the state-of-art methods on several real-world networks. Experimental results show that the two FNGCNs can identify the influential nodes more accurately than the compared methods, indicating that the proposed framework is effective in identifying influential nodes in complex networks.
Authors: Haoquan Lu, Hanzhe Liang, Jie Zhang, Chenxi Hu, Jinbao Wang, Can Gao
Abstract: 3D Anomaly Detection (AD) has shown great potential in detecting anomalies or defects of high-precision industrial products. However, existing methods are typically trained in a class-specific manner and also lack the capability of learning from emerging classes. In this study, we proposed a continual learning framework named Continual 3D Anomaly Detection (C3D-AD), which can not only learn generalized representations for multi-class point clouds but also handle new classes emerging over time.Specifically, in the feature extraction module, to extract generalized local features from diverse product types of different tasks efficiently, Kernel Attention with random feature Layer (KAL) is introduced, which normalizes the feature space. Then, to reconstruct data correctly and continually, an efficient Kernel Attention with learnable Advisor (KAA) mechanism is proposed, which learns the information from new categories while discarding redundant old information within both the encoder and decoder. Finally, to keep the representation consistency over tasks, a Reconstruction with Parameter Perturbation (RPP) module is proposed by designing a representation rehearsal loss function, which ensures that the model remembers previous category information and returns category-adaptive representation.Extensive experiments on three public datasets demonstrate the effectiveness of the proposed method, achieving an average performance of 66.4%, 83.1%, and 63.4% AUROC on Real3D-AD, Anomaly-ShapeNet, and MulSen-AD, respectively.
Authors: Marc Braun, Jose M. Pe\~na, Adel Daoud
Abstract: To reach human level intelligence, learning algorithms need to incorporate causal reasoning. But identifying causality, and particularly counterfactual reasoning, remains an elusive task. In this paper, we make progress on this task by utilizing instrumental variables (IVs). IVs are a classic tool for mitigating bias from unobserved confounders when estimating causal effects. While IV methods have been extended to non-separable structural models at the population level, existing approaches to counterfactual prediction typically assume additive noise in the outcome. In this paper, we show that under standard IV assumptions, along with the assumptions that latent noises in treatment and outcome are strictly monotonic and jointly Gaussian, the treatment-outcome relationship becomes uniquely identifiable from observed data. This enables counterfactual inference even in nonseparable models. We implement our approach by training a normalizing flow to maximize the likelihood of the observed data, demonstrating accurate recovery of the underlying outcome function. We call our method Flow IV.
Authors: Markus Pettersson, Connor T. Jerzak, Adel Daoud
Abstract: Machine learning models trained on Earth observation data, such as satellite imagery, have demonstrated significant promise in predicting household-level wealth indices, enabling the creation of high-resolution wealth maps that can be leveraged across multiple causal trials. However, because standard training objectives prioritize overall predictive accuracy, these predictions inherently suffer from shrinkage toward the mean, leading to attenuated estimates of causal treatment effects and limiting their utility in policy. Existing debiasing methods, such as Prediction-Powered Inference, can handle this attenuation bias but require additional fresh ground-truth data at the downstream stage of causal inference, which restricts their applicability in data-scarce environments. Here, we introduce and evaluate two correction methods -- linear calibration correction and Tweedie's correction -- that substantially reduce prediction bias without relying on newly collected labeled data. Linear calibration corrects bias through a straightforward linear transformation derived from held-out calibration data, whereas Tweedie's correction leverages empirical Bayes principles to directly address shrinkage-induced biases by exploiting score functions derived from the model's learning patterns. Through analytical exercises and experiments using Demographic and Health Survey data, we demonstrate that the proposed methods meet or outperform existing approaches that either require (a) adjustments to training pipelines or (b) additional labeled data. These approaches may represent a promising avenue for improving the reliability of causal inference when direct outcome measures are limited or unavailable, enabling a "one map, many trials" paradigm where a single upstream data creation team produces predictions usable by many downstream teams across diverse ML pipelines.
Authors: Xi Huang, Hongyi Zhou, Ge Li, Yucheng Tang, Weiran Liao, Bj\"orn Hein, Tamim Asfour, Rudolf Lioutikov
Abstract: We propose MoRe-ERL, a framework that combines Episodic Reinforcement Learning (ERL) and residual learning, which refines preplanned reference trajectories into safe, feasible, and efficient task-specific trajectories. This framework is general enough to incorporate into arbitrary ERL methods and motion generators seamlessly. MoRe-ERL identifies trajectory segments requiring modification while preserving critical task-related maneuvers. Then it generates smooth residual adjustments using B-Spline-based movement primitives to ensure adaptability to dynamic task contexts and smoothness in trajectory refinement. Experimental results demonstrate that residual learning significantly outperforms training from scratch using ERL methods, achieving superior sample efficiency and task performance. Hardware evaluations further validate the framework, showing that policies trained in simulation can be directly deployed in real-world systems, exhibiting a minimal sim-to-real gap.
Authors: Shitian Yang, Deyu Li, Xiaoke Jiang, Lei Zhang
Abstract: RGB-based 3D tasks, e.g., 3D detection, depth estimation, 3D keypoint estimation, still suffer from scarce, expensive annotations and a thin augmentation toolbox, since most image transforms, including resize and rotation, disrupt geometric consistency. In this paper, we introduce 3DRot, a plug-and-play augmentation that rotates and mirrors images about the camera's optical center while synchronously updating RGB images, camera intrinsics, object poses, and 3D annotations to preserve projective geometry-achieving geometry-consistent rotations and reflections without relying on any scene depth. We validate 3DRot with a classical 3D task, monocular 3D detection. On SUN RGB-D dataset, 3DRot raises $IoU_{3D}$ from 43.21 to 44.51, cuts rotation error (ROT) from 22.91$^\circ$ to 20.93$^\circ$, and boosts $mAP_{0.5}$ from 35.70 to 38.11. As a comparison, Cube R-CNN adds 3 other datasets together with SUN RGB-D for monocular 3D estimation, with a similar mechanism and test dataset, increases $IoU_{3D}$ from 36.2 to 37.8, boosts $mAP_{0.5}$ from 34.7 to 35.4. Because it operates purely through camera-space transforms, 3DRot is readily transferable to other 3D tasks.
Authors: Sadegh Ebrahimkhani, John Lataire
Abstract: The choice of parameterization in Nonlinear (NL) system models greatly affects the quality of the estimated model. Overly complex models can be impractical and hard to interpret, necessitating data-driven methods for simpler and more accurate representations. In this paper, we propose a data-driven approach to simplify a class of continuous-time NL system models using linear approximations around varying operating points. Specifically, for sparse additive NL models, our method identifies the number of NL subterms and their corresponding input spaces. Under small-signal operation, we approximate the unknown NL system as a trajectory-scheduled Linear Parameter-Varying (LPV) system, with LPV coefficients representing the gradient of the NL function and indicating input sensitivity. Using this sensitivity measure, we determine the NL system's structure through LPV model reduction by identifying non-zero LPV coefficients and selecting scheduling parameters. We introduce two sparse estimators within a vector-valued Reproducing Kernel Hilbert Space (RKHS) framework to estimate the LPV coefficients while preserving their structural relationships. The structure of the sparse additive NL model is then determined by detecting non-zero elements in the gradient vector (LPV coefficients) and the Hessian matrix (Jacobian of the LPV coefficients). We propose two computationally tractable RKHS-based estimators for this purpose. The sparsified Hessian matrix reveals the NL model's structure, with numerical simulations confirming the approach's effectiveness.
Authors: Faruk Alpay, Taylan Alpay, Bugra Kilictas
Abstract: We study the inverse problem of reconstructing high-dimensional trust embeddings from the one-dimensional Siamese trust scores that many distributed-security frameworks expose. Starting from two independent agents that publish time-stamped similarity scores for the same set of devices, we formalise the estimation task, derive an explicit direct-sum estimator that concatenates paired score series with four moment features, and prove that the resulting reconstruction map admits a unique fixed point under a contraction argument rooted in Banach theory. A suite of synthetic benchmarks (20 devices x 10 time steps) confirms that, even in the presence of Gaussian noise, the recovered embeddings preserve inter-device geometry as measured by Euclidean and cosine metrics; we complement these experiments with non-asymptotic error bounds that link reconstruction accuracy to score-sequence length. Beyond methodology, the paper demonstrates a practical privacy risk: publishing granular trust scores can leak latent behavioural information about both devices and evaluation models. We therefore discuss counter-measures -- score quantisation, calibrated noise, obfuscated embedding spaces -- and situate them within wider debates on transparency versus confidentiality in networked AI systems. All datasets, reproduction scripts and extended proofs accompany the submission so that results can be verified without proprietary code.
Authors: Alain Riou, Bernardo Torres, Ben Hayes, Stefan Lattner, Ga\"etan Hadjeres, Ga\"el Richard, Geoffroy Peeters
Abstract: In this paper, we introduce PESTO, a self-supervised learning approach for single-pitch estimation using a Siamese architecture. Our model processes individual frames of a Variable-$Q$ Transform (VQT) and predicts pitch distributions. The neural network is designed to be equivariant to translations, notably thanks to a Toeplitz fully-connected layer. In addition, we construct pitch-shifted pairs by translating and cropping the VQT frames and train our model with a novel class-based transposition-equivariant objective, eliminating the need for annotated data. Thanks to this architecture and training objective, our model achieves remarkable performances while being very lightweight ($130$k parameters). Evaluations on music and speech datasets (MIR-1K, MDB-stem-synth, and PTDB) demonstrate that PESTO not only outperforms self-supervised baselines but also competes with supervised methods, exhibiting superior cross-dataset generalization. Finally, we enhance PESTO's practical utility by developing a streamable VQT implementation using cached convolutions. Combined with our model's low latency (less than 10 ms) and minimal parameter count, this makes PESTO particularly suitable for real-time applications.
Authors: Rushin H. Gindra, Giovanni Palla, Mathias Nguyen, Sophia J. Wagner, Manuel Tran, Fabian J Theis, Dieter Saur, Lorin Crawford, Tingying Peng
Abstract: Spatial transcriptomics enables simultaneous measurement of gene expression and tissue morphology, offering unprecedented insights into cellular organization and disease mechanisms. However, the field lacks comprehensive benchmarks for evaluating multimodal learning methods that leverage both histology images and gene expression data. Here, we present HESCAPE, a large-scale benchmark for cross-modal contrastive pretraining in spatial transcriptomics, built on a curated pan-organ dataset spanning 6 different gene panels and 54 donors. We systematically evaluated state-of-the-art image and gene expression encoders across multiple pretraining strategies and assessed their effectiveness on two downstream tasks: gene mutation classification and gene expression prediction. Our benchmark demonstrates that gene expression encoders are the primary determinant of strong representational alignment, and that gene models pretrained on spatial transcriptomics data outperform both those trained without spatial data and simple baseline approaches. However, downstream task evaluation reveals a striking contradiction: while contrastive pretraining consistently improves gene mutation classification performance, it degrades direct gene expression prediction compared to baseline encoders trained without cross-modal objectives. We identify batch effects as a key factor that interferes with effective cross-modal alignment. Our findings highlight the critical need for batch-robust multimodal learning approaches in spatial transcriptomics. To accelerate progress in this direction, we release HESCAPE, providing standardized datasets, evaluation protocols, and benchmarking tools for the community
Authors: Bernardo Torres, Alain Riou, Ga\"el Richard, Geoffroy Peeters
Abstract: In this paper, we propose an Optimal Transport objective for learning one-dimensional translation-equivariant systems and demonstrate its applicability to single pitch estimation. Our method provides a theoretically grounded, more numerically stable, and simpler alternative for training state-of-the-art self-supervised pitch estimators.
Authors: Danial Ebrat, Tina Aminian, Sepideh Ahmadian, Luis Rueda
Abstract: Recommender systems are essential for guiding users through the vast and diverse landscape of digital content by delivering personalized and relevant suggestions. However, improving both personalization and interpretability remains a challenge, particularly in scenarios involving limited user feedback or heterogeneous item attributes. In this article, we propose a novel hybrid recommendation framework that combines Graph Attention Networks (GATs) with Large Language Models (LLMs) to address these limitations. LLMs are first used to enrich user and item representations by generating semantically meaningful profiles based on metadata such as titles, genres, and overviews. These enriched embeddings serve as initial node features in a user and movie bipartite graph, which is processed using a GAT based collaborative filtering model. To enhance ranking accuracy, we introduce a hybrid loss function that combines Bayesian Personalized Ranking (BPR), cosine similarity, and robust negative sampling. Post-processing involves reranking the GAT-generated recommendations using the LLM, which also generates natural-language justifications to improve transparency. We evaluated our model on benchmark datasets, including MovieLens 100k and 1M, where it consistently outperforms strong baselines. Ablation studies confirm that LLM-based embeddings and the cosine similarity term significantly contribute to performance gains. This work demonstrates the potential of integrating LLMs to improve both the accuracy and interpretability of recommender systems.
Authors: Nianyi Wang, Yu Chen, Shuai Zheng
Abstract: Learning-based fluid simulation networks have been proven as viable alternatives to traditional numerical solvers for the Navier-Stokes equations. Existing neural methods follow Smoothed Particle Hydrodynamics (SPH) frameworks, which inherently rely only on local inter-particle interactions. However, we emphasize that global context integration is also essential for learning-based methods to stabilize complex fluid simulations. We propose the first Fluid Attention Block (FAB) with a local-global hierarchy, where continuous convolutions extract local features while self-attention captures global dependencies. This fusion suppresses the error accumulation and models long-range physical phenomena. Furthermore, we pioneer the first Transformer architecture specifically designed for continuous fluid simulation, seamlessly integrated within a dual-pipeline architecture. Our method establishes a new paradigm for neural fluid simulation by unifying convolution-based local features with attention-based global context modeling. FluidFormer demonstrates state-of-the-art performance, with stronger stability in complex fluid scenarios.
Authors: Yujia Tong, Tian Zhang, Jingling Yuan, Yuze Wang, Chuang Hu
Abstract: Vision Transformers (ViTs) have revolutionized computer vision tasks with their exceptional performance. However, the introduction of privacy regulations such as GDPR and CCPA has brought new challenges to them. These laws grant users the right to withdraw their data, necessitating not only the deletion of data but also the complete removal of its influence from trained models. Machine unlearning emerges as a critical solution, with exact unlearning being computationally prohibitive and approximate methods offering a more practical approach. This work addresses the particularly challenging scenario of random data forgetting in ViTs, where the model must forget specific samples while retaining others, even within the same class. We first reveal the core characteristics of ViTs through selective masking experiments: when high-attention areas are masked, the model retains its recognition capability but significantly weakens its memorization ability. Based on the above insights, we propose LetheViT, a contrastive unlearning method tailored for ViTs. LetheViT uses masked image inputs to generate positive logits and original image inputs to generate negative logits, guiding the model to forget specific details while retaining the general cl category outlines. Experimental results demonstrate that LetheViT achieves state-of-the-art performance, effectively balancing privacy compliance with model efficacy.
Authors: Xuanran Zhai, Ce Hao
Abstract: Flow-matching-based policies have recently emerged as a promising approach for learning-based robot manipulation, offering significant acceleration in action sampling compared to diffusion-based policies. However, conventional flow-matching methods struggle with multi-modality, often collapsing to averaged or ambiguous behaviors in complex manipulation tasks. To address this, we propose the Variational Flow-Matching Policy (VFP), which introduces a variational latent prior for mode-aware action generation and effectively captures both task-level and trajectory-level multi-modality. VFP further incorporates Kantorovich Optimal Transport (K-OT) for distribution-level alignment and utilizes a Mixture-of-Experts (MoE) decoder for mode specialization and efficient inference. We comprehensively evaluate VFP on 41 tasks across four benchmark environments, demonstrating its effectiveness and sampling efficiency in both task and path multi-modality settings. Results show that VFP achieves a $49\%$ relative improvement in task success rate over standard flow-based baselines, while maintaining fast inference and compact model size. More details are available on our project page: https://sites.google.com/view/varfp/
Authors: Shai Kimhi, Avi Mendlson, Moshe Kimhi
Abstract: Adversarial patch attacks threaten the reliability of modern vision models. We present PatchMap, the first spatially exhaustive benchmark of patch placement, built by evaluating over 1.5e8 forward passes on ImageNet validation images. PatchMap reveals systematic hot-spots where small patches (as little as 2% of the image) induce confident misclassifications and large drops in model confidence. To demonstrate its utility, we propose a simple segmentation guided placement heuristic that leverages off the shelf masks to identify vulnerable regions without any gradient queries. Across five architectures-including adversarially trained ResNet50, our method boosts attack success rates by 8 to 13 percentage points compared to random or fixed placements. We publicly release PatchMap and the code implementation. The full PatchMap bench (6.5B predictions, multiple backbones) will be released soon to further accelerate research on location-aware defenses and adaptive attacks.
Authors: Raviraj Joshi, Rakesh Paul, Kanishk Singla, Anusha Kamath, Michael Evans, Katherine Luna, Shaona Ghosh, Utkarsh Vaidya, Eileen Long, Sanjay Singh Chauhan, Niranjan Wartikar
Abstract: The increasing use of Large Language Models (LLMs) in agentic applications highlights the need for robust safety guard models. While content safety in English is well-studied, non-English languages lack similar advancements due to the high cost of collecting culturally aligned labeled datasets. We present CultureGuard, a novel solution for curating culturally aligned, high-quality safety datasets across multiple languages. Our approach introduces a four-stage synthetic data generation and filtering pipeline: cultural data segregation, cultural data adaptation, machine translation, and quality filtering. This pipeline enables the conversion and expansion of the Nemotron-Content-Safety-Dataset-V2 English safety dataset into eight distinct languages: Arabic, German, Spanish, French, Hindi, Japanese, Thai, and Chinese. The resulting dataset, Nemotron-Content-Safety-Dataset-Multilingual-v1, comprises 386,661 samples in 9 languages and facilitates the training of Llama-3.1-Nemotron-Safety-Guard-Multilingual-8B-v1 via LoRA-based fine-tuning. The final model achieves state-of-the-art performance on several multilingual content safety benchmarks. We also benchmark the latest open LLMs on multilingual safety and observe that these LLMs are more prone to give unsafe responses when prompted in non-English languages. This work represents a significant step toward closing the safety gap in multilingual LLMs by enabling the development of culturally aware safety guard models.
Authors: Xin He, Junxi Shen, Zhenheng Tang, Xiaowen Chu, Bo Li, Ivor W. Tsang, Yew-Soon Ong
Abstract: Model merging via Mixture-of-Experts (MoE) has emerged as a scalable solution for consolidating multiple task-specific models into a unified sparse architecture, where each expert is derived from a model fine-tuned on a distinct task. While effective for multi-task integration, this paradigm introduces a critical yet underexplored challenge: how to attribute and protect the intellectual property (IP) of individual experts after merging. We propose RouteMark, a framework for IP protection in merged MoE models through the design of expert routing fingerprints. Our key insight is that task-specific experts exhibit stable and distinctive routing behaviors under probing inputs. To capture these patterns, we construct expert-level fingerprints using two complementary statistics: the Routing Score Fingerprint (RSF), quantifying the intensity of expert activation, and the Routing Preference Fingerprint (RPF), characterizing the input distribution that preferentially activates each expert. These fingerprints are reproducible, task-discriminative, and lightweight to construct. For attribution and tampering detection, we introduce a similarity-based matching algorithm that compares expert fingerprints between a suspect and a reference (victim) model. Extensive experiments across diverse tasks and CLIP-based MoE architectures show that RouteMark consistently yields high similarity for reused experts and clear separation from unrelated ones. Moreover, it remains robust against both structural tampering (expert replacement, addition, deletion) and parametric tampering (fine-tuning, pruning, permutation), outperforming weight- and activation-based baseliness. Our work lays the foundation for RouteMark as a practical and broadly applicable framework for IP verification in MoE-based model merging.
Authors: Jing Lan, Hexiao Ding, Hongzhao Chen, Yufeng Jiang, Ng Nga Chun, Gerald W. Y. Cheng, Zongxi Li, Jing Cai, Liang-ting Lin, Jung Sun Yoo
Abstract: Accurate prediction of protein-ligand interactions is essential for computer-aided drug discovery. However, existing methods often fail to capture solvent-dependent conformational changes and lack the ability to jointly learn multiple related tasks. To address these limitations, we introduce a pre-training method that incorporates ligand conformational ensembles generated under diverse solvent conditions as augmented input. This design enables the model to learn both structural flexibility and environmental context in a unified manner. The training process integrates molecular reconstruction to capture local geometry, interatomic distance prediction to model spatial relationships, and contrastive learning to build solvent-invariant molecular representations. Together, these components lead to significant improvements, including a 3.7% gain in binding affinity prediction, an 82% success rate on the PoseBusters Astex docking benchmarks, and an area under the curve of 97.1% in virtual screening. The framework supports solvent-aware, multi-task modeling and produces consistent results across benchmarks. A case study further demonstrates sub-angstrom docking accuracy with a root-mean-square deviation of 0.157 angstroms, offering atomic-level insight into binding mechanisms and advancing structure-based drug design.
Authors: Hwanwoo Kim, Simon Mak, Ann-Kathrin Schuetz, Alan Poon
Abstract: With advances in scientific computing, computer experiments are increasingly used for optimizing complex systems. However, for modern applications, e.g., the optimization of nuclear physics detectors, each experiment run can require hundreds of CPU hours, making the optimization of its black-box simulator over a high-dimensional space a challenging task. Given limited runs at inputs $\mathbf{x}_1, \cdots, \mathbf{x}_n$, the best solution from these evaluated inputs can be far from optimal, particularly as dimensionality increases. Existing black-box methods, however, largely employ this ''pick-the-winner'' (PW) solution, which leads to mediocre optimization performance. To address this, we propose a new Black-box Optimization via Marginal Means (BOMM) approach. The key idea is a new estimator of a global optimizer $\mathbf{x}^*$ that leverages the so-called marginal mean functions, which can be efficiently inferred with limited runs in high dimensions. Unlike PW, this estimator can select solutions beyond evaluated inputs for improved optimization performance. Assuming the objective function follows a generalized additive model with unknown link function and under mild conditions, we prove that the BOMM estimator not only is consistent for optimization, but also has an optimization rate that tempers the ''curse-of-dimensionality'' faced by existing methods, thus enabling better performance as dimensionality increases. We present a practical framework for implementing BOMM using the transformed additive Gaussian process surrogate model. Finally, we demonstrate the effectiveness of BOMM in numerical experiments and an application on neutrino detector optimization in nuclear physics.
Authors: Avishkar Behera, Riya Ann Easow, Venkatesh Parvathala, K. Sri Rama Murty
Abstract: This paper introduces a novel application of Test-Time Training (TTT) for Speech Enhancement, addressing the challenges posed by unpredictable noise conditions and domain shifts. This method combines a main speech enhancement task with a self-supervised auxiliary task in a Y-shaped architecture. The model dynamically adapts to new domains during inference time by optimizing the proposed self-supervised tasks like noise-augmented signal reconstruction or masked spectrogram prediction, bypassing the need for labeled data. We further introduce various TTT strategies offering a trade-off between adaptation and efficiency. Evaluations across synthetic and real-world datasets show consistent improvements across speech quality metrics, outperforming the baseline model. This work highlights the effectiveness of TTT in speech enhancement, providing insights for future research in adaptive and robust speech processing.
Authors: Junyi Mo, Jiayu Li, Duo Zhang, Elynn Chen
Abstract: Missing data in financial panels presents a critical obstacle, undermining asset-pricing models and reducing the effectiveness of investment strategies. Such panels are often inherently multi-dimensional, spanning firms, time, and financial variables, which adds complexity to the imputation task. Conventional imputation methods often fail by flattening the data's multidimensional structure, struggling with heterogeneous missingness patterns, or overfitting in the face of extreme data sparsity. To address these limitations, we introduce an Adaptive, Cluster-based Temporal smoothing tensor completion framework (ACT-Tensor) tailored for severely and heterogeneously missing multi-dimensional financial data panels. ACT-Tensor incorporates two key innovations: a cluster-based completion module that captures cross-sectional heterogeneity by learning group-specific latent structures; and a temporal smoothing module that proactively removes short-lived noise while preserving slow-moving fundamental trends. Extensive experiments show that ACT-Tensor consistently outperforms state-of-the-art benchmarks in terms of imputation accuracy across a range of missing data regimes, including extreme sparsity scenarios. To assess its practical financial utility, we evaluate the imputed data with an asset-pricing pipeline tailored for tensor-structured financial data. Results show that ACT-Tensor not only reduces pricing errors but also significantly improves risk-adjusted returns of the constructed portfolio. These findings confirm that our method delivers highly accurate and informative imputations, offering substantial value for financial decision-making.
Authors: Nicolas Langren\'e, Xavier Warin, Pierre Gruet
Abstract: To speed up Gaussian process inference, a number of fast kernel matrix-vector multiplication (MVM) approximation algorithms have been proposed over the years. In this paper, we establish an exact fast kernel MVM algorithm based on exact kernel decomposition into weighted empirical cumulative distribution functions, compatible with a class of kernels which includes multivariate Mat\'ern kernels with half-integer smoothness parameter. This algorithm uses a divide-and-conquer approach, during which sorting outputs are stored in a data structure. We also propose a new algorithm to take into account some linear fixed effects predictor function. Our numerical experiments confirm that our algorithm is very effective for low-dimensional Gaussian process inference problems with hundreds of thousands of data points. An implementation of our algorithm is available at https://gitlab.com/warin/fastgaussiankernelregression.git.
URLs: https://gitlab.com/warin/fastgaussiankernelregression.git.
Authors: Yang Sun, Wenbin Lu, Yi-Hui Zhou
Abstract: Recent developments in causal inference have greatly shifted the interest from estimating the average treatment effect to the individual treatment effect. In this article, we improve the predictive accuracy of representation learning and adversarial networks in estimating individual treatment effects by introducing a structure keeper which maintains the correlation between the baseline covariates and their corresponding representations in the high dimensional space. We train a discriminator at the end of representation layers to trade off representation balance and information loss. We show that the proposed discriminator minimizes an upper bound of the treatment estimation error. We can address the tradeoff between distribution balance and information loss by considering the correlations between the learned representation space and the original covariate feature space. We conduct extensive experiments with simulated and real-world observational data to show that our proposed Structure Maintained Representation Learning (SMRL) algorithm outperforms state-of-the-art methods. We also demonstrate the algorithms on real electronic health record data from the MIMIC-III database.
Authors: Haozhe Zhou, Riku Arakawa, Yuvraj Agarwal, Mayank Goel
Abstract: IMUs are regularly used to sense human motion, recognize activities, and estimate full-body pose. Users are typically required to place sensors in predefined locations that are often dictated by common wearable form factors and the machine learning model's training process. Consequently, despite the increasing number of everyday devices equipped with IMUs, the limited adaptability has seriously constrained the user experience to only using a few well-explored device placements (e.g., wrist and ears). In this paper, we rethink IMU-based motion sensing by acknowledging that signals can be captured from any point on the human body. We introduce IMU over Continuous Coordinates (IMUCoCo), a novel framework that maps signals from a variable number of IMUs placed on the body surface into a unified feature space based on their spatial coordinates. These features can be plugged into downstream models for pose estimation and activity recognition. Our evaluations demonstrate that IMUCoCo supports accurate pose estimation in a wide range of typical and atypical sensor placements. Overall, IMUCoCo supports significantly more flexible use of IMUs for motion sensing than the state-of-the-art, allowing users to place their sensors-laden devices according to their needs and preferences. The framework also supports the ability to change device locations depending on the context and suggests placement depending on the use case.
Authors: Akshay Paruchuri, Sinan Hersek, Lavisha Aggarwal, Qiao Yang, Xin Liu, Achin Kulshrestha, Andrea Colaco, Henry Fuchs, Ishan Chatterjee
Abstract: All-day smart glasses are likely to emerge as platforms capable of continuous contextual sensing, uniquely positioning them for unprecedented assistance in our daily lives. Integrating the multi-modal AI agents required for human memory enhancement while performing continuous sensing, however, presents a major energy efficiency challenge for all-day usage. Achieving this balance requires intelligent, context-aware sensor management. Our approach, EgoTrigger, leverages audio cues from the microphone to selectively activate power-intensive cameras, enabling efficient sensing while preserving substantial utility for human memory enhancement. EgoTrigger uses a lightweight audio model (YAMNet) and a custom classification head to trigger image capture from hand-object interaction (HOI) audio cues, such as the sound of a drawer opening or a medication bottle being opened. In addition to evaluating on the QA-Ego4D dataset, we introduce and evaluate on the Human Memory Enhancement Question-Answer (HME-QA) dataset. Our dataset contains 340 human-annotated first-person QA pairs from full-length Ego4D videos that were curated to ensure that they contained audio, focusing on HOI moments critical for contextual understanding and memory. Our results show EgoTrigger can use 54% fewer frames on average, significantly saving energy in both power-hungry sensing components (e.g., cameras) and downstream operations (e.g., wireless transmission), while achieving comparable performance on datasets for an episodic memory task. We believe this context-aware triggering strategy represents a promising direction for enabling energy-efficient, functional smart glasses capable of all-day use -- supporting applications like helping users recall where they placed their keys or information about their routine activities (e.g., taking medications).
Authors: Yaroslav Prytula, Illia Tsiporenko, Ali Zeynalli, Dmytro Fishman
Abstract: Instance segmentation is critical in biomedical imaging to accurately distinguish individual objects like cells, which often overlap and vary in size. Recent query-based methods, where object queries guide segmentation, have shown strong performance. While U-Net has been a go-to architecture in medical image segmentation, its potential in query-based approaches remains largely unexplored. In this work, we present IAUNet, a novel query-based U-Net architecture. The core design features a full U-Net architecture, enhanced by a novel lightweight convolutional Pixel decoder, making the model more efficient and reducing the number of parameters. Additionally, we propose a Transformer decoder that refines object-specific features across multiple scales. Finally, we introduce the 2025 Revvity Full Cell Segmentation Dataset, a unique resource with detailed annotations of overlapping cell cytoplasm in brightfield images, setting a new benchmark for biomedical instance segmentation. Experiments on multiple public datasets and our own show that IAUNet outperforms most state-of-the-art fully convolutional, transformer-based, and query-based models and cell segmentation-specific models, setting a strong baseline for cell instance segmentation tasks. Code is available at https://github.com/SlavkoPrytula/IAUNet
Authors: Andrea Dosi, Semanto Mondal, Rajib Chandra Ghosh, Massimo Brescia, Giuseppe Longo
Abstract: This work presents the results of a methodological transfer from remote sensing to healthcare, adapting AMBER -- a transformer-based model originally designed for multiband images, such as hyperspectral data -- to the task of 3D medical datacube segmentation. In this study, we use the AMBER architecture with Adaptive Fourier Neural Operators (AFNO) in place of the multi-head self-attention mechanism. While existing models rely on various forms of attention to capture global context, AMBER-AFNO achieves this through frequency-domain mixing, enabling a drastic reduction in model complexity. This design reduces the number of trainable parameters by over 80% compared to UNETR++, while maintaining a FLOPs count comparable to other state-of-the-art architectures. Model performance is evaluated on two benchmark 3D medical datasets -- ACDC and Synapse -- using standard metrics such as Dice Similarity Coefficient (DSC) and Hausdorff Distance (HD), demonstrating that AMBER-AFNO achieves competitive or superior accuracy with significant gains in training efficiency, inference speed, and memory usage.
Authors: Jiayi Wang, Jacqueline Jil Vallon, Neil Panjwani, Xi Ling, Sushmita Vij, Sandy Srinivas, John Leppert, Mark K. Buyyounouski, Mohsen Bayati
Abstract: Electronic health records (EHRs) contain rich unstructured clinical notes that could enhance predictive modeling, yet extracting meaningful features from these notes remains challenging. Current approaches range from labor-intensive manual clinician feature generation (CFG) to fully automated representational feature generation (RFG) that lack interpretability and clinical relevance. Here we introduce SNOW (Scalable Note-to-Outcome Workflow), a modular multi-agent system powered by large language models (LLMs) that autonomously generates structured clinical features from unstructured notes without human intervention. We evaluated SNOW against manual CFG, clinician-guided LLM approaches, and RFG methods for predicting 5-year prostate cancer recurrence in 147 patients from Stanford Healthcare. While manual CFG achieved the highest performance (AUC-ROC: 0.771), SNOW matched this performance (0.761) without requiring any clinical expertise, significantly outperforming both baseline features alone (0.691) and all RFG approaches. The clinician-guided LLM method also performed well (0.732) but still required expert input. SNOW's specialized agents handle feature discovery, extraction, validation, post-processing, and aggregation, creating interpretable features that capture complex clinical information typically accessible only through manual review. Our findings demonstrate that autonomous LLM systems can replicate expert-level feature engineering at scale, potentially transforming how clinical ML models leverage unstructured EHR data while maintaining the interpretability essential for clinical deployment.
Authors: Md Badsha Biswas, \"Ozlem Uzuner
Abstract: Social media, such as Twitter, provides opportunities for caregivers of dementia patients to share their experiences and seek support for a variety of reasons. Availability of this information online also paves the way for the development of internet-based interventions in their support. However, for this purpose, tweets written by caregivers of dementia patients must first be identified. This paper demonstrates our system for the SMM4H 2025 shared task 3, which focuses on detecting tweets posted by individuals who have a family member with dementia. The task is outlined as a binary classification problem, differentiating between tweets that mention dementia in the context of a family member and those that do not. Our solution to this problem explores large language models (LLMs) with various prompting methods. Our results show that a simple zero-shot prompt on a fine-tuned model yielded the best results. Our final system achieved a macro F1-score of 0.95 on the validation set and the test set. Our full code is available on GitHub.
Authors: Chungang Lin, Weiyao Zhang, Tianyu Zuo, Chao Zha, Yilong Jiang, Ruiqi Meng, Haitong Luo, Xuying Meng, Yujun Zhang
Abstract: Encrypted traffic classification is vital for modern network management and security. To reduce reliance on handcrafted features and labeled data, recent methods focus on learning generic representations through pre-training on large-scale unlabeled data. However, current pre-trained models face two limitations originating from the adopted Transformer architecture: (1) Limited model efficiency due to the self-attention mechanism with quadratic complexity; (2) Unstable traffic scalability to longer byte sequences, as the explicit positional encodings fail to generalize to input lengths not seen during pre-training. In this paper, we investigate whether convolutions, with linear complexity and implicit positional encoding, are competitive with Transformers in encrypted traffic classification with pre-training. We first conduct a systematic comparison, and observe that convolutions achieve higher efficiency and scalability, with lower classification performance. To address this trade-off, we propose NetConv, a novel pre-trained convolution model for encrypted traffic classification. NetConv employs stacked traffic convolution layers, which enhance the ability to capture localized byte-sequence patterns through window-wise byte scoring and sequence-wise byte gating. We design a continuous byte masking pre-training task to help NetConv learn protocol-specific patterns. Experimental results on four tasks demonstrate that NetConv improves average classification performance by 6.88% and model throughput by 7.41X over existing pre-trained models.
Authors: Ali Alkinoon, Trung Cuong Dang, Ahod Alghuried, Abdulaziz Alghamdi, Soohyeon Choi, Manar Mohaisen, An Wang, Saeed Salem, David Mohaisen
Abstract: The proper use of Android app permissions is crucial to the success and security of these apps. Users must agree to permission requests when installing or running their apps. Despite official Android platform documentation on proper permission usage, there are still many cases of permission abuse. This study provides a comprehensive analysis of the Android permission landscape, highlighting trends and patterns in permission requests across various applications from the Google Play Store. By distinguishing between benign and malicious applications, we uncover developers' evolving strategies, with malicious apps increasingly requesting fewer permissions to evade detection, while benign apps request more to enhance functionality. In addition to examining permission trends across years and app features such as advertisements, in-app purchases, content ratings, and app sizes, we leverage association rule mining using the FP-Growth algorithm. This allows us to uncover frequent permission combinations across the entire dataset, specific years, and 16 app genres. The analysis reveals significant differences in permission usage patterns, providing a deeper understanding of co-occurring permissions and their implications for user privacy and app functionality. By categorizing permissions into high-level semantic groups and examining their application across distinct app categories, this study offers a structured approach to analyzing the dynamics within the Android ecosystem. The findings emphasize the importance of continuous monitoring, user education, and regulatory oversight to address permission misuse effectively.
Authors: Zhihao Luo, Wentao Yan abd Jingyu Gong, Min Wang, Zhizhong Zhang, Xuhong Wang, Yuan Xie, Xin Tan
Abstract: Recent advances in Graphical User Interface (GUI) and embodied navigation have driven significant progress, yet these domains have largely evolved in isolation, with disparate datasets and training paradigms. In this paper, we observe that both tasks can be formulated as Markov Decision Processes (MDP), suggesting a foundational principle for their unification. Hence, we present NaviMaster, the first unified agent capable of seamlessly integrating GUI navigation and embodied navigation within a single framework. Specifically, NaviMaster (i) proposes a visual-target trajectory collection pipeline that generates trajectories for both GUI and embodied tasks in one formulation. (ii) employs a unified reinforcement learning framework on the mix data for better generalization. (iii) designs a novel distance-aware reward to ensure efficient learning from the trajectories. Through extensive experiments on out-of-domain benchmarks, NaviMaster is shown to outperform state-of-the-art agents in GUI navigation, spatial affordance prediction, and embodied navigation. Ablation studies further confirm the efficacy of our unified training strategy, data mixing strategy, and reward design.
Authors: Zhentao Xu, Fengyi Li, Albert Chen, Xiaofeng Wang
Abstract: In large-scale industrial LLM systems, prompt templates often expand to thousands of tokens as teams iteratively incorporate sections such as task instructions, few-shot examples, and heuristic rules to enhance robustness and coverage. This expansion leads to bloated prompts that are difficult to maintain and incur significant inference latency and serving costs. To address this, we introduce Prompt Compression via Attribution Estimation (ProCut), a flexible, LLM-agnostic, training-free framework that compresses prompts through attribution analysis. ProCut segments prompt templates into semantically meaningful units, quantifies their impact on task performance, and prunes low-utility components. Through extensive experiments on five public benchmark datasets and real-world industrial prompts, we show that ProCut achieves substantial prompt size reductions (78% fewer tokens in production) while maintaining or even slightly improving task performance (up to 62% better than alternative methods). We further introduce an LLM-driven attribution estimator that reduces compression latency by over 50%, and demonstrate that ProCut integrates seamlessly with existing prompt-optimization frameworks to produce concise, high-performing prompts.
Authors: Gustaf Ahdritz, Anat Kleiman
Abstract: The internet is rife with unattributed, deliberately misleading, or otherwise untrustworthy content. Though large language models (LLMs) are often tasked with autonomous web browsing, the extent to which they have learned the simple heuristics human researchers use to navigate this noisy environment is not currently known. In this paper, we introduce the Synthetic Media Literacy Test (SMeL Test), a minimal benchmark that tests the ability of language models to actively filter out untrustworthy information in context. We benchmark a variety of commonly used instruction-tuned LLMs, including reasoning models, and find that no model consistently trusts more reliable sources; while reasoning in particular is associated with higher scores, even the best API model we test hallucinates up to 70% of the time. Remarkably, larger and more capable models do not necessarily outperform their smaller counterparts. We hope our work sheds more light on this important form of hallucination and guides the development of new methods to combat it.
Authors: Jingze Shi, Yifan Wu, Bingheng Wu, Yiran Peng, Liangdong Wang, Guang Liu, Yuyu Luo
Abstract: In large language models, the demand for modeling long contexts is constantly increasing, but the quadratic complexity of the standard self-attention mechanism often becomes a bottleneck. Although existing sparse attention mechanisms have improved efficiency, they may still encounter issues such as static patterns or information loss. We introduce a trainable dynamic mask sparse attention mechanism, Dynamic Mask Attention, which effectively utilizes content-aware and position-aware sparsity. DMA achieves this through two key innovations: First, it dynamically generates content-aware sparse masks from value representations, enabling the model to identify and focus on critical information adaptively. Second, it implements position-aware sparse attention computation that effectively skips unnecessary calculation regions. This dual-sparsity design allows the model to significantly reduce the computational complexity of important information while retaining complete information, achieving an excellent balance between information fidelity and computational efficiency. We have verified the performance of DMA through comprehensive experiments. Comparative studies show that DMA outperforms multi-head attention, sliding window attention, multi-head latent attention, and native sparse attention in terms of perplexity under Chinchilla Scaling Law settings. Moreover, in challenging multi-query associative recall tasks, DMA also demonstrates superior performance and efficiency compared to these methods. Crucially, in the evaluation of a 1.7B parameter model, DMA significantly outperforms multi-head attention in both standard benchmark performance and the challenging needle-in-a-haystack task. These experimental results highlight its capability to balance model efficiency and long-context modeling ability effectively.
Authors: Dor Elimelech, Wasim Huleihel
Abstract: Detection of planted subgraphs in Erd\"os-R\'enyi random graphs has been extensively studied, leading to a rich body of results characterizing both statistical and computational thresholds. However, most prior work assumes a purely random generative model, making the resulting algorithms potentially fragile in the face of real-world perturbations. In this work, we initiate the study of semi-random models for the planted subgraph detection problem, wherein an adversary is allowed to remove edges outside the planted subgraph before the graph is revealed to the statistician. Crucially, the statistician remains unaware of which edges have been removed, introducing fundamental challenges to the inference task. We establish fundamental statistical limits for detection under this semi-random model, revealing a sharp dichotomy. Specifically, for planted subgraphs with strongly sub-logarithmic maximum density detection becomes information-theoretically impossible in the presence of an adversary, despite being possible in the classical random model. In stark contrast, for subgraphs with super-logarithmic density, the statistical limits remain essentially unchanged; we prove that the optimal (albeit computationally intractable) likelihood ratio test remains robust. Beyond these statistical boundaries, we design a new computationally efficient and robust detection algorithm, and provide rigorous statistical guarantees for its performance. Our results establish the first robust framework for planted subgraph detection and open new directions in the study of semi-random models, computational-statistical trade-offs, and robustness in graph inference problems.
Authors: Yuxuan Song, Zheng Zhang, Cheng Luo, Pengyang Gao, Fan Xia, Hao Luo, Zheng Li, Yuehang Yang, Hongli Yu, Xingwei Qu, Yuwei Fu, Jing Su, Ge Zhang, Wenhao Huang, Mingxuan Wang, Lin Yan, Xiaoying Jia, Jingjing Liu, Wei-Ying Ma, Ya-Qin Zhang, Yonghui Wu, Hao Zhou
Abstract: We present Seed Diffusion Preview, a large-scale language model based on discrete-state diffusion, offering remarkably fast inference speed. Thanks to non-sequential, parallel generation, discrete diffusion models provide a notable speedup to mitigate the inherent latency of token-by-token decoding, as demonstrated recently (e.g., Mercury Coder, Gemini Diffusion). Seed Diffusion Preview achieves an inference speed of 2,146 token/s over H20 GPUs while maintaining competitive performance across a sweep of standard code evaluation benchmarks, significantly faster than contemporary Mercury and Gemini Diffusion, establishing new state of the art on the speed-quality Pareto frontier for code models.
Authors: George Close, Kris Hong, Thomas Hain, Stefan Goetze
Abstract: There has been significant research effort developing neural-network-based predictors of SQ in recent years. While a primary objective has been to develop non-intrusive, i.e.~reference-free, metrics to assess the performance of SE systems, recent work has also investigated the direct inference of neural SQ predictors within the loss function of downstream speech tasks. To aid in the training of SQ predictors, several large datasets of audio with corresponding human labels of quality have been created. Recent work in this area has shown that speech representations derived from large unsupervised or semi-supervised foundational speech models are useful input feature representations for neural SQ prediction. In this work, a novel and robust SQ predictor is proposed based on feature representations extracted from an ASR model, found to be a powerful input feature for the SQ prediction task. The proposed system achieves higher correlation with human MOS ratings than recent approaches on all NISQA test sets and shows significantly better domain adaption compared to the commonly used DNSMOS metric.
Authors: Dongchi Huang, Zhirui Fang, Tianle Zhang, Yihang Li, Lin Zhao, Chunhe Xia
Abstract: Vision-Language-Action (VLA) models demonstrate significant potential for developing generalized policies in real-world robotic control. This progress inspires researchers to explore fine-tuning these models with Reinforcement Learning (RL). However, fine-tuning VLA models with RL still faces challenges related to sample efficiency, compatibility with action chunking, and training stability. To address these challenges, we explore the fine-tuning of VLA models through offline reinforcement learning incorporating action chunking. In this work, we propose Chunked RL, a novel reinforcement learning framework specifically designed for VLA models. Within this framework, we extend temporal difference (TD) learning to incorporate action chunking, a prominent characteristic of VLA models. Building upon this framework, we propose CO-RFT, an algorithm aimed at fine-tuning VLA models using a limited set of demonstrations (30 to 60 samples). Specifically, we first conduct imitation learning (IL) with full parameter fine-tuning to initialize both the backbone and the policy. Subsequently, we implement offline RL with action chunking to optimize the pretrained policy. Our empirical results in real-world environments demonstrate that CO-RFT outperforms previous supervised methods, achieving a 57% improvement in success rate and a 22.3% reduction in cycle time. Moreover, our method exhibits robust positional generalization capabilities, attaining a success rate of 44.3% in previously unseen positions.
Authors: Yang Li, Zhi Chen
Abstract: Generative modeling of high-frequency limit order book (LOB) dynamics is a critical yet unsolved challenge in quantitative finance, essential for robust market simulation and strategy backtesting. Existing approaches are often constrained by simplifying stochastic assumptions or, in the case of modern deep learning models like Transformers, rely on tokenization schemes that affect the high-precision, numerical nature of financial data through discretization and binning. To address these limitations, we introduce ByteGen, a novel generative model that operates directly on the raw byte streams of LOB events. Our approach treats the problem as an autoregressive next-byte prediction task, for which we design a compact and efficient 32-byte packed binary format to represent market messages without information loss. The core novelty of our work is the complete elimination of feature engineering and tokenization, enabling the model to learn market dynamics from its most fundamental representation. We achieve this by adapting the H-Net architecture, a hybrid Mamba-Transformer model that uses a dynamic chunking mechanism to discover the inherent structure of market messages without predefined rules. Our primary contributions are: 1) the first end-to-end, byte-level framework for LOB modeling; 2) an efficient packed data representation; and 3) a comprehensive evaluation on high-frequency data. Trained on over 34 million events from CME Bitcoin futures, ByteGen successfully reproduces key stylized facts of financial markets, generating realistic price distributions, heavy-tailed returns, and bursty event timing. Our findings demonstrate that learning directly from byte space is a promising and highly flexible paradigm for modeling complex financial systems, achieving competitive performance on standard market quality metrics without the biases of tokenization.
Authors: Arjun Kumar, Noppanat Wadlom, Jaeheon Kwak, Si-Hyuck Kang, Insik Shin
Abstract: Arrhythmia is a common cardiac condition that can precipitate severe complications without timely intervention. While continuous monitoring is essential for timely diagnosis, conventional approaches such as electrocardiogram and wearable devices are constrained by their reliance on specialized medical expertise and patient discomfort from their contact nature. Existing contactless monitoring, primarily designed for healthy subjects, face significant challenges when analyzing reflected signals from arrhythmia patients due to disrupted spatial stability and temporal consistency. In this paper, we introduce mCardiacDx, a radar-driven contactless system that accurately analyzes reflected signals and reconstructs heart pulse waveforms for arrhythmia monitoring and diagnosis. The key contributions of our work include a novel precise target localization (PTL) technique that locates reflected signals despite spatial disruptions, and an encoder-decoder model that transforms these signals into HPWs, addressing temporal inconsistencies. Our evaluation on a large dataset of healthy subjects and arrhythmia patients shows that both mCardiacDx and PTL outperform state-of-the-art approach in arrhythmia monitoring and diagnosis, also demonstrating improved performance in healthy subjects.
Authors: Samuele Grossi, Marco Letizia, Riccardo Torre
Abstract: The rise of generative models for scientific research calls for the development of new methods to evaluate their fidelity. A natural framework for addressing this problem is two-sample hypothesis testing, namely the task of determining whether two data sets are drawn from the same distribution. In large-scale and high-dimensional regimes, machine learning offers a set of tools to push beyond the limitations of standard statistical techniques. In this work, we put this claim to the test by comparing a recent proposal from the high-energy physics literature, the New Physics Learning Machine, to perform a classification-based two-sample test against a number of alternative approaches, following the framework presented in Grossi et al. (2025). We highlight the efficiency tradeoffs of the method and the computational costs that come from adopting learning-based approaches. Finally, we discuss the advantages of the different methods for different use cases.
Authors: Paul Zaha, Lars B\"ocking, Simeon Allmendinger, Leopold M\"uller, Niklas K\"uhl
Abstract: Medical image segmentation is crucial for disease diagnosis and treatment planning, yet developing robust segmentation models often requires substantial computational resources and large datasets. Existing research shows that pre-trained and finetuned foundation models can boost segmentation performance. However, questions remain about how particular image preprocessing steps may influence segmentation performance across different medical imaging modalities. In particular, edges-abrupt transitions in pixel intensity-are widely acknowledged as vital cues for object boundaries but have not been systematically examined in the pre-training of foundation models. We address this gap by investigating to which extend pre-training with data processed using computationally efficient edge kernels, such as kirsch, can improve cross-modality segmentation capabilities of a foundation model. Two versions of a foundation model are first trained on either raw or edge-enhanced data across multiple medical imaging modalities, then finetuned on selected raw subsets tailored to specific medical modalities. After systematic investigation using the medical domains Dermoscopy, Fundus, Mammography, Microscopy, OCT, US, and XRay, we discover both increased and reduced segmentation performance across modalities using edge-focused pre-training, indicating the need for a selective application of this approach. To guide such selective applications, we propose a meta-learning strategy. It uses standard deviation and image entropy of the raw image to choose between a model pre-trained on edge-enhanced or on raw data for optimal performance. Our experiments show that integrating this meta-learning layer yields an overall segmentation performance improvement across diverse medical imaging tasks by 16.42% compared to models pre-trained on edge-enhanced data only and 19.30% compared to models pre-trained on raw data only.
Authors: Wentao Zhang, Yilei Zhao, Chuqiao Zong, Xinrun Wang, Bo An
Abstract: Financial AI holds great promise for transforming modern finance, with the potential to support a wide range of tasks such as market forecasting, portfolio management, quantitative trading, and automated analysis. However, existing platforms remain limited in task coverage, lack robust multimodal data integration, and offer insufficient support for the training and deployment of large language models (LLMs). In response to these limitations, we present FinWorld, an all-in-one open-source platform that provides end-to-end support for the entire financial AI workflow, from data acquisition to experimentation and deployment. FinWorld distinguishes itself through native integration of heterogeneous financial data, unified support for diverse AI paradigms, and advanced agent automation, enabling seamless development and deployment. Leveraging data from 2 representative markets, 4 stock pools, and over 800 million financial data points, we conduct comprehensive experiments on 4 key financial AI tasks. These experiments systematically evaluate deep learning and reinforcement learning algorithms, with particular emphasis on RL-based finetuning for LLMs and LLM Agents. The empirical results demonstrate that FinWorld significantly enhances reproducibility, supports transparent benchmarking, and streamlines deployment, thereby providing a strong foundation for future research and real-world applications. Code is available at Github~\footnote{https://github.com/DVampire/FinWorld}.
Authors: Muhammad Aqeel, Shakiba Sharifi, Marco Cristani, Francesco Setti
Abstract: So-called unsupervised anomaly detection is better described as semi-supervised, as it assumes all training data are nominal. This assumption simplifies training but requires manual data curation, introducing bias and limiting adaptability. We propose Confident Meta-learning (CoMet), a novel training strategy that enables deep anomaly detection models to learn from uncurated datasets where nominal and anomalous samples coexist, eliminating the need for explicit filtering. Our approach integrates Soft Confident Learning, which assigns lower weights to low-confidence samples, and Meta-Learning, which stabilizes training by regularizing updates based on training validation loss covariance. This prevents overfitting and enhances robustness to noisy data. CoMet is model-agnostic and can be applied to any anomaly detection method trainable via gradient descent. Experiments on MVTec-AD, VIADUCT, and KSDD2 with two state-of-the-art models demonstrate the effectiveness of our approach, consistently improving over the baseline methods, remaining insensitive to anomalies in the training set, and setting a new state-of-the-art across all datasets.
Authors: Dmitrii Seletkov, Sophie Starck, Ayhan Can Erdur, Yundi Zhang, Daniel Rueckert, Rickmer Braren
Abstract: Reliable preclinical disease risk assessment is essential to move public healthcare from reactive treatment to proactive identification and prevention. However, image-based risk prediction algorithms often consider one condition at a time and depend on hand-crafted features obtained through segmentation tools. We propose a whole-body self-supervised representation learning method for the preclinical disease risk assessment under a competing risk modeling. This approach outperforms whole-body radiomics in multiple diseases, including cardiovascular disease (CVD), type 2 diabetes (T2D), chronic obstructive pulmonary disease (COPD), and chronic kidney disease (CKD). Simulating a preclinical screening scenario and subsequently combining with cardiac MRI, it sharpens further the prediction for CVD subgroups: ischemic heart disease (IHD), hypertensive diseases (HD), and stroke. The results indicate the translational potential of whole-body representations as a standalone screening modality and as part of a multi-modal framework within clinical workflows for early personalized risk stratification. The code is available at https://github.com/yayapa/WBRLforCR/
Authors: Yuzhuang Xu, Xu Han, Yuanchi Zhang, Yixuan Wang, Yijun Liu, Shiyu Ji, Qingfu Zhu, Wanxiang Che
Abstract: Large Language Models (LLMs) with Mixture-of-Experts (MoE) architectures are distinguished by their strong performance scaling with increasing parameters across a wide range of tasks, yet they also suffer from substantial computational and storage overheads. Notably, the performance gains of MoE models do not scale proportionally with the growth in expert parameters. While prior works attempt to reduce parameters via expert-level pruning, merging, or decomposition, they still suffer from challenges in both performance and computational efficiency. In this paper, we address these challenges by introducing micro-expert as a finer-grained compression unit that spans across matrices. We first establish a more fundamental perspective, viewing MoE layers as mixtures of micro-experts, and present CAMERA, a lightweight and training-free framework for identifying micro-expert redundancy. Our analysis uncovers significant variance in micro-expert contributions during decoding. Based on this insight, we further propose CAMERA-P, a structured micro-expert pruning framework, and CAMERA-Q, a mixed-precision quantization idea designed for micro-experts. Extensive experiments on nine downstream tasks show that CAMERA-P consistently outperforms strong baselines under pruning ratios ranging from 20% to 60%. Furthermore, CAMERA-Q achieves superior results under aggressive 2-bit quantization, surpassing existing matrix- and channel-level ideas. Notably, our method enables complete micro-expert analysis of Qwen2-57B-A14B in less than 5 minutes on a single NVIDIA A100-40GB GPU.
Authors: Jeanne I. M. Parmentier (Utrecht University, University of Twente, Inertia Technology B.V), Rhana M. Aarts (Utrecht University), Elin Hernlund (Swedish University of Agricultural Sciences), Marie Rhodin (Swedish University of Agricultural Sciences), Berend Jan van der Zwaag (University of Twente, Inertia Technology B.V)
Abstract: Monitoring respiration parameters such as respiratory rate could be beneficial to understand the impact of training on equine health and performance and ultimately improve equine welfare. In this work, we compare deep learning-based methods to an adapted signal processing method to automatically detect cyclic respiratory events and extract the dynamic respiratory rate from microphone recordings during high intensity exercise in Standardbred trotters. Our deep learning models are able to detect exhalation sounds (median F1 score of 0.94) in noisy microphone signals and show promising results on unlabelled signals at lower exercising intensity, where the exhalation sounds are less recognisable. Temporal convolutional networks were better at detecting exhalation events and estimating dynamic respiratory rates (median F1: 0.94, Mean Absolute Error (MAE) $\pm$ Confidence Intervals (CI): 1.44$\pm$1.04 bpm, Limits Of Agreements (LOA): 0.63$\pm$7.06 bpm) than long short-term memory networks (median F1: 0.90, MAE$\pm$CI: 3.11$\pm$1.58 bpm) and signal processing methods (MAE$\pm$CI: 2.36$\pm$1.11 bpm). This work is the first to automatically detect equine respiratory sounds and automatically compute dynamic respiratory rates in exercising horses. In the future, our models will be validated on lower exercising intensity sounds and different microphone placements will be evaluated in order to find the best combination for regular monitoring.
Authors: Cuno Sankey-Olsen, Rasmus Hvass Olesen, Tobias Oliver Eberhard, Andreas Triantafyllopoulos, Bj\"orn Schuller, Ilhan Aslan
Abstract: Chronic Obstructive Pulmonary Disease (COPD) is a serious and debilitating disease affecting millions around the world. Its early detection using non-invasive means could enable preventive interventions that improve quality of life and patient outcomes, with speech recently shown to be a valuable biomarker. Yet, its validity across different linguistic groups remains to be seen. To that end, audio data were collected from 96 Danish participants conducting three speech tasks (reading, coughing, sustained vowels). Half of the participants were diagnosed with different levels of COPD and the other half formed a healthy control group. Subsequently, we investigated different baseline models using openSMILE features and learnt x-vector embeddings. We obtained a best accuracy of 67% using openSMILE features and logistic regression. Our findings support the potential of speech-based analysis as a non-invasive, remote, and scalable screening tool as part of future COPD healthcare solutions.
Authors: Shuo Lu, Yanyin Chen, Wei Feng, Jiahao Fan, Fengheng Li, Zheng Zhang, Jingjing Lv, Junjie Shen, Ching Law, Jian Liang
Abstract: Layout generation plays a crucial role in enhancing both user experience and design efficiency. However, current approaches suffer from task-specific generation capabilities and perceptually misaligned evaluation metrics, leading to limited applicability and ineffective measurement. In this paper, we propose \textit{Uni-Layout}, a novel framework that achieves unified generation, human-mimicking evaluation and alignment between the two. For universal generation, we incorporate various layout tasks into a single taxonomy and develop a unified generator that handles background or element contents constrained tasks via natural language prompts. To introduce human feedback for the effective evaluation of layouts, we build \textit{Layout-HF100k}, the first large-scale human feedback dataset with 100,000 expertly annotated layouts. Based on \textit{Layout-HF100k}, we introduce a human-mimicking evaluator that integrates visual and geometric information, employing a Chain-of-Thought mechanism to conduct qualitative assessments alongside a confidence estimation module to yield quantitative measurements. For better alignment between the generator and the evaluator, we integrate them into a cohesive system by adopting Dynamic-Margin Preference Optimization (DMPO), which dynamically adjusts margins based on preference strength to better align with human judgments. Extensive experiments show that \textit{Uni-Layout} significantly outperforms both task-specific and general-purpose methods. Our code is publicly available at https://github.com/JD-GenX/Uni-Layout.
Authors: Xiao Wang, Hao Si, Fan Zhang, Xiaoya Zhou, Dengdi Sun, Wanli Lyu, Qingquan Yang, Jin Tang
Abstract: Multivariate time series analysis has long been one of the key research topics in the field of artificial intelligence. However, analyzing complex time series data remains a challenging and unresolved problem due to its high dimensionality, dynamic nature, and complex interactions among variables. Inspired by the strong structural modeling capability of hypergraphs, this paper proposes a novel hypergraph-based time series transformer backbone network, termed HGTS-Former, to address the multivariate coupling in time series data. Specifically, given the multivariate time series signal, we first normalize and embed each patch into tokens. Then, we adopt the multi-head self-attention to enhance the temporal representation of each patch. The hierarchical hypergraphs are constructed to aggregate the temporal patterns within each channel and fine-grained relations between different variables. After that, we convert the hyperedge into node features through the EdgeToNode module and adopt the feed-forward network to further enhance the output features. Extensive experiments conducted on two multivariate time series tasks and eight datasets fully validated the effectiveness of our proposed HGTS-Former. The source code will be released on https://github.com/Event-AHU/Time_Series_Analysis.
Authors: Nazmun N Khan, Taylor Sweet, Chase A Harvey, Calder Knapp, Dean J. Krusienski, David E Thompson
Abstract: The reliability of affective state estimation using EEG data is in question, given the variability in reported performance and the lack of standardized evaluation protocols. To investigate this, we reviewed 101 studies, focusing on the widely used DEAP dataset for emotion recognition. Our analysis revealed widespread methodological issues that include data leakage from improper segmentation, biased feature selection, flawed hyperparameter optimization, neglect of class imbalance, and insufficient methodological reporting. Notably, we found that nearly 87% of the reviewed papers contained one or more of these errors. Moreover, through experimental analysis, we observed that such methodological flaws can inflate the classification accuracy by up to 46%. These findings reveal fundamental gaps in standardized evaluation practices and highlight critical deficiencies in the peer review process for machine learning applications in neuroscience, emphasizing the urgent need for stricter methodological standards and evaluation protocols.
Authors: Akshay Dodwadmath, Setareh Maghsudi
Abstract: Stackelberg games and their resulting equilibria have received increasing attention in the multi-agent reinforcement learning literature. Each stage of a traditional Stackelberg game involves a leader(s) acting first, followed by the followers. In situations where the roles of leader(s) and followers can be interchanged, the designated role can have considerable advantages, for example, in first-mover advantage settings. Then the question arises: Who should be the leader and when? A bias in the leader selection process can lead to unfair outcomes. This problem is aggravated if the agents are self-interested and care only about their goals and rewards. We formally define this leader selection problem and show its relation to fairness in agents' returns. Furthermore, we propose a multi-agent reinforcement learning framework that maximizes fairness by integrating mediators. Mediators have previously been used in the simultaneous action setting with varying levels of control, such as directly performing agents' actions or just recommending them. Our framework integrates mediators in the Stackelberg setting with minimal control (leader selection). We show that the presence of mediators leads to self-interested agents taking fair actions, resulting in higher overall fairness in agents' returns.
Authors: Yu-Qin Chen, Shi-Xin Zhang
Abstract: The reliability of artificial intelligence hinges on the integrity of its training data, a foundation often compromised by noise and corruption. Here, through a comparative study of classical and quantum neural networks on both classical and quantum data, we reveal a fundamental difference in their response to data corruption. We find that classical models exhibit brittle memorization, leading to a failure in generalization. In contrast, quantum models demonstrate remarkable resilience, which is underscored by a phase transition-like response to increasing label noise, revealing a critical point beyond which the model's performance changes qualitatively. We further establish and investigate the field of quantum machine unlearning, the process of efficiently forcing a trained model to forget corrupting influences. We show that the brittle nature of the classical model forms rigid, stubborn memories of erroneous data, making efficient unlearning challenging, while the quantum model is significantly more amenable to efficient forgetting with approximate unlearning methods. Our findings establish that quantum machine learning can possess a dual advantage of intrinsic resilience and efficient adaptability, providing a promising paradigm for the trustworthy and robust artificial intelligence of the future.
Authors: Linyu Li, Zhi Jin, Yuanpeng He, Dongming Jin, Yichi Zhang, Haoran Duan, Nyima Tash
Abstract: Since knowledge graphs (KG) will continue to evolve in real scenarios, traditional KGE models are only suitable for static knowledge graphs. Therefore, continual knowledge graph embedding (CKGE) has attracted the attention of researchers. Currently, a key challenge facing CKGE is that the model is prone to "catastrophic forgetting", resulting in the loss of previously learned knowledge. In order to effectively alleviate this problem, we propose a new CKGE model BAKE. First, we note that the Bayesian posterior update principle provides a natural continual learning strategy that is insensitive to data order and can theoretically effectively resist the forgetting of previous knowledge during data evolution. Different from the existing CKGE method, BAKE regards each batch of new data as a Bayesian update of the model prior. Under this framework, as long as the posterior distribution of the model is maintained, the model can better preserve the knowledge of early snapshots even after evolving through multiple time snapshots. Secondly, we propose a continual clustering method for CKGE, which further directly combats knowledge forgetting by constraining the evolution difference (or change amplitude) between new and old knowledge between different snapshots. We conduct extensive experiments on BAKE on multiple datasets, and the results show that BAKE significantly outperforms existing baseline models.
Authors: Miaosen Luo, Jiesen Long, Zequn Li, Yunying Yang, Yuncheng Jiang, Sijie Mai
Abstract: Multimodal Affective Computing (MAC) aims to recognize and interpret human emotions by integrating information from diverse modalities such as text, video, and audio. Recent advancements in Multimodal Large Language Models (MLLMs) have significantly reshaped the landscape of MAC by offering a unified framework for processing and aligning cross-modal information. However, practical challenges remain, including performance variability across complex MAC tasks and insufficient understanding of how architectural designs and data characteristics impact affective analysis. To address these gaps, we conduct a systematic benchmark evaluation of state-of-the-art open-source MLLMs capable of concurrently processing audio, visual, and textual modalities across multiple established MAC datasets. Our evaluation not only compares the performance of these MLLMs but also provides actionable insights into model optimization by analyzing the influence of model architectures and dataset properties. Furthermore, we propose a novel hybrid strategy that combines generative knowledge prompting with supervised fine-tuning to enhance MLLMs' affective computing capabilities. Experimental results demonstrate that this integrated approach significantly improves performance across various MAC tasks, offering a promising avenue for future research and development in this field. Our code is released on https://github.com/LuoMSen/MLLM-MAC.
Authors: Dean Brandner, Sebastien Gros, Sergio Lucia
Abstract: Model predictive control (MPC) is widely used in process control due to its interpretability and ability to handle constraints. As a parametric policy in reinforcement learning (RL), MPC offers strong initial performance and low data requirements compared to black-box policies like neural networks. However, most RL methods rely on first-order updates, which scale well to large parameter spaces but converge at most linearly, making them inefficient when each policy update requires solving an optimal control problem, as is the case with MPC. While MPC policies are typically sparsely parameterized and thus amenable to second-order approaches, existing second-order methods demand second-order policy derivatives, which can be computationally and memory-wise intractable. This work introduces a Gauss-Newton approximation of the deterministic policy Hessian that eliminates the need for second-order policy derivatives, enabling superlinear convergence with minimal computational overhead. To further improve robustness, we propose a momentum-based Hessian averaging scheme for stable training under noisy estimates. We demonstrate the effectiveness of the approach on a nonlinear continuously stirred tank reactor (CSTR), showing faster convergence and improved data efficiency over state-of-the-art first-order methods.
Authors: Xinfang Chen, Siyang Xiao, Xianying Zhu, Junhong Xie, Ming Liang, Dajun Chen, Wei Jiang, Yong Li, Peng Di
Abstract: Code editing, including modifying, refactoring, and maintaining existing code, is the most frequent task in software development and has garnered significant attention from AI-powered tools. However, existing solutions that translate explicit natural language instructions into code edits face critical limitations, such as heavy reliance on human instruction input and high latency, which hinder their effective integration into a developer's workflow. We observe that developers' habitual behaviors and coding objectives are often reflected in their historical editing patterns, making this data key to addressing existing limitations. To leverage these insights, we propose NES (Next Edit Suggestion), an LLM-driven code editing framework that delivers an instruction-free and low-latency experience. Built on a dual-model architecture and trained with our high-quality SFT and DAPO datasets, NES enhances productivity by understanding developer intent while optimizing inference to minimize latency. NES is a scalable, industry-ready solution with a continuous Tab key interaction workflow, seamlessly adopted by a FinTech company with over 20,000 developers. Evaluations on real-world datasets show NES achieves 75.6% and 81.6% accuracy in two tasks of predicting next edit locations, alongside 91.36% ES and 27.7% EMR for intent-aligned edits, outperforming SOTA models. Our open-sourced SFT and DAPO datasets have been demonstrated to enhance the performance of open-source CodeLLMs. The demonstration of NES is available at https://youtu.be/yGoyYOe6fbY.
Authors: Zhan Qu, Shuzhou Yuan, Michael F\"arber
Abstract: This paper presents a systematic investigation into the constrained generation capabilities of large language models (LLMs) in producing Songci, a classical Chinese poetry form characterized by strict structural, tonal, and rhyme constraints defined by Cipai templates. We first develop a comprehensive, multi-faceted evaluation framework that includes: (i) a formal conformity score, (ii) automated quality assessment using LLMs, (iii) human evaluation, and (iv) classification-based probing tasks. Using this framework, we evaluate the generative performance of 18 LLMs, including 3 proprietary models and 15 open-source models across four families, under five prompting strategies: zero-shot, one-shot, completion-based, instruction-tuned, and chain-of-thought. Finally, we propose a Generate-Critic architecture in which the evaluation framework functions as an automated critic. Leveraging the critic's feedback as a reward signal, we fine-tune three lightweight open-source LLMs via supervised fine-tuning (SFT), resulting in improvements of up to 5.88% in formal conformity. Our findings offer new insights into the generative strengths and limitations of LLMs in producing culturally significant and formally constrained literary texts.
Authors: Karthik Peddi, Sai Ram Aditya Parisineni, Hemanth Macharla, Mayukha Pal
Abstract: Causal analysis helps us understand variables that are responsible for system failures. This improves fault detection and makes system more reliable. In this work, we present a new method that combines causal inference with machine learning to classify faults in electrical distribution systems (EDS) using graph-based models. We first build causal graphs using transfer entropy (TE). Each fault case is represented as a graph, where the nodes are features such as voltage and current, and the edges demonstrate how these features influence each other. Then, the graphs are classified using machine learning and GraphSAGE where the model learns from both the node values and the structure of the graph to predict the type of fault. To make the predictions understandable, we further developed an integrated approach using GNNExplainer and Captums Integrated Gradients to highlight the nodes (features) that influences the most on the final prediction. This gives us clear insights into the possible causes of the fault. Our experiments show high accuracy: 99.44% on the EDS fault dataset, which is better than state of art models. By combining causal graphs with machine learning, our method not only predicts faults accurately but also helps understand their root causes. This makes it a strong and practical tool for improving system reliability.
Authors: Jack Merullo, Arjun Khurana, Oliver McLaughlin
Abstract: Large language models demonstrate proficiency on phonetic tasks, such as rhyming, without explicit phonetic or auditory grounding. In this work, we investigate how \verb|Llama-3.2-1B-Instruct| represents token-level phonetic information. Our results suggest that Llama uses a rich internal model of phonemes to complete phonetic tasks. We provide evidence for high-level organization of phoneme representations in its latent space. In doing so, we also identify a ``phoneme mover head" which promotes phonetic information during rhyming tasks. We visualize the output space of this head and find that, while notable differences exist, Llama learns a model of vowels similar to the standard IPA vowel chart for humans, despite receiving no direct supervision to do so.
Authors: Karan Reddy, Mayukha Pal
Abstract: Standard transformer-based language models, while powerful for general text, often struggle with the fine-grained syntax and entity relationships in complex technical, engineering documents. To address this, we propose the Contextual Graph Transformer (CGT), a hybrid neural architecture that combines Graph Neural Networks (GNNs) and Transformers for domain-specific question answering. CGT constructs a dynamic graph over input tokens using sequential, skip-gram, and semantic similarity edges, which is processed by GATv2Conv layers for local structure learning. These enriched embeddings are then passed to a Transformer encoder to capture global dependencies. Unlike generic large models, technical domains often require specialized language models with stronger contextualization and structure awareness. CGT offers a parameter-efficient solution for such use cases. Integrated into a Retrieval-Augmented Generation (RAG) pipeline, CGT outperforms baselines like GPT-2 and BERT, achieving 24.7% higher accuracy than GPT-2 with 62.4% fewer parameters. This gain stems from CGTs ability to jointly model structural token interactions and long-range semantic coherence. The model is trained from scratch using a two-phase approach: pretraining on general text followed by fine-tuning on domain-specific manuals. This highlights CGTs adaptability to technical language, enabling better grounding, entity tracking, and retrieval-augmented responses in real-world applications.
Authors: Phillip Stephan, Florian Euchner, Stephan ten Brink
Abstract: The ability of modern telecommunication systems to locate users and objects in the radio environment raises justified privacy concerns. To prevent unauthorized localization, single-antenna transmitters can obfuscate the signal by convolving it with a randomized sequence prior to transmission, which alters the channel state information (CSI) estimated at the receiver. However, this strategy is only effective against CSI-based localization systems deploying single-antenna receivers. Inspired by the concept of blind multichannel identification, we propose a simple CSI recovery method for multi-antenna receivers to extract channel features that ensure reliable user localization regardless of the transmitted signal. We comparatively evaluate the impact of signal obfuscation and the proposed recovery method on the localization performance of CSI fingerprinting, channel charting, and classical triangulation using real-world channel measurements. This work aims to demonstrate the necessity for further efforts to protect the location privacy of users from adversarial radio-based localization systems.
Authors: Ali Noori, Pratik Devkota, Somya Mohanty, Prashanti Manda
Abstract: Automated annotation of clinical text with standardized medical concepts is critical for enabling structured data extraction and decision support. SNOMED CT provides a rich ontology for labeling clinical entities, but manual annotation is labor-intensive and impractical at scale. This study introduces a neural sequence labeling approach for SNOMED CT concept recognition using a Bidirectional GRU model. Leveraging a subset of MIMIC-IV, we preprocess text with domain-adapted SpaCy and SciBERT-based tokenization, segmenting sentences into overlapping 19-token chunks enriched with contextual, syntactic, and morphological features. The Bi-GRU model assigns IOB tags to identify concept spans and achieves strong performance with a 90 percent F1-score on the validation set. These results surpass traditional rule-based systems and match or exceed existing neural models. Qualitative analysis shows effective handling of ambiguous terms and misspellings. Our findings highlight that lightweight RNN-based architectures can deliver high-quality clinical concept annotation with significantly lower computational cost than transformer-based models, making them well-suited for real-world deployment.
Authors: Eman Alamoudi, Ellis Solaiman
Abstract: Arabic-language patient feedback remains under-analysed because dialect diversity and scarce aspect-level sentiment labels hinder automated assessment. To address this gap, we introduce EHSAN, a data-centric hybrid pipeline that merges ChatGPT pseudo-labelling with targeted human review to build the first explainable Arabic aspect-based sentiment dataset for healthcare. Each sentence is annotated with an aspect and sentiment label (positive, negative, or neutral), forming a pioneering Arabic dataset aligned with healthcare themes, with ChatGPT-generated rationales provided for each label to enhance transparency. To evaluate the impact of annotation quality on model performance, we created three versions of the training data: a fully supervised set with all labels reviewed by humans, a semi-supervised set with 50% human review, and an unsupervised set with only machine-generated labels. We fine-tuned two transformer models on these datasets for both aspect and sentiment classification. Experimental results show that our Arabic-specific model achieved high accuracy even with minimal human supervision, reflecting only a minor performance drop when using ChatGPT-only labels. Reducing the number of aspect classes notably improved classification metrics across the board. These findings demonstrate an effective, scalable approach to Arabic aspect-based sentiment analysis (SA) in healthcare, combining large language model annotation with human expertise to produce a robust and explainable dataset. Future directions include generalisation across hospitals, prompt refinement, and interpretable data-driven modelling.
Authors: Lei Zan, Keli Zhang, Ruichu Cai, Lujia Pan
Abstract: Large Language Models (LLMs) have demonstrated strong performance across a wide range of tasks, yet they still struggle with complex mathematical reasoning, a challenge fundamentally rooted in deep structural dependencies. To address this challenge, we propose \textbf{CA}usal \textbf{MA}thematician (\textbf{CAMA}), a two-stage causal framework that equips LLMs with explicit, reusable mathematical structure. In the learning stage, CAMA first constructs the \textbf{M}athematical \textbf{C}ausal \textbf{G}raph (\textbf{MCG}), a high-level representation of solution strategies, by combining LLM priors with causal discovery algorithms applied to a corpus of question-solution pairs. The resulting MCG encodes essential knowledge points and their causal dependencies. To better align the graph with downstream reasoning tasks, CAMA further refines the MCG through iterative feedback derived from a selected subset of the question-solution pairs. In the reasoning stage, given a new question, CAMA dynamically extracts a task-relevant subgraph from the MCG, conditioned on both the question content and the LLM's intermediate reasoning trace. This subgraph, which encodes the most pertinent knowledge points and their causal dependencies, is then injected back into the LLM to guide its reasoning process. Empirical results on real-world datasets show that CAMA significantly improves LLM performance on challenging mathematical problems. Furthermore, our experiments demonstrate that structured guidance consistently outperforms unstructured alternatives, and that incorporating asymmetric causal relationships yields greater improvements than using symmetric associations alone.
Authors: James Carzon, Luca Masserano, Joshua D. Ingram, Alex Shen, Antonio Carlos Herling Ribeiro Junior, Tommaso Dorigo, Michele Doro, Joshua S. Speagle, Rafael Izbicki, Ann B. Lee
Abstract: Generative artificial intelligence (AI) excels at producing complex data structures (text, images, videos) by learning patterns from training examples. Across scientific disciplines, researchers are now applying generative models to ``inverse problems'' to infer hidden parameters from observed data. While these methods can handle intractable models and large-scale studies, they can also produce biased or overconfident conclusions. We present a solution with Frequentist-Bayes (FreB), a mathematically rigorous protocol that reshapes AI-generated probability distributions into confidence regions that consistently include true parameters with the expected probability, while achieving minimum size when training and target data align. We demonstrate FreB's effectiveness by tackling diverse case studies in the physical sciences: identifying unknown sources under dataset shift, reconciling competing theoretical models, and mitigating selection bias and systematics in observational studies. By providing validity guarantees with interpretable diagnostics, FreB enables trustworthy scientific inference across fields where direct likelihood evaluation remains impossible or prohibitively expensive.
Authors: Yinghao Zhu, Yifan Qi, Zixiang Wang, Lei Gu, Dehao Sui, Haoran Hu, Xichen Zhang, Ziyi He, Liantao Ma, Lequan Yu
Abstract: The efficacy of AI agents in healthcare research is hindered by their reliance on static, predefined strategies. This creates a critical limitation: agents can become better tool-users but cannot learn to become better strategic planners, a crucial skill for complex domains like healthcare. We introduce HealthFlow, a self-evolving AI agent that overcomes this limitation through a novel meta-level evolution mechanism. HealthFlow autonomously refines its own high-level problem-solving policies by distilling procedural successes and failures into a durable, strategic knowledge base. To anchor our research and facilitate reproducible evaluation, we introduce EHRFlowBench, a new benchmark featuring complex, realistic health data analysis tasks derived from peer-reviewed clinical research. Our comprehensive experiments demonstrate that HealthFlow's self-evolving approach significantly outperforms state-of-the-art agent frameworks. This work marks a necessary shift from building better tool-users to designing smarter, self-evolving task-managers, paving the way for more autonomous and effective AI for scientific discovery.
Authors: Ziqin He, Mengqi Hu, Yifei Lou, Can Chen
Abstract: Dynamic mode decomposition (DMD) has become a powerful data-driven method for analyzing the spatiotemporal dynamics of complex, high-dimensional systems. However, conventional DMD methods are limited to matrix-based formulations, which might be inefficient or inadequate for modeling inherently multidimensional data including images, videos, and higher-order networks. In this letter, we propose tensor dynamic mode decomposition (TDMD), a novel extension of DMD to third-order tensors based on the recently developed T-product framework. By incorporating tensor factorization techniques, TDMD achieves more efficient computation and better preservation of spatial and temporal structures in multiway data for tasks such as state reconstruction and dynamic component separation, compared to standard DMD with data flattening. We demonstrate the effectiveness of TDMD on both synthetic and real-world datasets.
Authors: Enrique Valero-Leal, Pedro Larra\~naga, Concha Bielza
Abstract: Counterfactual explanations study what should have changed in order to get an alternative result, enabling end-users to understand machine learning mechanisms with counterexamples. Actionability is defined as the ability to transform the original case to be explained into a counterfactual one. We develop a method for actionable counterfactual explanations that, unlike predecessors, does not directly leverage training data. Rather, data is only used to learn a density estimator, creating a search landscape in which to apply path planning algorithms to solve the problem and masking the endogenous data, which can be sensitive or private. We put special focus on estimating the data density using Bayesian networks, demonstrating how their enhanced interpretability is useful in high-stakes scenarios in which fairness is raising concern. Using a synthetic benchmark comprised of 15 datasets, our proposal finds more actionable and simpler counterfactuals than the current state-of-the-art algorithms. We also test our algorithm with a real-world Environmental Protection Agency dataset, facilitating a more efficient and equitable study of policies to improve the quality of life in United States of America counties. Our proposal captures the interaction of variables, ensuring equity in decisions, as policies to improve certain domains of study (air, water quality, etc.) can be detrimental in others. In particular, the sociodemographic domain is often involved, where we find important variables related to the ongoing housing crisis that can potentially have a severe negative impact on communities.
Authors: Guy Blanc, Cl\'ement L. Canonne, Erik Waingarten
Abstract: In the uniformity testing task, an algorithm is provided with samples from an unknown probability distribution over a (known) finite domain, and must decide whether it is the uniform distribution, or, alternatively, if its total variation distance from uniform exceeds some input distance parameter. This question has received a significant amount of interest and its complexity is, by now, fully settled. Yet, we argue that it fails to capture many scenarios of interest, and that its very definition as a gap problem in terms of a prespecified distance may lead to suboptimal performance. To address these shortcomings, we introduce the problem of uniformity tracking, whereby an algorithm is required to detect deviations from uniformity (however they may manifest themselves) using as few samples as possible, and be competitive against an optimal algorithm knowing the distribution profile in hindsight. Our main contribution is a $\operatorname{polylog}(\operatorname{opt})$-competitive uniformity tracking algorithm. We obtain this result by leveraging new structural results on Poisson mixtures, which we believe to be of independent interest.
Authors: Vahe Gharakhanyan, Yi Yang, Luis Barroso-Luque, Muhammed Shuaibi, Daniel S. Levine, Kyle Michel, Viachaslau Bernat, Misko Dzamba, Xiang Fu, Meng Gao, Xingyu Liu, Keian Noori, Lafe J. Purvis, Tingling Rao, Brandon M. Wood, Ammar Rizvi, Matt Uyttendaele, Andrew J. Ouderkirk, Chiara Daraio, C. Lawrence Zitnick, Arman Boromand, Noa Marom, Zachary W. Ulissi, Anuroop Sriram
Abstract: Crystal Structure Prediction (CSP) of molecular crystals plays a central role in applications, such as pharmaceuticals and organic electronics. CSP is challenging and computationally expensive due to the need to explore a large search space with sufficient accuracy to capture energy differences of a few kJ/mol between polymorphs. Dispersion-inclusive density functional theory (DFT) provides the required accuracy but its computational cost is impractical for a large number of putative structures. We introduce FastCSP, an open-source, high-throughput CSP workflow based on machine learning interatomic potentials (MLIPs). FastCSP combines random structure generation using Genarris 3.0 with geometry relaxation and free energy calculations powered entirely by the Universal Model for Atoms (UMA) MLIP. We benchmark FastCSP on a curated set of 28 mostly rigid molecules, demonstrating that our workflow consistently generates known experimental structures and ranks them within 5 kJ/mol per molecule of the global minimum. Our results demonstrate that universal MLIPs can be used across diverse compounds without requiring system-specific tuning. Moreover, the speed and accuracy afforded by UMA eliminate the need for classical force fields in the early stages of CSP and for final re-ranking with DFT. The open-source release of the entire FastCSP workflow significantly lowers the barrier to accessing CSP. CSP results for a single system can be obtained within hours on tens of modern GPUs, making high-throughput crystal structure prediction feasible for a broad range of scientific applications.
Authors: Zheng Wang (Department of Computer Science, University of Science and Technology Beijing), Xiaojun Ye (School of Software, Tsinghua University), Chaokun Wang (School of Software, Tsinghua University), Jian Cui (Department of Computer Science, University of Science and Technology Beijing), Philip S. Yu (Department of Computer Science, University of Illinois at Chicago)
Abstract: Network embedding, aiming to project a network into a low-dimensional space, is increasingly becoming a focus of network research. Semi-supervised network embedding takes advantage of labeled data, and has shown promising performance. However, existing semi-supervised methods would get unappealing results in the completely-imbalanced label setting where some classes have no labeled nodes at all. To alleviate this, we propose two novel semi-supervised network embedding methods. The first one is a shallow method named RSDNE. Specifically, to benefit from the completely-imbalanced labels, RSDNE guarantees both intra-class similarity and inter-class dissimilarity in an approximate way. The other method is RECT which is a new class of graph neural networks. Different from RSDNE, to benefit from the completely-imbalanced labels, RECT explores the class-semantic knowledge. This enables RECT to handle networks with node features and multi-label setting. Experimental results on several real-world datasets demonstrate the superiority of the proposed methods. Code is available at https://github.com/zhengwang100/RECT.
Authors: Wenbin Zhang
Abstract: While artificial intelligence (AI)-based decision-making systems are increasingly popular, significant concerns on the potential discrimination during the AI decision-making process have been observed. For example, the distribution of predictions is usually biased and dependents on the sensitive attributes (e.g., gender and ethnicity). Numerous approaches have therefore been proposed to develop decision-making systems that are discrimination-conscious by-design, which are typically batch-based and require the simultaneous availability of all the training data for model learning. However, in the real-world, the data streams usually come on the fly which requires the model to process each input data once ``on arrival'' and without the need for storage and reprocessing. In addition, the data streams might also evolve over time, which further requires the model to be able to simultaneously adapt to non-stationary data distributions and time-evolving bias patterns, with an effective and robust trade-off between accuracy and fairness. In this paper, we propose a novel framework of online decision tree with fairness in the data stream with possible distribution drifting. Specifically, first, we propose two novel fairness splitting criteria that encode the data as well as possible, while simultaneously removing dependence on the sensitive attributes, and further adapts to non-stationary distribution with fine-grained control when needed. Second, we propose two fairness decision tree online growth algorithms that fulfills different online fair decision-making requirements. Our experiments show that our algorithms are able to deal with discrimination in massive and non-stationary streaming environments, with a better trade-off between fairness and predictive performance.
Authors: Bei Zhou, S{\o}ren Riis
Abstract: AlphaZero-style reinforcement learning (RL) algorithms have achieved superhuman performance in many complex board games such as Chess, Shogi, and Go. However, we showcase that these algorithms encounter significant and fundamental challenges when applied to impartial games, a class where players share game pieces and optimal strategy often relies on abstract mathematical principles. Specifically, we utilize the game of Nim as a concrete and illustrative case study to reveal critical limitations of AlphaZero-style and similar self-play RL algorithms. We introduce a novel conceptual framework distinguishing between champion and expert mastery to evaluate RL agent performance. Our findings reveal that while AlphaZero-style agents can achieve champion-level play on very small Nim boards, their learning progression severely degrades as the board size increases. This difficulty stems not merely from complex data distributions or noisy labels, but from a deeper representational bottleneck: the inherent struggle of generic neural networks to implicitly learn abstract, non-associative functions like parity, which are crucial for optimal play in impartial games. This limitation causes a critical breakdown in the positive feedback loop essential for self-play RL, preventing effective learning beyond rote memorization of frequently observed states. These results align with broader concerns regarding AlphaZero-style algorithms' vulnerability to adversarial attacks, highlighting their inability to truly master all legal game states. Our work underscores that simple hyperparameter adjustments are insufficient to overcome these challenges, establishing a crucial foundation for the development of fundamentally novel algorithmic approaches, potentially involving neuro-symbolic or meta-learning paradigms, to bridge the gap towards true expert-level AI in combinatorial games.
Authors: Xiao Han, Lu Zhang, Yongkai Wu, Shuhan Yuan
Abstract: Algorithmic recourse provides actionable recommendations to alter unfavorable predictions of machine learning models, enhancing transparency through counterfactual explanations. While significant progress has been made in algorithmic recourse for static data, such as tabular and image data, limited research explores recourse for multivariate time series, particularly for reversing abnormal time series. This paper introduces Recourse in time series Anomaly Detection (RecAD), a framework for addressing anomalies in multivariate time series using backtracking counterfactual reasoning. By modeling the causes of anomalies as external interventions on exogenous variables, RecAD predicts recourse actions to restore normal status as counterfactual explanations, where the recourse function, responsible for generating actions based on observed data, is trained using an end-to-end approach. Experiments on synthetic and real-world datasets demonstrate its effectiveness.
Authors: Wanteng Ma, Dong Xia, Jiashuo Jiang
Abstract: We investigate the contextual bandits with knapsack (CBwK) problem in a high-dimensional linear setting, where the feature dimension can be very large. Our goal is to harness sparsity to obtain sharper regret guarantees. To this end, we first develop an online variant of the hard thresholding algorithm that performs the sparse estimation in an online manner. We then embed this estimator in a primal-dual scheme: every knapsack constraint is paired with a dual variable, which is updated by an online learning rule to keep the cumulative resource consumption within budget. This integrated approach achieves a two-phase sub-linear regret that scales only logarithmically with the feature dimension, improving on the polynomial dependency reported in prior work. Furthermore, we show that either of the following structural assumptions is sufficient for a sharper regret bound of $\tilde{O}(s_{0} \sqrt{T})$: (i) a diverse-covariate condition; and (ii) a margin condition. When both conditions hold simultaneously, we can further control the regret to $O(s_{0}^{2} \log(dT)\log T)$ by a dual resolving scheme. As a by-product, applying our framework to high-dimensional contextual bandits without knapsack constraints recovers the optimal regret rates in both the data-poor and data-rich regimes. Finally, numerical experiments confirm the empirical efficiency of our algorithms in high-dimensional settings.
Authors: Zhichun Guo, Tong Zhao, Yozen Liu, Kaiwen Dong, William Shiao, Mingxuan Ju, Neil Shah, Nitesh V. Chawla
Abstract: Graph Neural Networks (GNNs) are prominent in graph machine learning and have shown state-of-the-art performance in Link Prediction (LP) tasks. Nonetheless, recent studies show that GNNs struggle to produce good results on low-degree nodes despite their overall strong performance. In practical applications of LP, like recommendation systems, improving performance on low-degree nodes is critical, as it amounts to tackling the cold-start problem of improving the experiences of users with few observed interactions. In this paper, we investigate improving GNNs' LP performance on low-degree nodes while preserving their performance on high-degree nodes and propose a simple yet surprisingly effective augmentation technique called NodeDup. Specifically, NodeDup duplicates low-degree nodes and creates links between nodes and their own duplicates before following the standard supervised LP training scheme. By leveraging a ''multi-view'' perspective for low-degree nodes, NodeDup shows significant LP performance improvements on low-degree nodes without compromising any performance on high-degree nodes. Additionally, as a plug-and-play augmentation module, NodeDup can be easily applied to existing GNNs with very light computational cost. Extensive experiments show that NodeDup achieves 38.49%, 13.34%, and 6.76% improvements on isolated, low-degree, and warm nodes, respectively, on average across all datasets compared to GNNs and state-of-the-art cold-start methods.
Authors: Georgia Papacharalampous, Hristos Tyralis, Nikolaos Doulamis, Anastasios Doulamis
Abstract: Predictions in the form of probability distributions are crucial for effective decision-making. Quantile regression enables such predictions within spatial prediction settings that aim to create improved precipitation datasets by merging remote sensing and gauge data. However, ensemble learning of quantile regression algorithms remains unexplored in this context and, at the same time, it has not been substantially developed so far in the broader machine learning research landscape. Here, we introduce nine quantile-based ensemble learners and address the aforementioned gap in precipitation dataset creation by presenting the first application of these learners to large precipitation datasets. We employed a novel feature engineering strategy, which reduces the number of predictors by using distance-weighted satellite precipitation at relevant locations, combined with location elevation. Our ensemble learners include six that are based on stacking ideas and three simple methods (mean, median, best combiner). Each of them combines the following six individual algorithms: quantile regression (QR), quantile regression forests (QRF), generalized random forests (GRF), gradient boosting machines (GBM), light gradient boosting machines (LightGBM), and quantile regression neural networks (QRNN). These algorithms serve as both base learners and combiners within different ensemble learning methods. We evaluated performance against a reference method (i.e., QR) using quantile scoring functions and a large dataset. The latter comprises 15 years of monthly gauge-measured and satellite precipitation in the contiguous United States (CONUS). Ensemble learning with QR and QRNN yielded the best results across the various investigated quantile levels, which range from 0.025 to 0.975, outperforming the reference method by 3.91% to 8.95%...
Authors: Huiling Meng, Ningyuan Chen, Xuefeng Gao
Abstract: Intensity control is a type of continuous-time dynamic optimization problems with many important applications in Operations Research including queueing and revenue management. In this study, we adapt the reinforcement learning framework to intensity control using choice-based network revenue management as a case study, which is a classical problem in revenue management that features a large state space, a large action space and a continuous time horizon. We show that by utilizing the inherent discretization of the sample paths created by the jump points, a unique and defining feature of intensity control, one does not need to discretize the time horizon in advance, which was believed to be necessary because most reinforcement learning algorithms are designed for discrete-time problems. As a result, the computation can be facilitated and the discretization error is significantly reduced. We lay the theoretical foundation for the Monte Carlo and temporal difference learning algorithms for policy evaluation and develop policy-gradient-based actor-critic algorithms for intensity control. Via a comprehensive numerical study, we demonstrate the benefit of our approach versus other state-of-the-art benchmarks.
Authors: Lukas Weigand, Tim Roith, Martin Burger
Abstract: A popular method to perform adversarial attacks on neuronal networks is the so-called fast gradient sign method and its iterative variant. In this paper, we interpret this method as an explicit Euler discretization of a differential inclusion, where we also show convergence of the discretization to the associated gradient flow. To do so, we consider the concept of p-curves of maximal slope in the case $p=\infty$. We prove existence of $\infty$-curves of maximum slope and derive an alternative characterization via differential inclusions. Furthermore, we also consider Wasserstein gradient flows for potential energies, where we show that curves in the Wasserstein space can be characterized by a representing measure on the space of curves in the underlying Banach space, which fulfill the differential inclusion. The application of our theory to the finite-dimensional setting is twofold: On the one hand, we show that a whole class of normalized gradient descent methods (in particular signed gradient descent) converge, up to subsequences, to the flow, when sending the step size to zero. On the other hand, in the distributional setting, we show that the inner optimization task of adversarial training objective can be characterized via $\infty$-curves of maximum slope on an appropriate optimal transport space.
Authors: Gyuejeong Lee, Daeyoung Choi
Abstract: Federated learning (FL) enables collaborative model training across distributed clients without centralizing data. However, existing approaches such as Federated Averaging (FedAvg) often perform poorly with heterogeneous data distributions, failing to achieve personalization owing to their inability to capture class-specific information effectively. We propose Class-wise Federated Averaging (cwFedAvg), a novel personalized FL (PFL) framework that performs Federated Averaging for each class, to overcome the personalization limitations of FedAvg. cwFedAvg creates class-specific global models via weighted aggregation of local models using class distributions, and subsequently combines them to generate personalized local models. We further propose Weight Distribution Regularizer (WDR), which encourages deep networks to encode class-specific information efficiently by aligning empirical and approximated class distributions derived from output layer weights, to facilitate effective class-wise aggregation. Our experiments demonstrate the superior performance of cwFedAvg with WDR over existing PFL methods through efficient personalization while maintaining the communication cost of FedAvg and avoiding additional local training and pairwise computations.
Authors: Wen-Tse Chen, Yuxuan Li, Shiyu Huang, Jiayu Chen, Jeff Schneider
Abstract: Multi-agent credit assignment is a fundamental challenge for cooperative multi-agent reinforcement learning (MARL), where a team of agents learn from shared reward signals. The Individual-Global-Max (IGM) condition is a widely used principle for multi-agent credit assignment, requiring that the joint action determined by individual Q-functions maximizes the global Q-value. Meanwhile, the principle of maximum entropy has been leveraged to enhance exploration in MARL. However, we identify a critical limitation in existing maximum entropy MARL methods: a misalignment arises between local policies and the joint policy that maximizes the global Q-value, leading to violations of the IGM condition. To address this misalignment, we propose an order-preserving transformation. Building on it, we introduce ME-IGM, a novel maximum entropy MARL algorithm compatible with any credit assignment mechanism that satisfies the IGM condition while enjoying the benefits of maximum entropy exploration. We empirically evaluate two variants of ME-IGM: ME-QMIX and ME-QPLEX, in non-monotonic matrix games, and demonstrate their state-of-the-art performance across 17 scenarios in SMAC-v2 and Overcooked.
Authors: Ivan Karpukhin, Foma Shipilov, Andrey Savchenko
Abstract: Forecasting multiple future events within a given time horizon is essential for applications in finance, retail, social networks, and healthcare. Marked Temporal Point Processes (MTPP) provide a principled framework to model both the timing and labels of events. However, most existing research focuses on predicting only the next event, leaving long-horizon forecasting largely underexplored. To address this gap, we introduce HoTPP, the first benchmark specifically designed to rigorously evaluate long-horizon predictions. We identify shortcomings in widely used evaluation metrics, propose a theoretically grounded T-mAP metric, present strong statistical baselines, and offer efficient implementations of popular models. Our empirical results demonstrate that modern MTPP approaches often underperform simple statistical baselines. Furthermore, we analyze the diversity of predicted sequences and find that most methods exhibit mode collapse. Finally, we analyze the impact of autoregression and intensity-based losses on prediction quality, and outline promising directions for future research. The HoTPP source code, hyperparameters, and full evaluation results are available at GitHub.
Authors: Kassi Muhammad, Teef David, Giulia Nassisid, Tina Farus
Abstract: As telecommunications networks become increasingly complex, the integration of advanced technologies such as network digital twins and generative artificial intelligence (AI) emerges as a pivotal solution to enhance network operations and resilience. This paper explores the synergy between network digital twins, which provide a dynamic virtual representation of physical networks, and generative AI, particularly focusing on Generative Adversarial Networks (GANs) and Variational Autoencoders (VAEs). We propose a novel architectural framework that incorporates these technologies to significantly improve predictive maintenance, network scenario simulation, and real-time data-driven decision-making. Through extensive simulations, we demonstrate how generative AI can enhance the accuracy and operational efficiency of network digital twins, effectively handling real-world complexities such as unpredictable traffic loads and network failures. The findings suggest that this integration not only boosts the capability of digital twins in scenario forecasting and anomaly detection but also facilitates a more adaptive and intelligent network management system.
Authors: Federico Arangath Joseph, Kilian Konstantin Haefeli, Noah Liniger, Caglar Gulcehre
Abstract: This work explores the in-context learning capabilities of State Space Models (SSMs) and presents, to the best of our knowledge, the first theoretical explanation of a possible underlying mechanism. We introduce a novel weight construction for SSMs, enabling them to predict the next state of any dynamical system after observing previous states without parameter fine-tuning. This is accomplished by extending the HiPPO framework to demonstrate that continuous SSMs can approximate the derivative of any input signal. Specifically, we find an explicit weight construction for continuous SSMs and provide an asymptotic error bound on the derivative approximation. The discretization of this continuous SSM subsequently yields a discrete SSM that predicts the next state. Finally, we demonstrate the effectiveness of our parameterization empirically. This work should be an initial step toward understanding how sequence models based on SSMs learn in context.
Authors: Gongpu Chen, Soung Chang Liew, Deniz Gunduz
Abstract: The restless multi-armed bandit (RMAB) framework is a popular model with applications across a wide variety of fields. However, its solution is hindered by the exponentially growing state space (with respect to the number of arms) and the combinatorial action space, making traditional reinforcement learning methods infeasible for large-scale instances. In this paper, we propose GINO-Q, a three-timescale stochastic approximation algorithm designed to learn an asymptotically optimal index policy for RMABs. GINO-Q mitigates the curse of dimensionality by decomposing the RMAB into a series of subproblems, each with the same dimension as a single arm, ensuring that complexity increases linearly with the number of arms. Unlike recently developed Whittle-index-based algorithms, GINO-Q does not require RMABs to be indexable, enhancing its flexibility and applicability. Our experimental results demonstrate that GINO-Q consistently learns near-optimal policies, even for non-indexable RMABs where Whittle-index-based algorithms perform poorly, and it converges significantly faster than existing baselines.
Authors: Kazuhisa Fujita
Abstract: We present three game-playing agents incorporating Vision Transformers (ViT) into the AlphaZero framework: AlphaViT, AlphaViD (AlphaViT with a transformer decoder), and AlphaVDA (AlphaViD with learnable action embeddings). These agents can play multiple board games of varying sizes using a single neural network with shared weights, thus overcoming AlphaZero's limitation of fixed board sizes. AlphaViT employs only a transformer encoder, whereas AlphaViD and AlphaVDA incorporate both a transformer encoder and a decoder. In AlphaViD, the decoder processes outputs from the encoder, whereas AlphaVDA uses learnable embeddings as the decoder inputs. The additional decoder in AlphaViD and AlphaVDA provides flexibility to adapt to various action spaces and board sizes. Experimental results show that the proposed agents, trained on either individual games or on multiple games simultaneously, consistently outperform traditional algorithms, such as Minimax and Monte Carlo Tree Search. They approach the performance of AlphaZero despite relying on a single deep neural network (DNN) with shared weights. In particular, AlphaViT performs strongly across all evaluated games. Furthermore, fine-tuning the DNN with weights pre-trained on small board games accelerates convergence and improves performance, particularly in Gomoku. Interestingly, simultaneous training on multiple games yields performance comparable to, or even surpassing, that of single-game training. These results indicate the potential of transformer-based architectures for developing more flexible and robust game-playing AI agents that excel in multiple games and dynamic environments.
Authors: Shuo Lu, Yingsheng Wang, Lijun Sheng, Lingxiao He, Aihua Zheng, Jian Liang
Abstract: Out-of-distribution (OOD) detection aims to detect test samples outside the training category space, which is an essential component in building reliable machine learning systems. Existing reviews on OOD detection primarily focus on method taxonomy, surveying the field by categorizing various approaches. However, many recent works concentrate on non-traditional OOD detection scenarios, such as test-time adaptation, multi-modal data sources and other novel contexts. In this survey, we uniquely review recent advances in OOD detection from the task-oriented perspective for the first time. According to the user's access to the model, that is, whether the OOD detection method is allowed to modify or retrain the model, we classify the methods as training-driven or training-agnostic. Besides, considering the rapid development of pre-trained models, large pre-trained model-based OOD detection is also regarded as an important category and discussed separately. Furthermore, we provide a discussion of the evaluation scenarios, a variety of applications, and several future research directions. We believe this survey with new taxonomy will benefit the proposal of new methods and the expansion of more practical scenarios. A curated list of related papers is provided in the Github repository: https://github.com/shuolucs/Awesome-Out-Of-Distribution-Detection.
URLs: https://github.com/shuolucs/Awesome-Out-Of-Distribution-Detection.
Authors: Zhonghao Shi, Xuan Shi, Anfeng Xu, Tiantian Feng, Harshvardhan Srivastava, Shrikanth Narayanan, Maja J. Matari\'c
Abstract: Automatic speech recognition (ASR) models often experience performance degradation due to data domain shifts introduced at test time, a challenge that is further amplified for child speakers. Test-time adaptation (TTA) methods have shown great potential in bridging this domain gap. However, the use of TTA to adapt ASR models to the individual differences in each child's speech has not yet been systematically studied. In this work, we investigate the effectiveness of two widely used TTA methods-SUTA, SGEM-in adapting off-the-shelf ASR models and their fine-tuned versions for child speech recognition, with the goal of enabling continuous, unsupervised adaptation at test time. Our findings show that TTA significantly improves the performance of both off-the-shelf and fine-tuned ASR models, both on average and across individual child speakers, compared to unadapted baselines. However, while TTA helps adapt to individual variability, it may still be limited with non-linguistic child speech.
Authors: Yinghui Huang, Zihao Tang, Xiangyu Chang
Abstract: The biases and discrimination of machine learning algorithms have attracted significant attention, leading to the development of various algorithms tailored to specific contexts. However, these solutions often fall short of addressing fairness issues inherent in machine learning operations. In this paper, we present an adaptive debiasing framework designed to find an optimal fair transformation of input data that maximally preserves data predictability under dynamic conditions. A distinctive feature of our approach is its flexibility and efficiency. It can be integrated with pretrained black-box classifiers, providing fairness guarantees with minimal retraining efforts, even in the face of frequent data drifts, evolving fairness requirements, and batches of similar tasks. To achieve this, we leverage the normalizing flows to enable efficient, information-preserving data transformation, ensuring that no critical information is lost during the debiasing process. Additionally, we incorporate the Wasserstein distance as the fairness measure to guide the optimization of data transformations. Finally, we introduce an efficient optimization algorithm with closed-formed gradient computations, making our framework scalable and suitable for dynamic, real-world environments.
Authors: Thao Nguyen, Kuan-Hao Huang, Ge Liu, Martin D. Burke, Ying Diao, Heng Ji
Abstract: We introduce Functional Group-Aware Representations for Small Molecules (FARM), a novel foundation model designed to bridge the gap between SMILES, natural language, and molecular graphs. The key innovation of FARM lies in its functional group-aware tokenization, which directly incorporates functional group information into SMILES, enriching SMILES with detailed chemical context. For example, instead of using "O" to represent all oxygen atoms, we use specific tokens like "O_ketone" and "O_hydroxyl" to differentiate oxygen atoms belonging to distinct functional groups. This tokenization expands the chemical lexicon, effectively bridging the gap between SMILES and natural language in terms of vocabulary size, ultimately enhancing the model's ability to predict molecular properties. FARM also represents molecules from two perspectives: by (1) using masked language modeling to capture atom-level features and (2) employing graph neural networks to encode the whole molecule topology. FARM leverages contrastive learning to aligns these two views of representations into a unified molecular embedding. We rigorously evaluate FARM on the MoleculeNet dataset, where it achieves state-of-the-art performance on 11 out of 13 tasks. These results highlight FARM's potential to improve molecular representation learning and demonstrate its strong transfer learning capabilities, paving the way for promising applications in drug discovery and pharmaceutical research.
Authors: Jun Li, Aaron Aguirre, Junior Moura, Che Liu, Lanhai Zhong, Chenxi Sun, Gari Clifford, Brandon Westover, Shenda Hong
Abstract: Artificial intelligence (AI) has demonstrated significant potential in ECG analysis and cardiovascular disease assessment. Recently, foundation models have played a remarkable role in advancing medical AI. The development of an ECG foundation model holds the promise of elevating AI-ECG research to new heights. However, building such a model faces several challenges, including insufficient database sample sizes and inadequate generalization across multiple domains. Additionally, there is a notable performance gap between single-lead and multi-lead ECG analyses. We introduced an ECG Foundation Model (ECGFounder), a general-purpose model that leverages real-world ECG annotations from cardiology experts to broaden the diagnostic capabilities of ECG analysis. ECGFounder was trained on over 10 million ECGs with 150 label categories from the Harvard-Emory ECG Database, enabling comprehensive cardiovascular disease diagnosis through ECG analysis. The model is designed to be both an effective out-of-the-box solution, and a to be fine-tunable for downstream tasks, maximizing usability. Importantly, we extended its application to lower rank ECGs, and arbitrary single-lead ECGs in particular. ECGFounder is applicable to supporting various downstream tasks in mobile monitoring scenarios. Experimental results demonstrate that ECGFounder achieves expert-level performance on internal validation sets, with AUROC exceeding 0.95 for eighty diagnoses. It also shows strong classification performance and generalization across various diagnoses on external validation sets. When fine-tuned, ECGFounder outperforms baseline models in demographic analysis, clinical event detection, and cross-modality cardiac rhythm diagnosis. The trained model and data will be publicly released upon publication through the bdsp.io. Our code is available at https://github.com/PKUDigitalHealth/ECGFounder
Authors: Tri Minh Triet Pham, Karthikeyan Premkumar, Mohamed Naili, Jinqiu Yang
Abstract: With the boom of machine learning (ML) techniques, software practitioners build ML systems to process the massive volume of streaming data for diverse software engineering tasks such as failure prediction in AIOps. Trained using historical data, such ML models encounter performance degradation caused by concept drift, i.e., data and inter-relationship (concept) changes between training and production. It is essential to use concept rift detection to monitor the deployed ML models and re-train the ML models when needed. In this work, we explore applying state-of-the-art (SOTA) concept drift detection techniques on synthetic and real-world datasets in an industrial setting. Such an industrial setting requires minimal manual effort in labeling and maximal generality in ML model architecture. We find that current SOTA semi-supervised methods not only require significant labeling effort but also only work for certain types of ML models. To overcome such limitations, we propose a novel model-agnostic technique (CDSeer) for detecting concept drift. Our evaluation shows that CDSeer has better precision and recall compared to the state-of-the-art while requiring significantly less manual labeling. We demonstrate the effectiveness of CDSeer at concept drift detection by evaluating it on eight datasets from different domains and use cases. Results from internal deployment of CDSeer on an industrial proprietary dataset show a 57.1% improvement in precision while using 99% fewer labels compared to the SOTA concept drift detection method. The performance is also comparable to the supervised concept drift detection method, which requires 100% of the data to be labeled. The improved performance and ease of adoption of CDSeer are valuable in making ML systems more reliable.
Authors: Peijia Qin, Ruiyi Zhang, Pengtao Xie
Abstract: Parameter-efficient fine-tuning (PEFT) is a flexible and efficient method for adapting large language models (LLMs) to downstream tasks. Among these methods, weight-decomposed low-rank adaptation (DoRA) is a promising approach that decomposes weight matrices into magnitude and direction components to mimic full fine-tuning (FT) better. However, DoRA's simultaneous optimization of these components makes it over-expressive, increases the risk of overfitting, and creates a coupled updating pattern that limits its learning capacity. To address these issues, we propose Bi-level Optimization-Based Weight-Decomposed Low-Rank Adaptation (BiDoRA), a novel PEFT method based on a bi-level optimization framework. BiDoRA fundamentally differs from DoRA by optimizing the magnitude and direction in two separate, asynchronous loops using distinct training and validation data splits. This decoupled optimization process effectively mitigates overfitting and allows for more flexible updates that align even more closely with FT. For instance, weight decomposition analysis shows BiDoRA achieves a magnitude-direction update correlation of $-8.042$, significantly closer to the FT ideal compared to $-1.784$ for DoRA. Evaluation of BiDoRA on diverse tasks spanning natural language understanding, generation, token classification, and extremely small biomedical datasets reveals that it consistently outperforms DoRA and a wide range of leading PEFT methods. This improvement is statistically significant, as demonstrated on the GLUE benchmark where BiDoRA surpasses DoRA with a p-value of $2.4\times10^{-4}$ in terms of the Wilcoxon signed-rank test. The code for BiDoRA is available at https://github.com/t2ance/BiDoRA.
Authors: Haoye Chai, Shiyuan Zhang, Xiaoqian Qi, Baohua Qiu, Yong Li
Abstract: Mobile traffic forecasting allows operators to anticipate network dynamics and performance in advance, offering substantial potential for enhancing service quality and improving user experience. However, existing models are often task-oriented and are trained with tailored data, which limits their effectiveness in diverse mobile network tasks of Base Station (BS) deployment, resource allocation, energy optimization, etc. and hinders generalization across different urban environments. Foundation models have made remarkable strides across various domains of NLP and CV due to their multi-tasking adaption and zero/few-shot learning capabilities. In this paper, we propose an innovative Foundation model for Mo}bile traffic forecasting (FoMo), aiming to handle diverse forecasting tasks of short/long-term predictions and distribution generation across multiple cities to support network planning and optimization. FoMo combines diffusion models and transformers, where various spatio-temporal masks are proposed to enable FoMo to learn intrinsic features of different tasks, and a contrastive learning strategy is developed to capture the correlations between mobile traffic and urban contexts, thereby improving its transfer learning capability. Extensive experiments on 9 real-world datasets demonstrate that FoMo outperforms current models concerning diverse forecasting tasks and zero/few-shot learning, showcasing a strong universality.
Authors: Xinyue Feng, Shuxin Zhong, Jinquan Hang, Wenjun Lyu, Yuequn Zhang, Guang Yang, Haotian Wang, Desheng Zhang, Guang Wang
Abstract: Customer expansion, i.e., growing a business existing customer base by acquiring new customers, is critical for scaling operations and sustaining the long-term profitability of logistics companies. Although state-of-the-art works model this task as a single-node classification problem under a heterogeneous graph learning framework and achieve good performance, they struggle with extremely positive label sparsity issues in our scenario. Multi-task learning (MTL) offers a promising solution by introducing a correlated, label-rich task to enhance the label-sparse task prediction through knowledge sharing. However, existing MTL methods result in performance degradation because they fail to discriminate task-shared and task-specific structural patterns across tasks. This issue arises from their limited consideration of the inherently complex structure learning process of heterogeneous graph neural networks, which involves the multi-layer aggregation of multi-type relations. To address the challenge, we propose a Structure-Aware Hierarchical Information Sharing Framework (SrucHIS), which explicitly regulates structural information sharing across tasks in logistics customer expansion. SrucHIS breaks down the structure learning phase into multiple stages and introduces sharing mechanisms at each stage, effectively mitigating the influence of task-specific structural patterns during each stage. We evaluate StrucHIS on both private and public datasets, achieving a 51.41% average precision improvement on the private dataset and a 10.52% macro F1 gain on the public dataset. StrucHIS is further deployed at one of the largest logistics companies in China and demonstrates a 41.67% improvement in the success contract-signing rate over existing strategies, generating over 453K new orders within just two months.
Authors: Ayan Sengupta, Vaibhav Seth, Arinjay Pathak, Aastha Verma, Natraj Raman, Sriram Gopalakrishnan, Niladri Chatterjee, Tanmoy Chakraborty
Abstract: Large Language Models (LLMs) are highly resource-intensive to fine-tune due to their enormous size. While low-rank adaptation is a prominent parameter-efficient fine-tuning approach, it suffers from sensitivity to hyperparameter choices, leading to instability in model performance on fine-tuning downstream tasks. This paper highlights the importance of effective parameterization in low-rank fine-tuning to reduce estimator variance and enhance the stability of final model outputs. We propose MonteCLoRA, an efficient fine-tuning technique that employs Monte Carlo estimation to learn an unbiased posterior estimation of low-rank parameters with low expected variance, stabilizing fine-tuned LLMs with only O(r) additional parameters, for a given rank r. MonteCLoRA shows 0.5% and 1.6% improvements in accuracy and robustness over unregularized low-rank adaptation method on natural language understanding tasks with pre-trained RoBERTa-base. Furthermore, in generative tasks with pre-trained LLaMA-1-7B and LLaMA-3.2-3B-Instruct, MonteCLoRA demonstrates robust performance with 50% and 62% lower spreads respectively than the contemporary efficient fine-tuning methods. The theoretical and empirical results presented in the paper underscore how parameterization and hyperpriors balance exploration-exploitation in the low-rank parametric space, therefore leading to more optimal and robust parameter estimation during efficient fine-tuning.
Authors: Jesse He, Helen Jenne, Herman Chau, Davis Brown, Mark Raugas, Sara Billey, Henry Kvinge
Abstract: Machine learning is becoming an increasingly valuable tool in mathematics, enabling one to identify subtle patterns across collections of examples so vast that they would be impossible for a single researcher to feasibly review and analyze. In this work, we use graph neural networks to investigate \emph{quiver mutation} -- an operation that transforms one quiver (or directed multigraph) into another -- which is central to the theory of cluster algebras with deep connections to geometry, topology, and physics. In the study of cluster algebras, the question of \emph{mutation equivalence} is of fundamental concern: given two quivers, can one efficiently determine if one quiver can be transformed into the other through a sequence of mutations? In this paper, we use graph neural networks and AI explainability techniques to independently discover mutation equivalence criteria for quivers of type $\tilde{D}$. Along the way, we also show that even without explicit training to do so, our model captures structure within its hidden representation that allows us to reconstruct known criteria from type $D$, adding to the growing evidence that modern machine learning models are capable of learning abstract and parsimonious rules from mathematical data.
Authors: Junjie Shan, Ziqi Zhao, Jialin Lu, Rui Zhang, Siu Ming Yiu, Ka-Ho Chow
Abstract: Foundation models that bridge vision and language have made significant progress. While they have inspired many life-enriching applications, their potential for abuse in creating new threats remains largely unexplored. In this paper, we reveal that vision-language models (VLMs) can be weaponized to enhance gradient inversion attacks (GIAs) in federated learning (FL), where an FL server attempts to reconstruct private data samples from gradients shared by victim clients. Despite recent advances, existing GIAs struggle to reconstruct high-resolution images when the victim has a large local data batch. One promising direction is to focus reconstruction on valuable samples rather than the entire batch, but current methods lack the flexibility to target specific data of interest. To address this gap, we propose Geminio, the first approach to transform GIAs into semantically meaningful, targeted attacks. It enables a brand new privacy attack experience: attackers can describe, in natural language, the data they consider valuable, and Geminio will prioritize reconstruction to focus on those high-value samples. This is achieved by leveraging a pretrained VLM to guide the optimization of a malicious global model that, when shared with and optimized by a victim, retains only gradients of samples that match the attacker-specified query. Geminio can be launched at any FL round and has no impact on normal training (i.e., the FL server can steal clients' data while still producing a high-utility ML model as in benign scenarios). Extensive experiments demonstrate its effectiveness in pinpointing and reconstructing targeted samples, with high success rates across complex datasets and large batch sizes with resilience against defenses.
Authors: Divya Anand Sinha, Ruijie Du, Yezi Liu, Athina Markopolou, Yanning Shen
Abstract: Graph federated learning is of essential importance for training over large graph datasets while protecting data privacy, where each client stores a subset of local graph data, while the server collects the local gradients and broadcasts only the aggregated gradients. Recent studies reveal that a malicious attacker can steal private image data from the gradient exchange of neural networks during federated learning. However, the vulnerability of graph data and graph neural networks under such attacks, i.e., reconstructing both node features and graph structure from gradients, remains largely underexplored. To answer this question, this paper studies the problem of whether private data can be reconstructed from leaked gradients in both node classification and graph classification tasks and proposes a novel attack named Graph Leakage from Gradients (GLG). Two widely used GNN frameworks are analyzed, namely GCN and GraphSAGE. The effects of different model settings on reconstruction are extensively discussed. Theoretical analysis and empirical validation demonstrate that, by leveraging the unique properties of graph data and GNNs, GLG achieves more accurate reconstruction of both nodal features and graph structure from gradients.
Authors: Long Qian, Bingke Zhu, Yingying Chen, Ming Tang, Jinqiao Wang
Abstract: Overfitting has traditionally been viewed as detrimental to anomaly detection, where excessive generalization often limits models' sensitivity to subtle anomalies. Our work challenges this conventional view by introducing Controllable Overfitting-based Anomaly Detection (COAD), a novel framework that strategically leverages overfitting to enhance anomaly discrimination capabilities. We propose the Aberrance Retention Quotient (ARQ), a novel metric that systematically quantifies the extent of overfitting, enabling the identification of an optimal golden overfitting interval wherein model sensitivity to anomalies is maximized without sacrificing generalization. To comprehensively capture how overfitting affects detection performance, we further propose the Relative Anomaly Distribution Index (RADI), a metric superior to traditional AUROC by explicitly modeling the separation between normal and anomalous score distributions. Theoretically, RADI leverages ARQ to track and evaluate how overfitting impacts anomaly detection, offering an integrated approach to understanding the relationship between overfitting dynamics and model efficacy. We also rigorously validate the statistical efficacy of Gaussian noise as pseudo-anomaly generators, reinforcing the method's broad applicability. Empirical evaluations demonstrate that our controllable overfitting method achieves State-Of-The-Art(SOTA) performance in both one-class and multi-class anomaly detection tasks, thus redefining overfitting as a powerful strategy rather than a limitation.
Authors: Aman Patel, Arpita Singhal, Austin Wang, Anusri Pampari, Maya Kasowski, Anshul Kundaje
Abstract: Recent advances in self-supervised models for natural language, vision, and protein sequences have inspired the development of large genomic DNA language models (DNALMs). These models aim to learn generalizable representations of diverse DNA elements, potentially enabling various genomic prediction, interpretation and design tasks. Despite their potential, existing benchmarks do not adequately assess the capabilities of DNALMs on key downstream applications involving an important class of non-coding DNA elements critical for regulating gene activity. In this study, we introduce DART-Eval, a suite of representative benchmarks specifically focused on regulatory DNA to evaluate model performance across zero-shot, probed, and fine-tuned scenarios against contemporary ab initio models as baselines. Our benchmarks target biologically meaningful downstream tasks such as functional sequence feature discovery, predicting cell-type specific regulatory activity, and counterfactual prediction of the impacts of genetic variants. We find that current DNALMs exhibit inconsistent performance and do not offer compelling gains over alternative baseline models for most tasks, while requiring significantly more computational resources. We discuss potentially promising modeling, data curation, and evaluation strategies for the next generation of DNALMs. Our code is available at https://github.com/kundajelab/DART-Eval.
Authors: Zijian Zhao, Fanyi Meng, Zhonghao Lyu, Hang Li, Xiaoyang Li, Guangxu Zhu
Abstract: Channel state information (CSI) is a fundamental component in both wireless communication and sensing systems, enabling critical functions such as radio resource optimization and environmental perception. In wireless sensing, data scarcity and packet loss hinder efficient model training, while in wireless communication, high-dimensional CSI matrices and short coherent times caused by high mobility present challenges in CSI estimation.To address these issues, we propose a unified framework named CSI-BERT2 for CSI prediction and classification tasks. Building on CSI-BERT, we introduce a two-stage training method that first uses a mask language model (MLM) to enable the model to learn general feature extraction from scarce datasets in an unsupervised manner, followed by fine-tuning for specific downstream tasks. Specifically, we extend MLM into a mask prediction model (MPM), which efficiently addresses the CSI prediction task. We also introduce an adaptive re-weighting layer (ARL) to enhance subcarrier representation and a multi-layer perceptron (MLP) based temporal embedding module to mitigate permutation invariance issues in time-series CSI data. This significantly improves the CSI classification performance of the original CSI-BERT model. Extensive experiments on both real-world collected and simulated datasets demonstrate that CSI-BERT2 achieves state-of-the-art performance across all tasks. Our results further show that CSI-BERT2 generalizes effectively across varying sampling rates and robustly handles discontinuous CSI sequences caused by packet loss-challenges that conventional methods fail to address.
Authors: Pengxiang Li, Lu Yin, Shiwei Liu
Abstract: Large Language Models (LLMs) have achieved remarkable success, yet recent findings reveal that their deeper layers often contribute minimally and can be pruned without affecting overall performance. While some view this as an opportunity for model compression, we identify it as a training shortfall rooted in the widespread use of Pre-Layer Normalization (Pre-LN). We demonstrate that Pre-LN, commonly employed in models like GPT and LLaMA, leads to diminished gradient norms in its deeper layers, reducing their effectiveness. In contrast, Post-Layer Normalization (Post-LN) preserves larger gradient norms in deeper layers but suffers from vanishing gradients in earlier layers. To address this, we introduce Mix-LN, a novel normalization technique that combines the strengths of Pre-LN and Post-LN within the same model. Mix-LN applies Post-LN to the earlier layers and Pre-LN to the deeper layers, ensuring more uniform gradients across layers. This allows all parts of the network--both shallow and deep layers--to contribute effectively to training. Extensive experiments with various model sizes from 70M to 7B demonstrate that Mix-LN consistently outperforms both Pre-LN and Post-LN, promoting more balanced, healthier gradient norms throughout the network, and enhancing the overall quality of LLM pre-training. Furthermore, we demonstrate that models pre-trained with Mix-LN learn better compared to those using Pre-LN or Post-LN during supervised fine-tuning (SFT) and reinforcement learning from human feedback (RLHF), highlighting the critical importance of high-quality deep layers. By effectively addressing the inefficiencies of deep layers in current LLMs, Mix-LN unlocks their potential, enhancing model capacity without increasing model size. Our code is available at https://github.com/pixeli99/MixLN.
Authors: Yinbin Han, Meisam Razaviyayn, Renyuan Xu
Abstract: Diffusion models have emerged as powerful tools for generative modeling, demonstrating exceptional capability in capturing target data distributions from large datasets. However, fine-tuning these massive models for specific downstream tasks, constraints, and human preferences remains a critical challenge. While recent advances have leveraged reinforcement learning algorithms to tackle this problem, much of the progress has been empirical, with limited theoretical understanding. To bridge this gap, we propose a stochastic control framework for fine-tuning diffusion models. Building on denoising diffusion probabilistic models as the pre-trained reference dynamics, our approach integrates linear dynamics control with Kullback-Leibler regularization. We establish the well-posedness and regularity of the stochastic control problem and develop a policy iteration algorithm (PI-FT) for numerical solution. We show that PI-FT achieves global convergence at a linear rate. Unlike existing work that assumes regularities throughout training, we prove that the control and value sequences generated by the algorithm maintain the regularity. Additionally, we explore extensions of our framework to parametric settings and continuous-time formulations.
Authors: Niklas Pfister, V\'aclav Volhejn, Manuel Knott, Santiago Arias, Julia Bazi\'nska, Mykhailo Bichurin, Alan Commike, Janet Darling, Peter Dienes, Matthew Fiedler, David Haber, Matthias Kraft, Marco Lancini, Max Mathys, Dami\'an Pascual-Ortiz, Jakub Podolak, Adri\`a Romero-L\'opez, Kyriacos Shiarlis, Andreas Signer, Zsolt Terek, Athanasios Theocharis, Daniel Timbrell, Samuel Trautwein, Samuel Watts, Yun-Han Wu, Mateo Rojas-Carulla
Abstract: Current evaluations of defenses against prompt attacks in large language model (LLM) applications often overlook two critical factors: the dynamic nature of adversarial behavior and the usability penalties imposed on legitimate users by restrictive defenses. We propose D-SEC (Dynamic Security Utility Threat Model), which explicitly separates attackers from legitimate users, models multi-step interactions, and expresses the security-utility in an optimizable form. We further address the shortcomings in existing evaluations by introducing Gandalf, a crowd-sourced, gamified red-teaming platform designed to generate realistic, adaptive attack. Using Gandalf, we collect and release a dataset of 279k prompt attacks. Complemented by benign user data, our analysis reveals the interplay between security and utility, showing that defenses integrated in the LLM (e.g., system prompts) can degrade usability even without blocking requests. We demonstrate that restricted application domains, defense-in-depth, and adaptive defenses are effective strategies for building secure and useful LLM applications.
Authors: Amogh Raj, Carol Eunice Gudumotou, Sakol Bun, Keerthana Srinivasa, Arash Sarshar
Abstract: We present a novel framework combining Deep Operator Networks (DeepONets) with Physics-Informed Neural Networks (PINNs) to solve partial differential equations (PDEs) and estimate their unknown parameters. By integrating data-driven learning with physical constraints, our method achieves robust and accurate solutions across diverse scenarios. Bayesian training is implemented through variational inference, allowing for comprehensive uncertainty quantification for both aleatoric and epistemic uncertainties. This ensures reliable predictions and parameter estimates even in noisy conditions or when some of the physical equations governing the problem are missing. The framework demonstrates its efficacy in solving forward and inverse problems, including the 1D unsteady heat equation and 2D reaction-diffusion equations, as well as regression tasks with sparse, noisy observations. This approach provides a computationally efficient and generalizable method for addressing uncertainty quantification in PDE surrogate modeling.
Authors: Sota Mashiko, Yuji Kawamata, Tomoru Nakayama, Tetsuya Sakurai, Yukihiko Okada
Abstract: Anomaly detection is crucial in financial auditing, and effective detection requires large volumes of data from multiple organizations. However, journal entry data is highly sensitive, making it infeasible to share them directly across audit firms. To address this challenge, journal entry anomaly detection methods based on model share-type federated learning (FL) have been proposed. These methods require multiple rounds of communication with external servers to exchange model parameters, which necessitates connecting devices storing confidential data to external networks -- a practice not recommended for sensitive data such as journal entries. To overcome these limitations, a novel anomaly detection framework based on data collaboration (DC) analysis, a non-model share-type FL approach, is proposed. The method first transforms raw journal entry data into secure intermediate representations via dimensionality reduction and then constructs a collaboration representation used to train an anomaly detection autoencoder. Notably, the approach does not require raw data to be exposed or devices to be connected to external networks, and the entire process needs only a single round of communication. The proposed method was evaluated on both synthetic and real-world journal entry data collected from eight healthcare organizations. The experimental results demonstrated that the framework not only outperforms the baseline trained on individual data but also achieves higher detection performance than model-sharing FL methods such as FedAvg and FedProx, particularly under non-i.i.d. settings that simulate practical audit environments. This study addresses the critical need to integrate organizational knowledge while preserving data confidentiality, contributing to the development of practical intelligent auditing systems.
Authors: Bernardo \'Avila Pires, Mark Rowland, Diana Borsa, Zhaohan Daniel Guo, Khimya Khetarpal, Andr\'e Barreto, David Abel, R\'emi Munos, Will Dabney
Abstract: We introduce distributional dynamic programming (DP) methods for optimizing statistical functionals of the return distribution, with standard reinforcement learning as a special case. Previous distributional DP methods could optimize the same class of expected utilities as classic DP. To go beyond, we combine distributional DP with stock augmentation, a technique previously introduced for classic DP in the context of risk-sensitive RL, where the MDP state is augmented with a statistic of the rewards obtained since the first time step. We find that a number of recently studied problems can be formulated as stock-augmented return distribution optimization, and we show that we can use distributional DP to solve them. We analyze distributional value and policy iteration, with bounds and a study of what objectives these distributional DP methods can or cannot optimize. We describe a number of applications outlining how to use distributional DP to solve different stock-augmented return distribution optimization problems, for example maximizing conditional value-at-risk, and homeostatic regulation. To highlight the practical potential of stock-augmented return distribution optimization and distributional DP, we introduce an agent that combines DQN and the core ideas of distributional DP, and empirically evaluate it for solving instances of the applications discussed.
Authors: Mitch Kosieradzki, Seongjin Choi
Abstract: For intelligent transportation systems and autonomous vehicles to operate safely and efficiently, they must reliably predict the future motion and trajectory of surrounding agents within complex traffic environments. At the same time, the motion of these agents is inherently uncertain, making accurate prediction difficult. In this paper, we propose \textbf{TrajFlow}, a generative framework for estimating the occupancy density of dynamic agents. Our framework utilizes a causal encoder to extract semantically meaningful embeddings of the observed trajectory, as well as a normalizing flow to decode these embeddings and determine the most likely future location of an agent at some time point in the future. Our formulation differs from existing approaches because we model the marginal distribution of spatial locations instead of the joint distribution of unobserved trajectories. The advantages of a marginal formulation are numerous. First, we demonstrate that the marginal formulation produces higher accuracy on challenging trajectory forecasting benchmarks. Second, the marginal formulation allows for fully continuous sampling of future locations. Finally, marginal densities are better suited for downstream tasks as they allow for the computation of per-agent motion trajectories and occupancy grids, the two most commonly used representations for motion forecasting. We present a novel architecture based entirely on neural differential equations as an implementation of this framework and provide ablations to demonstrate the advantages of a continuous implementation over a more traditional discrete neural network based approach. The code is available at https://github.com/UMN-Choi-Lab/TrajFlow.
Authors: Christian Wald, Gabriele Steidl
Abstract: Among generative neural models, flow matching techniques stand out for their simple applicability and good scaling properties. Here, velocity fields of curves connecting a simple latent and a target distribution are learned. Then the corresponding ordinary differential equation can be used to sample from a target distribution, starting in samples from the latent one. This paper reviews from a mathematical point of view different techniques to learn the velocity fields of absolutely continuous curves in the Wasserstein geometry. We show how the velocity fields can be characterized and learned via i) transport plans (couplings) between latent and target distributions, ii) Markov kernels and iii) stochastic processes, where the latter two include the coupling approach, but are in general broader. Besides this main goal, we show how flow matching can be used for solving Bayesian inverse problems, where the definition of conditional Wasserstein distances plays a central role. Finally, we briefly address continuous normalizing flows and score matching techniques, which approach the learning of velocity fields of curves from other directions.
Authors: Zhixuan Li, Naipeng Chen, Seonghwa Choi, Sanghoon Lee, Weisi Lin
Abstract: Time-series forecasting is crucial for numerous real-world applications including weather prediction and financial market modeling. While temporal-domain methods remain prevalent, frequency-domain approaches can effectively capture multi-scale periodic patterns, reduce sequence dependencies, and naturally denoise signals. However, existing approaches typically train model components for all frequencies under a unified training objective, often leading to mismatched learning speeds: high-frequency components converge faster and risk overfitting, while low-frequency components underfit due to insufficient training time. To deal with this challenge, we propose BEAT (Balanced frEquency Adaptive Tuning), a novel framework that dynamically monitors the training status for each frequency and adaptively adjusts their gradient updates. By recognizing convergence, overfitting, or underfitting for each frequency, BEAT dynamically reallocates learning priorities, moderating gradients for rapid learners and increasing those for slower ones, alleviating the tension between competing objectives across frequencies and synchronizing the overall learning process. Extensive experiments on seven real-world datasets demonstrate that BEAT consistently outperforms state-of-the-art approaches.
Authors: Wenyun Li, Wenjie Huang, Chen Sun
Abstract: In many real-world scenarios, reward signal for agents are exceedingly sparse, making it challenging to learn an effective reward function for reward shaping. To address this issue, the proposed approach in this paper performs reward shaping not only by utilizing non-zero-reward transitions but also by employing the \emph{Semi-Supervised Learning} (SSL) technique combined with a novel data augmentation to learn trajectory space representations from the majority of transitions, {i.e}., zero-reward transitions, thereby improving the efficacy of reward shaping. Experimental results in Atari and robotic manipulation demonstrate that our method outperforms supervised-based approaches in reward inference, leading to higher agent scores. Notably, in more sparse-reward environments, our method achieves up to twice the peak scores compared to supervised baselines. The proposed double entropy data augmentation enhances performance, showcasing a 15.8\% increase in best score over other augmentation methods
Authors: Jan Schuchardt, Mina Dalirrooyfard, Jed Guzelkabaagac, Anderson Schneider, Yuriy Nevmyvaka, Stephan G\"unnemann
Abstract: Many forms of sensitive data, such as web traffic, mobility data, or hospital occupancy, are inherently sequential. The standard method for training machine learning models while ensuring privacy for units of sensitive information, such as individual hospital visits, is differentially private stochastic gradient descent (DP-SGD). However, we observe in this work that the formal guarantees of DP-SGD are incompatible with time series specific tasks like forecasting, since they rely on the privacy amplification attained by training on small, unstructured batches sampled from an unstructured dataset. In contrast, batches for forecasting are generated by (1) sampling sequentially structured time series from a dataset, (2) sampling contiguous subsequences from these series, and (3) partitioning them into context and ground-truth forecast windows. We theoretically analyze the privacy amplification attained by this structured subsampling to enable the training of forecasting models with sound and tight event- and user-level privacy guarantees. Towards more private models, we additionally prove how data augmentation amplifies privacy in self-supervised training of sequence models. Our empirical evaluation demonstrates that amplification by structured subsampling enables the training of forecasting models with strong formal privacy guarantees.
Authors: Yik Siu Chan, Narutatsu Ri, Yuxin Xiao, Marzyeh Ghassemi
Abstract: Despite extensive safety alignment efforts, large language models (LLMs) remain vulnerable to jailbreak attacks that elicit harmful behavior. While existing studies predominantly focus on attack methods that require technical expertise, two critical questions remain underexplored: (1) Are jailbroken responses truly useful in enabling average users to carry out harmful actions? (2) Do safety vulnerabilities exist in more common, simple human-LLM interactions? In this paper, we demonstrate that LLM responses most effectively facilitate harmful actions when they are both actionable and informative--two attributes easily elicited in multi-step, multilingual interactions. Using this insight, we propose HarmScore, a jailbreak metric that measures how effectively an LLM response enables harmful actions, and Speak Easy, a simple multi-step, multilingual attack framework. Notably, by incorporating Speak Easy into direct request and jailbreak baselines, we see an average absolute increase of 0.319 in Attack Success Rate and 0.426 in HarmScore in both open-source and proprietary LLMs across four safety benchmarks. Our work reveals a critical yet often overlooked vulnerability: Malicious users can easily exploit common interaction patterns for harmful intentions.
Authors: Keke Gai, Mohan Wang, Jing Yu, Dongjue Wang, Qi Wu
Abstract: Multimodal Federated Learning (MFL) with mixed modalities enables unimodal and multimodal clients to collaboratively train models while ensuring clients' privacy. As a representative sample of local data, prototypes offer an approach with low resource consumption and no reliance on prior knowledge for MFL with mixed modalities. However, existing prototype-based MFL methods assume unified labels across clients and identical tasks per client, which is impractical in MFL with mixed modalities. In this work, we propose an Adaptive prototype-based Multimodal Federated Learning (AproMFL) framework for mixed modalities to address the aforementioned issues. Our AproMFL transfers knowledge through adaptively-constructed prototypes without unified labels. Clients adaptively select prototype construction methods in line with labels; server converts client prototypes into unified multimodal prototypes and cluster them to form global prototypes. To address model aggregation issues in task heterogeneity, we develop a client relationship graph-based scheme to dynamically adjust aggregation weights. Furthermore, we propose a global prototype knowledge transfer loss and a global model knowledge transfer loss to enable the transfer of global knowledge to local knowledge. Experimental results show that AproMFL outperforms four baselines on three highly heterogeneous datasets ($\alpha=0.1$) and two heterogeneous tasks, with the optimal results in accuracy and recall being 0.42%~6.09% and 1.6%~3.89% higher than those of FedIoT (FedAvg-based MFL), respectively.
Authors: Nikos Zarifis, Puqian Wang, Ilias Diakonikolas, Jelena Diakonikolas
Abstract: We study the task of learning Generalized Linear models (GLMs) in the agnostic model under the Gaussian distribution. We give the first polynomial-time algorithm that achieves a constant-factor approximation for \textit{any} monotone Lipschitz activation. Prior constant-factor GLM learners succeed for a substantially smaller class of activations. Our work resolves a well-known open problem, by developing a robust counterpart to the classical GLMtron algorithm (Kakade et al., 2011). Our robust learner applies more generally, encompassing all monotone activations with bounded $(2+\zeta)$-moments, for any fixed $\zeta>0$ -- a condition that is essentially necessary. To obtain our results, we leverage a novel data augmentation technique with decreasing Gaussian noise injection and prove a number of structural results that may be useful in other settings.
Authors: Theodoros Kouzelis, Ioannis Kakogeorgiou, Spyros Gidaris, Nikos Komodakis
Abstract: Latent generative models have emerged as a leading approach for high-quality image synthesis. These models rely on an autoencoder to compress images into a latent space, followed by a generative model to learn the latent distribution. We identify that existing autoencoders lack equivariance to semantic-preserving transformations like scaling and rotation, resulting in complex latent spaces that hinder generative performance. To address this, we propose EQ-VAE, a simple regularization approach that enforces equivariance in the latent space, reducing its complexity without degrading reconstruction quality. By finetuning pre-trained autoencoders with EQ-VAE, we enhance the performance of several state-of-the-art generative models, including DiT, SiT, REPA and MaskGIT, achieving a 7 speedup on DiT-XL/2 with only five epochs of SD-VAE fine-tuning. EQ-VAE is compatible with both continuous and discrete autoencoders, thus offering a versatile enhancement for a wide range of latent generative models. Project page and code: https://eq-vae.github.io/.
Authors: Xu Shen, Yixin Liu, Yili Wang, Rui Miao, Yiwei Dai, Shirui Pan, Yi Chang, Xin Wang
Abstract: Out-of-distribution (OOD) generalization has emerged as a critical challenge in graph learning, as real-world graph data often exhibit diverse and shifting environments that traditional models fail to generalize across. A promising solution to address this issue is graph invariant learning (GIL), which aims to learn invariant representations by disentangling label-correlated invariant subgraphs from environment-specific subgraphs. However, existing GIL methods face two major challenges: (1) the difficulty of capturing and modeling diverse environments in graph data, and (2) the semantic cliff, where invariant subgraphs from different classes are difficult to distinguish, leading to poor class separability and increased misclassifications. To tackle these challenges, we propose a novel method termed Multi-Prototype Hyperspherical Invariant Learning (MPHIL), which introduces two key innovations: (1) hyperspherical invariant representation extraction, enabling robust and highly discriminative hyperspherical invariant feature extraction, and (2) multi-prototype hyperspherical classification, which employs class prototypes as intermediate variables to eliminate the need for explicit environment modeling in GIL and mitigate the semantic cliff issue. Derived from the theoretical framework of GIL, we introduce two novel objective functions: the invariant prototype matching loss to ensure samples are matched to the correct class prototypes, and the prototype separation loss to increase the distinction between prototypes of different classes in the hyperspherical space. Extensive experiments on 11 OOD generalization benchmark datasets demonstrate that MPHIL achieves state-of-the-art performance, significantly outperforming existing methods across graph data from various domains and with different distribution shifts.
Authors: Soumyajit Guin, Vivek S. Borkar, Shalabh Bhatnagar
Abstract: In this paper, we consider the risk-sensitive cost criterion with exponentiated costs for Markov decision processes and develop a model-free policy gradient algorithm in this setting. Unlike additive cost criteria such as average or discounted cost, the risk-sensitive cost criterion is less studied due to the complexity resulting from the multiplicative structure of the resulting Bellman equation. We develop an actor-critic algorithm with function approximation in this setting and provide its asymptotic convergence analysis. We also show the results of numerical experiments that demonstrate the superiority in performance of our algorithm over other recent algorithms in the literature.
Authors: Saman Khamesian, Asiful Arefeen, Maria Adela Grando, Bithika M. Thompson, Hassan Ghasemzadeh
Abstract: Managing Type 1 Diabetes (T1D) demands constant vigilance as individuals strive to regulate their blood glucose levels to avoid the harmful effects of dysglycemia, including both hyperglycemia and hypoglycemia. Despite the development of advanced technologies such as automated insulin delivery (AID) systems, achieving optimal glycemic control remains challenging. AID systems combine continuous subcutaneous insulin infusion with data from continuous glucose monitors (CGMs), offering potential benefits in reducing glucose variability and increasing time-in-range. However, these systems still frequently fail to prevent dysglycemia, partly due to limitations in their prediction algorithms, which lack the accuracy needed to avert abnormal glucose events. This shortcoming highlights the need for more advanced glucose forecasting methods. To address this need, we introduce GLIMMER, Glucose Level Indicator Model with Modified Error Rate, a machine learning-based model for predicting blood glucose levels. GLIMMER classifies glucose values into normal and abnormal ranges and employs a novel custom loss function that prioritizes accuracy in dysglycemic regions, where patient safety is most critical. To evaluate GLIMMER's effectiveness for T1D management, we used both a publicly available dataset and a newly collected dataset involving 25 individuals with T1D. In forecasting glucose levels for the next hour, GLIMMER achieved a root mean square error (RMSE) of 23.97 (+/-3.77) and a mean absolute error (MAE) of 15.83 (+/-2.09) mg/dL. These results represent a 23% improvement in RMSE and a 31% improvement in MAE compared to the best previously reported models.
Authors: Qiuhai Zeng, Claire Jin, Xinyue Wang, Yuhan Zheng, Qunhua Li
Abstract: Large language models (LLMs) are increasingly used to automate data analysis through executable code generation. Yet, data science tasks often admit multiple statistically valid solutions, e.g. different modeling strategies, making it critical to understand the reasoning behind analyses, not just their outcomes. While manual review of LLM-generated code can help ensure statistical soundness, it is labor-intensive and requires expertise. A more scalable approach is to evaluate the underlying workflows - the logical plans guiding code generation. However, it remains unclear how to assess whether a LLM-generated workflow supports reproducible implementations. To address this, we present $\it{AIRepr}$, an $\it{A}$nalyst - $\it{I}$nspector framework for automatically evaluating and improving the $\it{Repr}$oducibility of LLM-generated data analysis workflows. Our framework is grounded in statistical principles and supports scalable, automated assessment. We introduce two novel reproducibility-enhancing prompting strategies and benchmark them against standard prompting across 15 analyst-inspector LLM pairs and 1,032 tasks from three public benchmarks. Our findings show that workflows with higher reproducibility also yield more accurate analyses, and that reproducibility-enhancing prompts substantially improve both metrics. This work provides a foundation for more transparent, reliable, and efficient human-AI collaboration in data science. Our code is publicly available.
Authors: Weilin Chen, Ruichu Cai, Jie Qiao, Yuguang Yan, Jos\'e Miguel Hern\'andez-Lobato
Abstract: Estimating causal effects under networked interference from observational data is a crucial yet challenging problem. Most existing methods mainly rely on the networked unconfoundedness assumption, which guarantees the identification of networked effects. However, this assumption is often violated due to the latent confounders inherent in observational data, thereby hindering the identification of networked effects. To address this issue, we leverage the rich interaction patterns between units in networks, which provide valuable information for recovering these latent confounders. Building on this insight, we develop a confounder recovery framework that explicitly characterizes three categories of latent confounders in networked settings: those affecting only the unit, those affecting only the unit's neighbors, and those influencing both. Based on this framework, we design a networked effect estimator using identifiable representation learning techniques. From a theoretical standpoint, we prove the identifiability of all three types of latent confounders and, by leveraging the recovered confounders, establish a formal identification result for networked effects. Extensive experiments validate our theoretical findings and demonstrate the effectiveness of the proposed method.
Authors: Laziz Abdullaev, Tan M. Nguyen
Abstract: Transformer-based deep learning models have achieved state-of-the-art performance across numerous language and vision tasks. While the self-attention mechanism, a core component of transformers, has proven capable of handling complex data patterns, it has been observed that the representational capacity of the attention matrix degrades significantly across transformer layers, thereby hurting its overall performance. In this work, we leverage the connection between self-attention computations and low-pass non-local means (NLM) smoothing filters and propose the Twicing Attention, a novel attention mechanism that uses kernel twicing procedure in nonparametric regression to alleviate the low-pass behavior of associated NLM smoothing with compelling theoretical guarantees and enhanced adversarial robustness. This approach enables the extraction and reuse of meaningful information retained in the residuals following the imperfect smoothing operation at each layer. Our proposed method offers two key advantages over standard self-attention: 1) a provably slower decay of representational capacity and 2) improved robustness and accuracy across various data modalities and tasks. We empirically demonstrate the performance gains of our model over baseline transformers on multiple tasks and benchmarks, including image classification and language modeling, on both clean and corrupted data.
Authors: Kasra Arabi, R. Teal Witter, Chinmay Hegde, Niv Cohen
Abstract: Generative models have rapidly evolved to generate realistic outputs. However, their synthetic outputs increasingly challenge the clear distinction between natural and AI-generated content, necessitating robust watermarking techniques. Watermarks are typically expected to preserve the integrity of the target image, withstand removal attempts, and prevent unauthorized replication onto unrelated images. To address this need, recent methods embed persistent watermarks into images produced by diffusion models using the initial noise. Yet, to do so, they either distort the distribution of generated images or rely on searching through a long dictionary of used keys for detection. In this paper, we propose a novel watermarking method that embeds semantic information about the generated image directly into the watermark, enabling a distortion-free watermark that can be verified without requiring a database of key patterns. Instead, the key pattern can be inferred from the semantic embedding of the image using locality-sensitive hashing. Furthermore, conditioning the watermark detection on the original image content improves robustness against forgery attacks. To demonstrate that, we consider two largely overlooked attack strategies: (i) an attacker extracting the initial noise and generating a novel image with the same pattern; (ii) an attacker inserting an unrelated (potentially harmful) object into a watermarked image, possibly while preserving the watermark. We empirically validate our method's increased robustness to these attacks. Taken together, our results suggest that content-aware watermarks can mitigate risks arising from image-generative models.
Authors: Olawale Salaudeen, Nicole Chiou, Shiny Weng, Sanmi Koyejo
Abstract: Spurious correlations, unstable statistical shortcuts a model can exploit, are expected to degrade performance out-of-distribution (OOD). However, across many popular OOD generalization benchmarks, vanilla empirical risk minimization (ERM) often achieves the highest OOD accuracy. Moreover, gains in in-distribution accuracy generally improve OOD accuracy, a phenomenon termed accuracy on the line, which contradicts the expected harm of spurious correlations. We show that these observations are an artifact of misspecified OOD datasets that do not include shifts in spurious correlations that harm OOD generalization, the setting they are meant to evaluate. Consequently, current practice evaluates "robustness" without truly stressing the spurious signals we seek to eliminate; our work pinpoints when that happens and how to fix it. Contributions. (i) We derive necessary and sufficient conditions for a distribution shift to reveal a model's reliance on spurious features; when these conditions hold, "accuracy on the line" disappears. (ii) We audit leading OOD datasets and find that most still display accuracy on the line, suggesting they are misspecified for evaluating robustness to spurious correlations. (iii) We catalog the few well-specified datasets and summarize generalizable design principles, such as identifying datasets of natural interventions (e.g., a pandemic), to guide future well-specified benchmarks.
Authors: Minwei Zhao, Sanja Scepanovic, Stephen Law, Ivica Obadic, Cai Wu, Daniele Quercia
Abstract: Understanding how urban socio-demographic and environmental factors relate with health is essential for public health and urban planning. However, traditional statistical methods struggle with nonlinear effects, while machine learning models often fail to capture geographical (nearby areas being more similar) and topological (unequal connectivity between places) effects in an interpretable way. To address this, we propose MedGNN, a spatio-topologically explicit framework that constructs a 2-hop spatial graph, integrating positional and locational node embeddings with urban characteristics in a graph neural network. Applied to MEDSAT, a comprehensive dataset covering over 150 environmental and socio-demographic factors and six prescription outcomes (depression, anxiety, diabetes, hypertension, asthma, and opioids) across 4,835 Greater London neighborhoods, MedGNN improved predictions by over 25% on average compared to baseline methods. Using depression prescriptions as a case study, we analyzed graph embeddings via geographical principal component analysis, identifying findings that: align with prior research (e.g., higher antidepressant prescriptions among older and White populations), contribute to ongoing debates (e.g., greenery linked to higher and NO2 to lower prescriptions), and warrant further study (e.g., canopy evaporation correlated with fewer prescriptions). These results demonstrate MedGNN's potential, and more broadly, of carefully applied machine learning, to advance transdisciplinary public health research.
Authors: Juzheng Zhang, Jiacheng You, Ashwinee Panda, Tom Goldstein
Abstract: Low-Rank Adaptation (LoRA) has emerged as a popular parameter-efficient fine-tuning (PEFT) method for Large Language Models (LLMs), yet it still incurs notable overhead and suffers from parameter interference in multi-task scenarios. We propose LoRA with Reduced Interference (LoRI), a simple yet effective approach that freezes the projection matrices $A$ as random projections and sparsifies the matrices $B$ using task-specific masks. This design substantially reduces the number of trainable parameters while maintaining strong task performance. Moreover, LoRI minimizes cross-task interference in adapter merging by leveraging the orthogonality between adapter subspaces, and supports continual learning by using sparsity to mitigate catastrophic forgetting. Extensive experiments across natural language understanding, mathematical reasoning, code generation, and safety alignment tasks demonstrate that LoRI outperforms full fine-tuning and existing PEFT methods, while using up to 95% fewer trainable parameters than LoRA. In multi-task experiments, LoRI enables effective adapter merging and continual learning with reduced cross-task interference. Code is available at: https://github.com/juzhengz/LoRI
Authors: Hsing-Huan Chung, Shravan Chaudhari, Xing Han, Yoav Wald, Suchi Saria, Joydeep Ghosh
Abstract: Dynamic graph learning is essential for applications involving temporal networks and requires effective modeling of temporal relationships. Seminal attention-based models like TGAT and DyGFormer rely on sinusoidal time encoders to capture temporal dependencies between edge events. Prior work justified sinusoidal encodings because their inner products depend on the time spans between events, which are crucial features for modeling inter-event relations. However, sinusoidal encodings inherently lose temporal information due to their many-to-one nature and therefore require high dimensions. In this paper, we rigorously study a simpler alternative: the linear time encoder, which avoids temporal information loss caused by sinusoidal functions and reduces the need for high-dimensional time encoders. We show that the self-attention mechanism can effectively learn to compute time spans between events from linear time encodings and extract relevant temporal patterns. Through extensive experiments on six dynamic graph datasets, we demonstrate that the linear time encoder improves the performance of TGAT and DyGFormer in most cases. Moreover, the linear time encoder can lead to significant savings in model parameters with minimal performance loss. For example, compared to a 100-dimensional sinusoidal time encoder, TGAT with a 2-dimensional linear time encoder saves 43% of parameters and achieves higher average precision on five datasets. While both encoders can be used simultaneously, our study highlights the often-overlooked advantages of linear time features in modern dynamic graph models. These findings can positively impact the design choices of various dynamic graph learning architectures and eventually benefit temporal network applications such as recommender systems, communication networks, and traffic forecasting.
Authors: Moritz Piening, Robert Beinert
Abstract: Gaussian mixture models (GMMs) are widely used in machine learning for tasks such as clustering, classification, image reconstruction, and generative modeling. A key challenge in working with GMMs is defining a computationally efficient and geometrically meaningful metric. The mixture Wasserstein (MW) distance adapts the Wasserstein metric to GMMs and has been applied in various domains, including domain adaptation, dataset comparison, and reinforcement learning. However, its high computational cost -- arising from repeated Wasserstein distance computations involving matrix square root estimations and an expensive linear program -- limits its scalability to high-dimensional and large-scale problems. To address this, we propose multiple novel slicing-based approximations to the MW distance that significantly reduce computational complexity while preserving key optimal transport properties. From a theoretical viewpoint, we establish several weak and strong equivalences between the introduced metrics, and show the relations to the original MW distance and the well-established sliced Wasserstein distance. Furthermore, we validate the effectiveness of our approach through numerical experiments, demonstrating computational efficiency and applications in clustering, perceptual image comparison, and GMM minimization
Authors: Haoran Gu, Handing Wang, Yi Mei, Mengjie Zhang, Yaochu Jin
Abstract: Aligning large language models with multiple human expectations and values is crucial for ensuring that they adequately serve a variety of user needs. To this end, offline multiobjective alignment algorithms such as the Rewards-in-Context algorithm have shown strong performance and efficiency. However, inappropriate preference representations and training with imbalanced reward scores limit the performance of such algorithms. In this work, we introduce ParetoHqD that addresses the above issues by representing human preferences as preference directions in the objective space and regarding data near the Pareto front as ''high-quality'' data. For each preference, ParetoHqD follows a two-stage supervised fine-tuning process, where each stage uses an individual Pareto high-quality training set that best matches its preference direction. The experimental results have demonstrated the superiority of ParetoHqD over five baselines on two multiobjective alignment tasks.
Authors: Mateo Espinosa Zarlenga, Gabriele Dominici, Pietro Barbiero, Zohreh Shams, Mateja Jamnik
Abstract: In this paper, we investigate how concept-based models (CMs) respond to out-of-distribution (OOD) inputs. CMs are interpretable neural architectures that first predict a set of high-level concepts (e.g., stripes, black) and then predict a task label from those concepts. In particular, we study the impact of concept interventions (i.e., operations where a human expert corrects a CM's mispredicted concepts at test time) on CMs' task predictions when inputs are OOD. Our analysis reveals a weakness in current state-of-the-art CMs, which we term leakage poisoning, that prevents them from properly improving their accuracy when intervened on for OOD inputs. To address this, we introduce MixCEM, a new CM that learns to dynamically exploit leaked information missing from its concepts only when this information is in-distribution. Our results across tasks with and without complete sets of concept annotations demonstrate that MixCEMs outperform strong baselines by significantly improving their accuracy for both in-distribution and OOD samples in the presence and absence of concept interventions.
Authors: Samantha Petti, Carlos Mart\'i-G\'omez, Justin B. Kinney, Juannan Zhou, David M. McCandlish
Abstract: Mappings from biological sequences (DNA, RNA, protein) to quantitative measures of sequence functionality play an important role in contemporary biology. We are interested in the related tasks of (i) inferring predictive sequence-to-function maps and (ii) decomposing sequence-function maps to elucidate the contributions of individual subsequences. Because each sequence-function map can be written as a weighted sum over subsequences in multiple ways, meaningfully interpreting these weights requires ``gauge-fixing,'' i.e., defining a unique representation for each map. Recent work has established that most existing gauge-fixed representations arise as the unique solutions to $L_2$-regularized regression in an overparameterized ``weight space'' where the choice of regularizer defines the gauge. Here, we establish the relationship between regularized regression in overparameterized weight space and Gaussian process approaches that operate in ``function space,'' i.e.~the space of all real-valued functions on a finite set of sequences. We disentangle how weight space regularizers both impose an implicit prior on the learned function and restrict the optimal weights to a particular gauge. We show how to construct regularizers that correspond to arbitrary explicit Gaussian process priors combined with a wide variety of gauges and characterize the implicit function space priors associated with the most common weight space regularizers. Finally, we derive the posterior distribution of a broad class of sequence-to-function statistics, including gauge-fixed weights and multiple systems for expressing higher-order epistatic coefficients. We show that such distributions can be efficiently computed for product-kernel priors using a kernel trick.
Authors: Antonio A. Ginart, Naveen Kodali, Jason Lee, Caiming Xiong, Silvio Savarese, John R. Emmons
Abstract: We introduce the LZ penalty, a penalty specialized for reducing degenerate repetitions in autoregressive language models without loss of capability. The penalty is based on the codelengths in the LZ77 universal lossless compression algorithm. Through the lens of the prediction-compression duality, decoding the LZ penalty has the interpretation of sampling from the residual distribution after removing the information that is highly compressible. We demonstrate the LZ penalty enables state-of-the-art open-source reasoning models to operate with greedy (temperature zero) decoding without loss of capability and without instances of degenerate repetition. Both the industry-standard frequency penalty and repetition penalty are ineffective, incurring degenerate repetition rates of up to 4%.
Authors: Shunxian Gu, Chaoqun You, Bangbang Ren, Lailong Luo, Junxu Xia, Deke Guo
Abstract: Scaling deep neural network (DNN) training to more devices can reduce time-to-solution. However, it is impractical for users with limited computing resources. FOSI, as a hybrid order optimizer, converges faster than conventional optimizers by taking advantage of both gradient information and curvature information when updating the DNN model. Therefore, it provides a new chance for accelerating DNN training in the resource-constrained setting. In this paper, we explore its distributed design, namely DHO$_2$, including distributed calculation of curvature information and model update with partial curvature information to accelerate DNN training with a low memory burden. To further reduce the training time, we design a novel strategy to parallelize the calculation of curvature information and the model update on different devices. Experimentally, our distributed design can achieve an approximate linear reduction of memory burden on each device with the increase of the device number. Meanwhile, it achieves $1.4\times\sim2.1\times$ speedup in the total training time compared with other distributed designs based on conventional first- and second-order optimizers.
Authors: Junyu Xue, Xudong Wang, Xiaoling He, Shicheng Liu, Yi Wang, Guoming Tang
Abstract: Non-intrusive load monitoring (NILM) aims to disaggregate total electricity consumption into individual appliance usage, thus enabling more effective energy management. While deep learning has advanced NILM, it remains limited by its dependence on labeled data, restricted generalization, and lack of explainability. This paper introduces the first prompt-based NILM framework that leverages large language models (LLMs) with in-context learning. We design and evaluate prompt strategies that integrate appliance features, contextual information, and representative time-series examples through extensive case studies. Extensive experiments on the REDD and UK-DALE datasets show that LLMs guided solely by prompts deliver only basic NILM capabilities, with performance that lags behind traditional deep-learning models in complex scenarios. However, the experiments also demonstrate strong generalization across different houses and even regions by simply adapting the injected appliance features. It also provides clear, human-readable explanations for the inferred appliance states. Our findings define the capability boundaries of using prompt-only LLMs for NILM tasks. Their strengths in generalization and explainability present a promising new direction for the field.
Authors: Junfeng Jiao, Seung Gyu Baik, Seung Jun Choi, Yiming Xu
Abstract: This paper quantitatively investigates the crash severity of Autonomous Vehicles (AVs) with spatially localized machine learning and macroscopic measures of the urban built environment. Extending beyond the microscopic effects of individual infrastructure elements, we focus on the city-scale land use and behavioral patterns, while addressing spatial heterogeneity and spatial autocorrelation. We implemented a spatially localized machine learning technique called Geographical Random Forest (GRF) on the California AV collision dataset. Analyzing multiple urban measures, including points of interest, building footprint, and land use, we built a GRF model and visualized it as a crash severity risk map of San Francisco. This paper presents three findings. First, spatially localized machine learning outperformed regular machine learning in predicting AV crash severity. The bias-variance tradeoff was evident as we adjusted the localization weight hyperparameter. Second, land use was the most important predictor, compared to intersections, building footprints, public transit stops, and Points Of Interest (POIs). Third, AV crashes were more likely to result in low-severity incidents in city center areas with greater diversity and commercial activities, than in residential neighborhoods. Residential land use is likely associated with higher severity due to human behavior and less restrictive environments. Counterintuitively, residential areas were associated with higher crash severity, compared to more complex areas such as commercial and mixed-use areas. When robotaxi operators train their AV systems, it is recommended to: (1) consider where their fleet operates and make localized algorithms for their perception system, and (2) design safety measures specific to residential neighborhoods, such as slower driving speeds and more alert sensors.
Authors: Mohammadmahdi Ghasemloo, Alireza Moradi
Abstract: With the widespread adoption of Large Language Models (LLMs), there is a growing need to establish best practices for leveraging their capabilities beyond traditional natural language tasks. In this paper, a novel cross-domain knowledge transfer framework is proposed to enhance the performance of LLMs in time series forecasting -- a task of increasing relevance in fields such as energy systems, finance, and healthcare. The approach systematically infuses LLMs with structured temporal information to improve their forecasting accuracy. This study evaluates the proposed method on a real-world time series dataset and compares it to a naive baseline where the LLM receives no auxiliary information. Results show that knowledge-informed forecasting significantly outperforms the uninformed baseline in terms of predictive accuracy and generalization. These findings highlight the potential of knowledge transfer strategies to bridge the gap between LLMs and domain-specific forecasting tasks.
Authors: Zheng Wu, Pengzhou Cheng, Zongru Wu, Lingzhong Dong, Zhuosheng Zhang
Abstract: Graphical user interface (GUI) agents have recently emerged as an intriguing paradigm for human-computer interaction, capable of automatically executing user instructions to operate intelligent terminal devices. However, when encountering out-of-distribution (OOD) instructions that violate environmental constraints or exceed the current capabilities of agents, GUI agents may suffer task breakdowns or even pose security threats. Therefore, effective OOD detection for GUI agents is essential. Traditional OOD detection methods perform suboptimally in this domain due to the complex embedding space and evolving GUI environments. In this work, we observe that the in-distribution input semantic space of GUI agents exhibits a clustering pattern with respect to the distance from the centroid. Based on the finding, we propose GEM, a novel method based on fitting a Gaussian mixture model over input embedding distances extracted from the GUI agent that reflect its capability boundary. Evaluated on eight datasets spanning smartphones, computers, and web browsers, our method achieves an average accuracy improvement of 23.70\% over the best-performing baseline while only increasing training time by 4.9\% and testing time by 6.5\%. We also experimentally demonstrate that GEM can improve the step-wise success rate by 9.40\% by requesting assistance from the cloud model when encountering OOD samples. Analysis verifies the generalization ability of our method through experiments on nine different backbones. The codes are available at https://github.com/Wuzheng02/GEM-OODforGUIagents.
Authors: Jack Chen, Fazhong Liu, Naruto Liu, Yuhan Luo, Erqu Qin, Harry Zheng, Tian Dong, Haojin Zhu, Yan Meng, Xiao Wang
Abstract: Large language models (LLMs) excel at mathematical reasoning and logical problem-solving. The current popular training paradigms primarily use supervised fine-tuning (SFT) and reinforcement learning (RL) to enhance the models' reasoning abilities. However, when using SFT or RL alone, there are respective challenges: SFT may suffer from overfitting, while RL is prone to mode collapse. The state-of-the-art methods have proposed hybrid training schemes. However, static switching faces challenges such as poor generalization across different tasks and high dependence on data quality. In response to these challenges, inspired by the curriculum learning-quiz mechanism in human reasoning cultivation, We propose SASR, a step-wise adaptive hybrid training framework that theoretically unifies SFT and RL and dynamically balances the two throughout optimization. SASR uses SFT for initial warm-up to establish basic reasoning skills, and then uses an adaptive dynamic adjustment algorithm based on gradient norm and divergence relative to the original distribution to seamlessly integrate SFT with the online RL method GRPO. By monitoring the training status of LLMs and adjusting the training process in sequence, SASR ensures a smooth transition between training schemes, maintaining core reasoning abilities while exploring different paths. Experimental results demonstrate that SASR outperforms SFT, RL, and static hybrid training methods.
Authors: Qian Tan, Dongzhan Zhou, Peng Xia, Wanhao Liu, Wanli Ouyang, Lei Bai, Yuqiang Li, Tianfan Fu
Abstract: Multimodal large language models (MLLMs) have made impressive progress in many applications in recent years. However, chemical MLLMs that can handle cross-modal understanding and generation remain underexplored. To fill this gap, we propose ChemMLLM, a unified chemical multimodal large language model for molecule understanding and generation. Also, we design five multimodal tasks across text, molecular SMILES strings, and image, and curate the datasets. We benchmark ChemMLLM against a range of general leading MLLMs and Chemical LLMs on these tasks. Experimental results show that ChemMLLM achieves superior performance across all evaluated tasks. For example, in molecule image optimization task, ChemMLLM outperforms the best baseline (GPT-4o) by 116.75\% (4.27 vs 1.97 property improvement). The code is publicly available at https://github.com/bbsbz/ChemMLLM.git.
Authors: Takashi Ishida, Thanawat Lodkaew, Ikko Yamane
Abstract: Publishing a large language model (LLM) benchmark on the Internet risks contaminating future LLMs: the benchmark may be unintentionally (or intentionally) used to train or select a model. A common mitigation is to keep the benchmark private and let participants submit their models or predictions to the organizers. However, this strategy will require trust in a single organization and still permits test-set overfitting through repeated queries. To overcome this issue, we propose a way to publish benchmarks without completely disclosing the ground-truth answers to the questions, while still maintaining the ability to openly evaluate LLMs. Our main idea is to inject randomness to the answers by preparing several logically correct answers, and only include one of them as the solution in the benchmark. This reduces the best possible accuracy, i.e., Bayes accuracy, of the benchmark. Not only is this helpful to keep us from disclosing the ground truth, but this approach also offers a test for detecting data contamination. In principle, even fully capable models should not surpass the Bayes accuracy. If a model surpasses this ceiling despite this expectation, this is a strong signal of data contamination. We present experimental evidence that our method can detect data contamination accurately on a wide range of benchmarks, models, and training methodologies.
Authors: Quan Zhou, Mark Kozdoba, Shie Mannor
Abstract: We study the problem of selecting a subset from a large action space shared by a family of bandits, with the goal of achieving performance nearly matching that of using the full action space. We assume that similar actions tend to have related payoffs, modeled by a Gaussian process. To exploit this structure, we propose a simple epsilon-net algorithm to select a representative subset. We provide theoretical guarantees for its performance and compare it empirically to Thompson Sampling and Upper Confidence Bound.
Authors: Jimeng Shi, Sizhe Zhou, Bowen Jin, Wei Hu, Runchu Tian, Shaowen Wang, Giri Narasimhan, Jiawei Han
Abstract: Large language models (LLMs) often need to incorporate external knowledge to solve theme-specific problems. Retrieval-augmented generation (RAG) has shown its high promise, empowering LLMs to generate more qualified responses with retrieved external data and knowledge. However, most RAG methods retrieve relevant documents based on either sparse or dense retrieval methods or their combinations, which overlooks the essential, multi-dimensional, and structured semantic information present in documents. This structured information plays a critical role in finding concise yet highly relevant information for domain knowledge-intensive tasks, such as scientific question-answering (QA). In this work, we introduce a multi-dimensional (cube) structure, Hypercube, which can index and allocate documents in a pre-defined multi-dimensional space. Built on the hypercube, we further propose Hypercube-RAG, a novel RAG framework for precise and efficient retrieval. Given a query, Hypercube-RAG first decomposes it based on its entities, phrases, and topics along with pre-defined hypercube dimensions, and then retrieves relevant documents from cubes by aligning these decomposed components with corresponding dimensions. Experiments on three datasets across different domains demonstrate that our method improves response accuracy by 3.7% and retrieval accuracy by 5.3% over the strongest RAG baseline. It also boosts retrieval efficiency (speed) by one or two magnitudes faster than graph-based RAG. Notably, our Hypercube-RAG inherently offers explainability by revealing those underlying dimensions used for retrieval. The code and data are available at https://github.com/JimengShi/Hypercube-RAG.
Authors: Qianyi Xu, Gousia Habib, Dilruk Perera, Mengling Feng
Abstract: Timely and personalized treatment decisions are essential across a wide range of healthcare settings where patient responses can vary significantly and evolve over time. Clinical data used to support these treatment decisions are often irregularly sampled, where missing data frequencies may implicitly convey information about the patient's condition. Existing Reinforcement Learning (RL) based clinical decision support systems often ignore the missing patterns and distort them with coarse discretization and simple imputation. They are also predominantly model-free and largely depend on retrospective data, which could lead to insufficient exploration and bias by historical behaviors. To address these limitations, we propose medDreamer, a novel model-based reinforcement learning framework for personalized treatment recommendation. medDreamer contains a world model with an Adaptive Feature Integration module that simulates latent patient states from irregular data and a two-phase policy trained on a hybrid of real and imagined trajectories. This enables learning optimal policies that go beyond the sub-optimality of historical clinical decisions, while remaining close to real clinical data. We evaluate medDreamer on both sepsis and mechanical ventilation treatment tasks using two large-scale Electronic Health Records (EHRs) datasets. Comprehensive evaluations show that medDreamer significantly outperforms model-free and model-based baselines in both clinical outcomes and off-policy metrics.
Authors: Antonio Tudisco, Deborah Volpe, Giovanna Turvani
Abstract: Accurate and reliable diagnosis of diseases is crucial in enabling timely medical treatment and enhancing patient survival rates. In recent years, Machine Learning has revolutionized diagnostic practices by creating classification models capable of identifying diseases. However, these classification problems often suffer from significant class imbalances, which can inhibit the effectiveness of traditional models. Therefore, the interest in Quantum models has arisen, driven by the captivating promise of overcoming the limitations of the classical counterpart thanks to their ability to express complex patterns by mapping data in a higher-dimensional computational space.
Authors: Zhengyu Fang, Zhimeng Jiang, Huiyuan Chen, Xiaoge Zhang, Kaiyu Tang, Xiao Li, Jing Li
Abstract: Diffusion models have shown strong performance in generating high-quality tabular data, but they carry privacy risks by reproducing exact training samples. While prior work focuses on dataset-level augmentation to reduce memorization, little is known about which individual samples contribute most. We present the first data-centric study of memorization dynamics in tabular diffusion models. We quantify memorization for each real sample based on how many generated samples are flagged as replicas, using a relative distance ratio. Our empirical analysis reveals a heavy-tailed distribution of memorization counts: a small subset of samples contributes disproportionately to leakage, confirmed via sample-removal experiments. To understand this, we divide real samples into top- and non-top-memorized groups and analyze their training-time behaviors. We track when each sample is first memorized and monitor per-epoch memorization intensity (AUC). Memorized samples are memorized slightly earlier and show stronger signals in early training. Based on these insights, we propose DynamicCut, a two-stage, model-agnostic mitigation method: (a) rank samples by epoch-wise intensity, (b) prune a tunable top fraction, and (c) retrain on the filtered dataset. Across multiple tabular datasets and models, DynamicCut reduces memorization with minimal impact on data diversity and downstream performance. It also complements augmentation-based defenses. Furthermore, DynamicCut enables cross-model transferability: high-ranked samples identified from one model (e.g., a diffusion model) are also effective for reducing memorization when removed from others, such as GANs and VAEs.
Authors: Sahil Rajesh Dhayalkar
Abstract: We present a formal and constructive simulation framework for nondeterministic finite automata (NFAs) using standard feedforward neural networks. Unlike prior approaches that rely on recurrent architectures or post hoc extraction methods, our formulation symbolically encodes automaton states as binary vectors, transitions as sparse matrix transformations, and nondeterministic branching-including $\varepsilon$-closures-as compositions of shared thresholded updates. We prove that every regular language can be recognized exactly by a depth-unrolled feedforward network with shared parameters, independent of input length. Our construction yields not only formal equivalence between NFAs and neural networks, but also practical trainability: we demonstrate that these networks can learn NFA acceptance behavior through gradient descent using standard supervised data. Extensive experiments validate all theoretical results, achieving perfect or near-perfect agreement on acceptance, state propagation, and closure dynamics. This work establishes a new bridge between symbolic automata theory and modern neural architectures, showing that feedforward networks can perform precise, interpretable, and trainable symbolic computation.
Authors: Menglin Kong, Vincent Zhihao Zheng, Xudong Wang, Lijun Sun
Abstract: This paper introduces a data-driven time embedding method for modeling long-range seasonal dependencies in spatiotemporal forecasting tasks. The proposed approach employs Dynamic Mode Decomposition (DMD) to extract temporal modes directly from observed data, eliminating the need for explicit timestamps or hand-crafted time features. These temporal modes serve as time representations that can be seamlessly integrated into deep spatiotemporal forecasting models. Unlike conventional embeddings such as time-of-day indicators or sinusoidal functions, our method captures complex multi-scale periodicity through spectral analysis of spatiotemporal data. Extensive experiments on urban mobility, highway traffic, and climate datasets demonstrate that the DMD-based embedding consistently improves long-horizon forecasting accuracy, reduces residual correlation, and enhances temporal generalization. The method is lightweight, model-agnostic, and compatible with any architecture that incorporates time covariates.
Authors: Thong Q. Nguyen, Shubhang Desai, Raja Hasnain Anwar, Firoz Shaik, Vishwas Suryanarayanan, Vishal Chowdhary
Abstract: Continual memory augmentation lets computer-using agents (CUAs) learn from prior interactions, but unvetted memories can encode domain-inappropriate or unsafe heuristics--spurious rules that drift from user intent and safety constraints. We introduce VerificAgent, a scalable oversight framework that treats persistent memory as an explicit alignment surface. VerificAgent combines (1) an expert-curated seed of domain knowledge, (2) iterative, trajectory-based memory growth during training, and (3) a post-hoc human fact-checking pass to sanitize accumulated memories before deployment. Evaluated on OSWorld productivity tasks and additional adversarial stress tests, VerificAgent improves task reliability, reduces hallucination-induced failures, and preserves interpretable, auditable guidance--without additional model fine-tuning. By letting humans correct high-impact errors once, the verified memory acts as a frozen safety contract that future agent actions must satisfy. Our results suggest that domain-scoped, human-verified memory offers a scalable oversight mechanism for CUAs, complementing broader alignment strategies by limiting silent policy drift and anchoring agent behavior to the norms and safety constraints of the target domain.
Authors: Patryk Krukowski, {\L}ukasz Gorczyca, Piotr Helm, Kamil Ksi\k{a}\.zek, Przemys{\l}aw Spurek
Abstract: Continual learning under adversarial conditions remains an open problem, as existing methods often compromise either robustness, scalability, or both. We propose a novel framework that integrates Interval Bound Propagation (IBP) with a hypernetwork-based architecture to enable certifiably robust continual learning across sequential tasks. Our method, SHIELD, generates task-specific model parameters via a shared hypernetwork conditioned solely on compact task embeddings, eliminating the need for replay buffers or full model copies and enabling efficient over time. To further enhance robustness, we introduce Interval MixUp, a novel training strategy that blends virtual examples represented as $\ell_{\infty}$ balls centered around MixUp points. Leveraging interval arithmetic, this technique guarantees certified robustness while mitigating the wrapping effect, resulting in smoother decision boundaries. We evaluate SHIELD under strong white-box adversarial attacks, including PGD and AutoAttack, across multiple benchmarks. It consistently outperforms existing robust continual learning methods, achieving state-of-the-art average accuracy while maintaining both scalability and certification. These results represent a significant step toward practical and theoretically grounded continual learning in adversarial settings.
Authors: Jackson Eshbaugh
Abstract: Neural networks excel as function approximators, but their complexity often obscures what kinds of functions they learn. We introduce the linearity score $\lambda(f)$, a simple and interpretable diagnostic that quantifies how well a regression network's output can be mimicked by a linear model. Defined as the $R^2$ value between the network's predictions and those of a trained linear surrogate, $\lambda(f)$ measures linear decodability: the extent to which the network's behavior aligns with a structurally simple model. We evaluate this framework on both synthetic ($y = x \cdot \sin(x) + \epsilon$) and real-world datasets (Medical Insurance, Concrete, California Housing), using dataset-specific networks and surrogates. Our findings show that high $\lambda(f)$ scores reliably indicate alignment with the network's outputs -- but do not guarantee accuracy with respect to the ground truth. These results highlight the risk of using surrogate fidelity as a proxy for model understanding -- especially in high-stakes regression tasks.
Authors: Sunil Kumar, Bowen Zhao, Leo Dirac, Paulina Varshavskaya
Abstract: Despite tremendous recent advances in large model reasoning ability, vision-language models (VLMs) still struggle with detailed visual reasoning, especially when compute resources are limited. To address this challenge, we draw inspiration from methods like Deepseek-r1 for VLMs and train smaller-scale models with Group Relative Policy Optimization (GRPO) to use external tools such as zoom. The greatest benefit is obtained with a combination of GRPO learning, a simple reward structure, a simplified tool-calling interface, allocating additional tokens to the result of the tool call, and a training data mix that over-represents visually difficult examples. Compared to similarly-sized baseline models, our method achieves better performance on some visual question-answering (VQA) tasks, thanks to the detailed visual information gathered from the external tool.
Authors: Jittarin Jetwiriyanon, Teo Susnjak, Surangika Ranathunga
Abstract: This study investigates zero-shot forecasting capabilities of Time Series Foundation Models (TSFMs) for macroeconomic indicators. We apply TSFMs to forecasting economic indicators under univariate conditions, bypassing the need for train bespoke econometric models using and extensive training datasets. Our experiments were conducted on a case study dataset, without additional customisation. We rigorously back-tested three state-of-the-art TSFMs (Chronos, TimeGPT and Moirai) under data-scarce conditions and structural breaks. Our results demonstrate that appropriately engineered TSFMs can internalise rich economic dynamics, accommodate regime shifts, and deliver well-behaved uncertainty estimates out of the box, while matching state-of-the-art multivariate models on this domain. Our findings suggest that, without any fine-tuning, TSFMs can match or exceed classical models during stable economic conditions. However, they are vulnerable to degradation in performances during periods of rapid shocks. The findings offer guidance to practitioners on when zero-shot deployments are viable for macroeconomic monitoring and strategic planning.
Authors: Naoki Sato, Hiroki Naganuma, Hideaki Iiduka
Abstract: Muon, a recently proposed optimizer that leverages the inherent matrix structure of neural network parameters, has demonstrated strong empirical performance, indicating its potential as a successor to standard optimizers such as AdamW. This paper presents theoretical analysis to support its practical success. We provide convergence proofs for Muon across four practical settings, systematically examining its behavior with and without the inclusion of Nesterov momentum and weight decay. Our analysis covers the standard configuration using both, thereby elucidating its real-world performance. We then demonstrate that the addition of weight decay yields strictly tighter theoretical bounds and clarify the interplay between the weight decay coefficient and the learning rate. Finally, we derive the critical batch size for Muon that minimizes the computational cost of training. Our analysis identifies the hyperparameters governing this value, and our experiments validate the corresponding theoretical findings.
Authors: Shiyi Liu, Buwen Liang, Yuetong Fang, Zixuan Jiang, Renjing Xu
Abstract: Recent advances in AI for science have highlighted the power of contrastive learning in bridging heterogeneous biological data modalities. Building on this paradigm, we propose HIPPO (HIerarchical Protein-Protein interaction prediction across Organisms), a hierarchical contrastive framework for protein-protein interaction(PPI) prediction, where protein sequences and their hierarchical attributes are aligned through multi-tiered biological representation matching. The proposed approach incorporates hierarchical contrastive loss functions that emulate the structured relationship among functional classes of proteins. The framework adaptively incorporates domain and family knowledge through a data-driven penalty mechanism, enforcing consistency between the learned embedding space and the intrinsic hierarchy of protein functions. Experiments on benchmark datasets demonstrate that HIPPO achieves state-of-the-art performance, outperforming existing methods and showing robustness in low-data regimes. Notably, the model demonstrates strong zero-shot transferability to other species without retraining, enabling reliable PPI prediction and functional inference even in less characterized or rare organisms where experimental data are limited. Further analysis reveals that hierarchical feature fusion is critical for capturing conserved interaction determinants, such as binding motifs and functional annotations. This work advances cross-species PPI prediction and provides a unified framework for interaction prediction in scenarios with sparse or imbalanced multi-species data.
Authors: Md Tanvirul Alam, Aritran Piplai, Nidhi Rastogi
Abstract: Machine learning models are commonly used for malware classification; however, they suffer from performance degradation over time due to concept drift. Adapting these models to changing data distributions requires frequent updates, which rely on costly ground truth annotations. While active learning can reduce the annotation burden, leveraging unlabeled data through semi-supervised learning remains a relatively underexplored approach in the context of malware detection. In this research, we introduce \texttt{ADAPT}, a novel pseudo-labeling semi-supervised algorithm for addressing concept drift. Our model-agnostic method can be applied to various machine learning models, including neural networks and tree-based algorithms. We conduct extensive experiments on five diverse malware detection datasets spanning Android, Windows, and PDF domains. The results demonstrate that our method consistently outperforms baseline models and competitive benchmarks. This work paves the way for more effective adaptation of machine learning models to concept drift in malware detection.
Authors: Runze Yang, Longbing Cao, Xin You, Kun Fang, Jianxun Li, Jie Yang
Abstract: The integration of Fourier transform and deep learning opens new avenues for time series forecasting. We reconsider the Fourier transform from a basis functions perspective. Specifically, the real and imaginary parts of the frequency components can be regarded as the coefficients of cosine and sine basis functions at tiered frequency levels, respectively. We find that existing Fourier-based methods face inconsistent starting cycles and inconsistent series length issues. They fail to interpret frequency components precisely and overlook temporal information. Accordingly, the novel Fourier Basis Mapping (FBM) method addresses these issues by integrating time-frequency features through Fourier basis expansion and mapping in the time-frequency space. Our approach extracts explicit frequency features while preserving temporal characteristics. FBM supports plug-and-play integration with various types of neural networks by only adjusting the first initial projection layer for better performance. First, we propose FBM-L, FBM-NL, and FBM-NP to enhance linear, MLP-based, and Transformer-based models, respectively, demonstrating the effectiveness of time-frequency features. Next, we propose a synergetic model architecture, termed FBM-S, which decomposes the seasonal, trend, and interaction effects into three separate blocks, each designed to model time-frequency features in a specialized manner. Finally, we introduce several techniques tailored for time-frequency features, including interaction masking, centralization, patching, rolling window projection, and multi-scale down-sampling. The results are validated on diverse real-world datasets for both long-term and short-term forecasting tasks with SOTA performance.
Authors: Yunfeng Li, Junhong Liu, Zhaohui Yang, Guofu Liao, Chuyun Zhang
Abstract: Deep learning models have been widely adopted for False Data Injection Attack (FDIA) detection in smart grids due to their ability to capture unstructured and sparse features. However, the increasing system scale and data dimensionality introduce significant computational and memory burdens, particularly in large-scale industrial datasets, limiting detection efficiency. To address these issues, this paper proposes Rec-AD, a computationally efficient framework that integrates Tensor Train decomposition with the Deep Learning Recommendation Model (DLRM). Rec-AD enhances training and inference efficiency through embedding compression, optimized data access via index reordering, and a pipeline training mechanism that reduces memory communication overhead. Fully compatible with PyTorch, Rec-AD can be integrated into existing FDIA detection systems without code modifications. Experimental results show that Rec-AD significantly improves computational throughput and real-time detection performance, narrowing the attack window and increasing attacker cost. These advancements strengthen edge computing capabilities and scalability, providing robust technical support for smart grid security.
Authors: Yunfeng Li, Junhong Liu, Zhaohui Yang, Guofu Liao, Chuyun Zhang
Abstract: False Data Injection Attacks (FDIAs) pose severe security risks to smart grids by manipulating measurement data collected from spatially distributed devices such as SCADA systems and PMUs. These measurements typically exhibit Non-Independent and Identically Distributed (Non-IID) characteristics across different regions, which significantly challenges the generalization ability of detection models. Traditional centralized training approaches not only face privacy risks and data sharing constraints but also incur high transmission costs, limiting their scalability and deployment feasibility. To address these issues, this paper proposes a privacy-preserving federated learning framework, termed Federated Cluster Average (FedClusAvg), designed to improve FDIA detection in Non-IID and resource-constrained environments. FedClusAvg incorporates cluster-based stratified sampling and hierarchical communication (client-subserver-server) to enhance model generalization and reduce communication overhead. By enabling localized training and weighted parameter aggregation, the algorithm achieves accurate model convergence without centralizing sensitive data. Experimental results on benchmark smart grid datasets demonstrate that FedClusAvg not only improves detection accuracy under heterogeneous data distributions but also significantly reduces communication rounds and bandwidth consumption. This work provides an effective solution for secure and efficient FDIA detection in large-scale distributed power systems.
Authors: Xing Hu, Haodong Chen, Qianqian Duan, Choon Ki Ahn, Huiliang Shang, Dawei Zhang Zhang
Abstract: With the global population increasing and arable land resources becoming increasingly limited, smart and precision agriculture have emerged as essential directions for sustainable agricultural development. Artificial intelligence (AI), particularly deep learning models, has been widely adopted in applications such as crop monitoring, pest detection, and yield prediction. Among recent generative models, diffusion models have demonstrated considerable potential in agricultural image processing, data augmentation, and remote sensing analysis. Compared to traditional generative adversarial networks (GANs), diffusion models exhibit greater training stability and superior image generation quality, effectively addressing challenges such as limited annotated datasets and imbalanced sample distributions in agricultural scenarios. This paper reviews recent advancements in the application of diffusion models within agriculture, focusing on their roles in crop disease and pest detection, remote sensing image enhancement, crop growth prediction, and agricultural resource management. Empirical studies show that diffusion models significantly enhance the performance of downstream models by improving accuracy, robustness, and generalization in tasks involving image synthesis, augmentation, and denoising under complex environmental conditions. Despite ongoing challenges in computational efficiency and domain generalization, diffusion models are expected to play an increasingly important role in the future of intelligent agriculture. As the technology continues to evolve, it holds substantial promise for addressing pressing global issues in food security and environmental sustainability.
Authors: Camilo Tamayo-Rousseau, Yunjia Zhao, Yiqun Zhang, Randall Balestriero
Abstract: Self-attention mechanisms are foundational to Transformer architectures, supporting their impressive success in a wide range of tasks. While there are many self-attention variants, their robustness to noise and spurious correlations has not been well studied. This study evaluates Softmax, Sigmoid, Linear, Doubly Stochastic, and Cosine attention within Vision Transformers under different data corruption scenarios. Through testing across the CIFAR-10, CIFAR-100, and Imagenette datasets, we show that Doubly Stochastic attention is the most robust. It consistently outperformed the next best mechanism by $0.1\%-5.1\%$ when training data, or both training and testing data, were corrupted. Our findings inform self-attention selection in contexts with imperfect data. The code used is available at https://github.com/ctamayor/NeurIPS-Robustness-ViT.
Authors: Guangchen Lan, Sipeng Zhang, Tianle Wang, Yuwei Zhang, Daoan Zhang, Xinpeng Wei, Xiaoman Pan, Hongming Zhang, Dong-Jun Han, Christopher G. Brinton
Abstract: As the era of large language models (LLMs) on behalf of users unfolds, Preference Optimization (PO) methods have become a central approach to aligning LLMs with human preferences and improving performance. We propose Maximum a Posteriori Preference Optimization (MaPPO), a framework for learning from preferences that explicitly incorporates prior reward knowledge into the optimization objective. While existing methods such as Direct Preference Optimization (DPO) and its variants treat preference learning as a Maximum Likelihood Estimation (MLE) problem, MaPPO extends this paradigm by integrating prior reward estimates into a principled Maximum a Posteriori (MaP) objective. This not only generalizes DPO and its variants, but also enhances alignment by mitigating the oversimplified binary classification of responses. More importantly, MaPPO introduces no additional hyperparameter, and supports preference optimization in both offline and online settings. In addition, MaPPO can be used as a plugin with consistent improvement on DPO variants, including widely used SimPO, IPO, and CPO. Extensive empirical evaluations of different model sizes and model series on three standard benchmarks, including MT-Bench, AlpacaEval 2.0, and Arena-Hard, demonstrate consistent improvements in alignment performance without sacrificing computational efficiency.
Authors: Shiva Raja, Cansu Demirkiran, Aakash Sarkar, Milos Popovic, Ajay Joshi
Abstract: Sequence modeling is crucial for AI to understand temporal data and detect complex time-dependent patterns. While recurrent neural networks (RNNs), convolutional neural networks (CNNs), and Transformers have advanced in capturing long-range dependencies, they struggle with achieving high accuracy with very long sequences due to limited memory retention (fixed context window). State-Space Models (SSMs) leverage exponentially decaying memory enabling lengthy context window and so they process very long data sequences more efficiently than recurrent and Transformer-based models. Unlike traditional neural models like CNNs and RNNs, SSM-based models require solving differential equations through continuous integration, making training and inference both compute- and memory-intensive on conventional CPUs and GPUs. In this paper we introduce a specialized hardware accelerator, EpochCore, for accelerating SSMs. EpochCore is based on systolic arrays (SAs) and is designed to enhance the energy efficiency and throughput of inference of SSM-based models for long-range sequence tasks. Within the SA, we propose a versatile processing element (PE) called LIMA-PE to perform traditional and specialized MAC operations to support traditional DNNs and SSMs. To complement the EpochCore microarchitecture, we propose a novel dataflow, ProDF, which enables highly efficient execution of SSM-based models. By leveraging the LIMA-PE microarchitecture and ProDF, EpochCore achieves on average 250x gains in performance and 45x improvement in energy efficiency, at the expense of 2x increase in area cost over traditional SA-based accelerators, and around ~2,000x improvement in latency/inference on LRA datasets compared to GPU kernel operations.
Authors: Xueying Wu, Baijun Zhou, Zhihui Gao, Yuzhe Fu, Qilin Zheng, Yintao He, Hai Li
Abstract: Large language model (LLM) inference poses significant challenges due to its intensive memory and computation demands. Weight and activation quantization (WAQ) offers a promising solution by reducing both memory footprint and arithmetic complexity. Traditional WAQ designs rely on uniform integer quantization for hardware efficiency, but often suffer from significant model performance degradation at low precision. In contrast, K-Means quantization, a non-uniform technique, achieves higher accuracy by aligning with the Gaussian-like distributions of weights and activations in LLMs. However, two key challenges prevent the efficient deployment of K-Means-based WAQ designs for LLM inference: (1) The non-uniform structure of K-Means-quantized data precludes direct execution on low-precision compute units, necessitating dequantization and floating-point matrix multiplications (MatMuls) during inference. (2) Activation outliers hinder effective low-precision quantization. Offline thresholding methods for outlier detection degrade model performance substantially, while existing online detection techniques introduce significant runtime overhead. To address the aforementioned challenges and fully unleash the potential of K-Means-based WAQ for LLM inference, in this paper, we propose KLLM, an LLM inference accelerator for efficient execution with K-Means-quantized weights and activations. KLLM features an index-based computation scheme for efficient execution of MatMuls and nonlinear operations on K-Means-quantized data, which avoids most of the dequantization and full-precision computations. Moreover, KLLM incorporates a lightweight outlier detection engine, Orizuru, that efficiently identifies the top-$k$ largest and smallest elements in the activation data stream during online inference.
Authors: Jinan Zhou, Rajat Ghosh, Vaishnavi Bhargava, Debojyoti Dutta, Aryan Singhal
Abstract: When designing LLM services, practitioners care about three key properties: inference-time budget, factual authenticity, and reasoning capacity. However, our analysis shows that no model can simultaneously optimize for all three. We formally prove this trade-off and propose a principled framework named The BAR Theorem for LLM-application design.
Authors: Yuetian Chen, Zhiqi Wang, Nathalie Baracaldo, Swanand Ravindra Kadhe, Lei Yu
Abstract: Membership inference attacks (MIAs) pose a critical threat to the privacy of training data in deep learning. Despite significant progress in attack methodologies, our understanding of when and how models encode membership information during training remains limited. This paper presents a dynamic analytical framework for dissecting and quantifying privacy leakage dynamics at the individual sample level. By tracking per-sample vulnerabilities on an FPR-TPR plane throughout training, our framework systematically measures how factors such as dataset complexity, model architecture, and optimizer choice influence the rate and severity at which samples become vulnerable. Crucially, we discover a robust correlation between a sample's intrinsic learning difficulty, and find that the privacy risk of samples highly vulnerable in the final trained model is largely determined early during training. Our results thus provide a deeper understanding of how privacy risks dynamically emerge during training, laying the groundwork for proactive, privacy-aware model training strategies.
Authors: Shervin Rahimzadeh Arashloo
Abstract: We generalise the recently introduced large-margin $\ell_p$-SVDD approach to exploit the geometry of data distribution via manifold regularising for time series anomaly detection. Specifically, we formulate a manifold-regularised variant of the $\ell_p$-SVDD method to encourage label smoothness on the underlying manifold to capture structural information for improved detection performance. Drawing on an existing Representer theorem, we then provide an effective optimisation technique for the proposed method. We theoretically study the proposed approach using Rademacher complexities to analyse its generalisation performance and also provide an experimental assessment of the proposed method across various data sets to compare its performance against other methods.
Authors: Md. Ehsanul Haque, S. M. Jahidul Islam, Shakil Mia, Rumana Sharmin, Ashikuzzaman, Md Samir Morshed, Md. Tahmidul Huque
Abstract: Liver diseases are a serious health concern in the world, which requires precise and timely diagnosis to enhance the survival chances of patients. The current literature implemented numerous machine learning and deep learning models to classify liver diseases, but most of them had some issues like high misclassification error, poor interpretability, prohibitive computational expense, and lack of good preprocessing strategies. In order to address these drawbacks, we introduced StackLiverNet in this study; an interpretable stacked ensemble model tailored to the liver disease detection task. The framework uses advanced data preprocessing and feature selection technique to increase model robustness and predictive ability. Random undersampling is performed to deal with class imbalance and make the training balanced. StackLiverNet is an ensemble of several hyperparameter-optimized base classifiers, whose complementary advantages are used through a LightGBM meta-model. The provided model demonstrates excellent performance, with the testing accuracy of 99.89%, Cohen Kappa of 0.9974, and AUC of 0.9993, having only 5 misclassifications, and efficient training and inference speeds that are amenable to clinical practice (training time 4.2783 seconds, inference time 0.1106 seconds). Besides, Local Interpretable Model-Agnostic Explanations (LIME) are applied to generate transparent explanations of individual predictions, revealing high concentrations of Alkaline Phosphatase and moderate SGOT as important observations of liver disease. Also, SHAP was used to rank features by their global contribution to predictions, while the Morris method confirmed the most influential features through sensitivity analysis.
Authors: Antonio Tudisco, Andrea Marchesin, Maurizio Zamboni, Mariagrazia Graziano, Giovanna Turvani
Abstract: Recent advancements in Quantum Computing and Machine Learning have increased attention to Quantum Machine Learning (QML), which aims to develop machine learning models by exploiting the quantum computing paradigm. One of the widely used models in this area is the Variational Quantum Circuit (VQC), a hybrid model where the quantum circuit handles data inference while classical optimization adjusts the parameters of the circuit. The quantum circuit consists of an encoding layer, which loads data into the circuit, and a template circuit, known as the ansatz, responsible for processing the data. This work involves performing an analysis by considering both Amplitude- and Angle-encoding models, and examining how the type of rotational gate applied affects the classification performance of the model. This comparison is carried out by training the different models on two datasets, Wine and Diabetes, and evaluating their performance. The study demonstrates that, under identical model topologies, the difference in accuracy between the best and worst models ranges from 10% to 30%, with differences reaching up to 41%. Moreover, the results highlight how the choice of rotational gates used in encoding can significantly impact the model's classification performance. The findings confirm that the embedding represents a hyperparameter for VQC models.
Authors: Andrea Coccaro, Marco Letizia, Humberto Reyes-Gonzalez, Riccardo Torre
Abstract: Normalizing flows have emerged as a powerful brand of generative models, as they not only allow for efficient sampling of complicated target distributions but also deliver density estimation by construction. We propose here an in-depth comparison of coupling and autoregressive flows, both based on symmetric (affine) and non-symmetric (rational quadratic spline) bijectors, considering four different architectures: real-valued non-Volume preserving (RealNVP), masked autoregressive flow (MAF), coupling rational quadratic spline (C-RQS), and autoregressive rational quadratic spline (A-RQS). We focus on a set of multimodal target distributions of increasing dimensionality ranging from 4 to 400. The performances were compared by means of different test statistics for two-sample tests, built from known distance measures: the sliced Wasserstein distance, the dimension-averaged one-dimensional Kolmogorov--Smirnov test, and the Frobenius norm of the difference between correlation matrices. Furthermore, we included estimations of the variance of both the metrics and the trained models. Our results indicate that the A-RQS algorithm stands out both in terms of accuracy and training speed. Nonetheless, all the algorithms are generally able, without too much fine-tuning, to learn complicated distributions with limited training data and in a reasonable time of the order of hours on a Tesla A40 GPU. The only exception is the C-RQS, which takes significantly longer to train, does not always provide good accuracy, and becomes unstable for large dimensionalities. All algorithms were implemented using \textsc{TensorFlow2} and \textsc{TensorFlow Probability} and have been made available on \href{https://github.com/NF4HEP/NormalizingFlowsHD}{GitHub}.
Authors: Xuan Ren, Zeyu Zhang, Lingqiao Liu
Abstract: Small language models like T5 excel in generating high-quality text for data-to-text tasks, offering adaptability and cost-efficiency compared to Large Language Models (LLMs). However, they frequently miss keywords, which is considered one of the most severe and common errors in this task. In this work, we explore the potential of using feedback systems to enhance semantic fidelity in smaller language models for data-to-text generation tasks, through our Verification and Correction Prompting (VCP) approach. In the inference stage, our approach involves a multi-step process, including generation, verification, and regeneration stages. During the verification stage, we implement a simple rule to check for the presence of every keyword in the prediction. Recognizing that this rule can be inaccurate, we have developed a carefully designed training procedure, which enabling the model to incorporate feedback from the error-correcting prompt effectively, despite its potential inaccuracies. The VCP approach effectively reduces the Semantic Error Rate (SER) while maintaining the text's quality.
Authors: Prateek Mittal, Puneet Goyal, Joohi Chauhan
Abstract: This study introduces a novel multimodal food recognition framework that effectively combines visual and textual modalities to enhance classification accuracy and robustness. The proposed approach employs a dynamic multimodal fusion strategy that adaptively integrates features from unimodal visual inputs and complementary textual metadata. This fusion mechanism is designed to maximize the use of informative content, while mitigating the adverse impact of missing or inconsistent modality data. The framework was rigorously evaluated on the UPMC Food-101 dataset and achieved unimodal classification accuracies of 73.60% for images and 88.84% for text. When both modalities were fused, the model achieved an accuracy of 97.84%, outperforming several state-of-the-art methods. Extensive experimental analysis demonstrated the robustness, adaptability, and computational efficiency of the proposed settings, highlighting its practical applicability to real-world multimodal food-recognition scenarios.
Authors: Yuping Wang, Jier Chen
Abstract: In autonomous driving, deep learning enabled motion prediction is a popular topic. A critical gap in traditional motion prediction methodologies lies in ensuring equivariance under Euclidean geometric transformations and maintaining invariant interaction relationships. This research introduces a groundbreaking solution by employing EqMotion, a theoretically geometric equivariant and interaction invariant motion prediction model for particles and humans, plus integrating agent-equivariant high-definition (HD) map features for context aware motion prediction in autonomous driving. The use of EqMotion as backbone marks a significant departure from existing methods by rigorously ensuring motion equivariance and interaction invariance. Equivariance here implies that an output motion must be equally transformed under the same Euclidean transformation as an input motion, while interaction invariance preserves the manner in which agents interact despite transformations. These properties make the network robust to arbitrary Euclidean transformations and contribute to more accurate prediction. In addition, we introduce an equivariant method to process the HD map to enrich the spatial understanding of the network while preserving the overall network equivariance property. By applying these technologies, our model is able to achieve high prediction accuracy while maintain a lightweight design and efficient data utilization.
Authors: Yuping Wang, Jier Chen
Abstract: Forecasting vehicular motions in autonomous driving requires a deep understanding of agent interactions and the preservation of motion equivariance under Euclidean geometric transformations. Traditional models often lack the sophistication needed to handle the intricate dynamics inherent to autonomous vehicles and the interaction relationships among agents in the scene. As a result, these models have a lower model capacity, which then leads to higher prediction errors and lower training efficiency. In our research, we employ EqMotion, a leading equivariant particle, and human prediction model that also accounts for invariant agent interactions, for the task of multi-agent vehicle motion forecasting. In addition, we use a multi-modal prediction mechanism to account for multiple possible future paths in a probabilistic manner. By leveraging EqMotion, our model achieves state-of-the-art (SOTA) performance with fewer parameters (1.2 million) and a significantly reduced training time (less than 2 hours).
Authors: Lei Zhao, Aishwarya Natarjan, Luca Buonanno, Archit Gajjar, Ron M. Roth, Sergey Serebryakov, John Moon, Jim Ignowski, Giacomo Pedretti
Abstract: Transformer models represent the cutting edge of Deep Neural Networks (DNNs) and excel in a wide range of machine learning tasks. However, processing these models demands significant computational resources and results in a substantial memory footprint. While In-memory Computing (IMC)offers promise for accelerating Vector-Matrix Multiplications(VMMs) with high computational parallelism and minimal data movement, employing it for other crucial DNN operators remains a formidable task. This challenge is exacerbated by the extensive use of complex activation functions, Softmax, and data-dependent matrix multiplications (DMMuls) within Transformer models. To address this challenge, we introduce a Reconfigurable Analog Computing Engine (RACE) by enhancing Analog Content Addressable Memories (ACAMs) to support broader operations. Based on the RACE, we propose the RACE-IT accelerator (meaning RACE for In-memory Transformers) to enable efficient analog-domain execution of all core operations of Transformer models. Given the flexibility of our proposed RACE in supporting arbitrary computations, RACE-IT is well-suited for adapting to emerging and non-traditional DNN architectures without requiring hardware modifications. We compare RACE-IT with various accelerators. Results show that RACE-IT increases performance by 453x and 15x, and reduces energy by 354x and 122x over the state-of-the-art GPUs and existing Transformer-specific IMC accelerators, respectively.
Authors: Zhi-Feng Wei, Pablo Moriano, Ramakrishnan Kannan
Abstract: This study investigates the robustness of graph embedding methods for community detection in the face of network perturbations, specifically edge deletions. Graph embedding techniques, which represent nodes as low-dimensional vectors, are widely used for various graph machine learning tasks due to their ability to capture structural properties of networks effectively. However, the impact of perturbations on the performance of these methods remains relatively understudied. The research considers state-of-the-art graph embedding methods from two families: matrix factorization (e.g., LE, LLE, HOPE, M-NMF) and random walk-based (e.g., DeepWalk, LINE, node2vec). Through experiments conducted on both synthetic and real-world networks, the study reveals varying degrees of robustness within each family of graph embedding methods. The robustness is found to be influenced by factors such as network size, initial community partition strength, and the type of perturbation. Notably, node2vec and LLE consistently demonstrate higher robustness for community detection across different scenarios, including networks with degree and community size heterogeneity. These findings highlight the importance of selecting an appropriate graph embedding method based on the specific characteristics of the network and the task at hand, particularly in scenarios where robustness to perturbations is crucial.
Authors: Bolian Li, Yifan Wang, Anamika Lochab, Ananth Grama, Ruqi Zhang
Abstract: Aligning large language models (LLMs) with human preferences is essential for their applications. Recently, decoding-time alignment has emerged as an effective plug-and-play technique that avoids fine-tuning model parameters. This approach retains the general utility of pretrained LLMs but often suffers from significant inefficiencies during decoding, primarily due to wasted token generation and excessive reward evaluations. To address these challenges, we introduce Cascade Reward Sampling (CARDS) to resolve both efficiency bottlenecks in decoding-time alignment. Specifically, we develop a segment-level rejection sampling algorithm that minimizes redundant computations of both LLMs and reward models (RMs). Central to CARDS is an uncertainty-based segmentation mechanism, which ensures the accuracy of RMs evaluations on incomplete segments. Furthermore, we provide a detailed analysis of reward scores on segments to elucidate the improved alignment performance. Experimental results demonstrate that CARDS significantly improves decoding efficiency, alignment quality, and general utility compared to existing decoding-time alignment methods, achieving approximately a 70% reduction in decoding time and over 90% win-ties in utility and safety benchmarks.
Authors: Heng Gao, Jun Li
Abstract: Out-of-distribution (OOD) detection is crucial for the reliable deployment of machine learning models in real-world scenarios, enabling the identification of unknown samples or objects. A prominent approach to enhance OOD detection performance involves leveraging auxiliary datasets for training. Recent efforts have explored using generative models, such as Stable Diffusion (SD), to synthesize outlier data in the pixel space. However, synthesizing OOD data in the pixel space can lead to reduced robustness due to over-generation. To address this challenge, we propose Outlier-Aware Learning (OAL), a novel framework that generates synthetic OOD training data within the latent space, taking a further step to study how to utilize Stable Diffusion for developing a latent-based outlier synthesis approach. This improvement facilitates network training with fewer outliers and less computational cost. Besides, to regularize the model's decision boundary, we develop a mutual information-based contrastive learning module (MICL) that amplifies the distinction between In-Distribution (ID) and collected OOD data. Moreover, we develop a knowledge distillation module to prevent the degradation of ID classification accuracy when training with OOD data. The superior performance of our method on several benchmark datasets demonstrates its efficiency and effectiveness. Source code is available in https://github.com/HengGao12/OAL.
Authors: Yan Sun, Pratik Chaudhari, Ian J. Barnett, Edgar Dobriban
Abstract: Recent advances in machine learning have significantly improved prediction accuracy in various applications. However, ensuring the calibration of probabilistic predictions remains a significant challenge. Despite efforts to enhance model calibration, the rigorous statistical evaluation of model calibration remains less explored. In this work, we develop confidence intervals the $\ell_2$ Expected Calibration Error (ECE). We consider top-1-to-$k$ calibration, which includes both the popular notion of confidence calibration as well as full calibration. For a debiased estimator of the ECE, we show asymptotic normality, but with different convergence rates and asymptotic variances for calibrated and miscalibrated models. We develop methods to construct asymptotically valid confidence intervals for the ECE, accounting for this behavior as well as non-negativity. Our theoretical findings are supported through extensive experiments, showing that our methods produce valid confidence intervals with shorter lengths compared to those obtained by resampling-based methods.
Authors: Jiawei Lian, Shaohui Mei, Xiaofei Wang, Yi Wang, Lefan Wang, Yingjie Lu, Mingyang Ma, Lap-Pui Chau
Abstract: It has been widely substantiated that deep neural networks (DNNs) are susceptible and vulnerable to adversarial perturbations. Existing studies mainly focus on performing attacks by corrupting targeted objects (physical attack) or images (digital attack), which is intuitively acceptable and understandable in terms of the attack's effectiveness. In contrast, our focus lies in conducting background adversarial attacks in both digital and physical domains, without causing any disruptions to the targeted objects themselves. Specifically, an effective background adversarial attack framework is proposed to attack anything, by which the attack efficacy generalizes well between diverse objects, models, and tasks. Technically, we approach the background adversarial attack as an iterative optimization problem, analogous to the process of DNN learning. Besides, we offer a theoretical demonstration of its convergence under a set of mild but sufficient conditions. To strengthen the attack efficacy and transferability, we propose a new ensemble strategy tailored for adversarial perturbations and introduce an improved smooth constraint for the seamless connection of integrated perturbations. We conduct comprehensive and rigorous experiments in both digital and physical domains across various objects, models, and tasks, demonstrating the effectiveness of attacking anything of the proposed method. The findings of this research substantiate the significant discrepancy between human and machine vision on the value of background variations, which play a far more critical role than previously recognized, necessitating a reevaluation of the robustness and reliability of DNNs. The code will be publicly available at https://github.com/JiaweiLian/Attack_Anything
Authors: Yihe Wang, Nadia Mammone, Darina Petrovsky, Alexandros T. Tzallas, Francesco C. Morabito, Xiang Zhang
Abstract: Electroencephalography (EEG) has emerged as a cost-effective and efficient tool to support neurologists in the detection of Alzheimer's Disease (AD). However, most existing approaches rely heavily on manual feature engineering or data transformation. While such techniques may provide benefits when working with small-scale datasets, they often lead to information loss and distortion when applied to large-scale data, ultimately limiting model performance. Moreover, the limited subject scale and demographic diversity of datasets used in prior studies hinder comprehensive evaluation of model robustness and generalizability, thus restricting their applicability in real-world clinical settings. To address these challenges, we propose ADformer, a novel multi-granularity spatial-temporal transformer designed to capture both temporal and spatial features from raw EEG signals, enabling effective end-to-end representation learning. Our model introduces multi-granularity embedding strategies across both spatial and temporal dimensions, leveraging a two-stage intra-inter granularity self-attention mechanism to learn both local patterns within each granularity and global dependencies across granularities. We evaluate ADformer on 4 large-scale datasets comprising a total of 1,713 subjects, representing one of the largest corpora for EEG-based AD detection to date, under a cross-validated, subject-independent setting. Experimental results demonstrate that ADformer consistently outperforms existing methods, achieving subject-level F1 scores of 92.82%, 89.83%, 67.99%, and 83.98% on the 4 datasets, respectively, in distinguishing AD from healthy control (HC) subjects.
Authors: Xin Bing, Florentina Bunea, Jonathan Niles-Weed, Marten Wegkamp
Abstract: Softmax mixture models (SMMs) are discrete $K$-mixtures introduced to model the probability of choosing an attribute $x_j \in \RR^L$ from $p$ candidates, in heterogeneous populations. They have been known as mixed multinomial logits in the econometrics literature, and are gaining traction in the LLM literature, where single softmax models are routinely used in the final layer of a neural network. This paper provides a comprehensive analysis of the EM algorithm for SMMs in high dimensions. Its population-level theoretical analysis forms the basis for proving (i) local identifiability, in SSMs with generic features and, further, via a stochastic argument, (ii) full identifiability in SSMs with random features, when $p$ is large enough. These are the first results in this direction for SSMs with $L > 1$. The population-level EM analysis characterizes the initialization radius for algorithmic convergence. This also guides the construction of warm starts of the sample level EM. Under suitable initialization, the EM algorithm is shown to recover the mixture atoms of the SSM at near-parametric rate. We provide two main directions for warm start construction, both based on a new method for estimating the moments of the mixing measure underlying an SSM with random design. First, we construct a method of moments (MoM) estimator of the mixture parameters, and provide its first theoretical analysis. While MoM can enjoy parametric rates of convergence, and thus can serve as a warm-start, the estimator's quality degrades exponentially in $K$. Our recommendation, when $K$ is not small, is to run the EM algorithm several times with random initializations. We again make use of the novel latent moments estimation method to estimate the $K$-dimensional subspace of the mixture atoms. Sampling from this subspace reduces substantially the number of required draws.
Authors: Tuan-Dung Bui, Duc-Thieu Luu-Van, Thanh-Phat Nguyen, Thu-Trang Nguyen, Son Nguyen, Hieu Dinh Vo
Abstract: Code completion is essential in software development, helping developers by predicting code snippets based on context. Among completion tasks, Method Body Completion (MBC) is particularly challenging as it involves generating complete method bodies based on their signatures and context. This task becomes significantly harder in large repositories, where method bodies must integrate repositoryspecific elements such as custom APIs, inter-module dependencies, and project-specific conventions. In this paper, we introduce RAMBO, a novel RAG-based approach for repository-level MBC. Instead of retrieving similar method bodies, RAMBO identifies essential repository-specific elements, such as classes, methods, and variables/fields, and their relevant usages. By incorporating these elements and their relevant usages into the code generation process, RAMBO ensures more accurate and contextually relevant method bodies. Our experimental results with leading code LLMs across 40 Java projects show that RAMBO significantly outperformed the state-of-the-art repository-level MBC approaches, with the improvements of up to 46% in BLEU, 57% in CodeBLEU, 36% in Compilation Rate, and up to 3X in Exact Match. Notably, RAMBO surpassed RepoCoder Oracle method by up to 12% in Exact Match, setting a new benchmark for repository-level MBC.
Authors: Taras Kucherenko, Derek Peristy, Judith B\"utepage
Abstract: Animation data is often obtained through optical motion capture systems, which utilize a multitude of cameras to establish the position of optical markers. However, system errors or occlusions can result in missing markers, the manual cleaning of which can be time-consuming. This has sparked interest in machine learning-based solutions for missing marker reconstruction in the academic community. Most academic papers utilize a simplistic mean square error as the main metric. In this paper, we show that this metric does not correlate with subjective perception of the fill quality. Additionally, we introduce and evaluate a set of better-correlated metrics that can drive progress in the field.
Authors: Mathilde Faanes, Ragnhild Holden Helland, Ole Solheim, S\'ebastien Muller, Ingerid Reinertsen
Abstract: Automatic segmentation of brain tumors in intra-operative ultrasound (iUS) images could facilitate localization of tumor tissue during resection surgery. The lack of large annotated datasets limits the current models performances. In this paper, we investigated the use of tumor annotations in magnetic resonance imaging (MRI) scans, which are more accessible than annotations in iUS images, for training of deep learning models for iUS brain tumor segmentation. We used 180 annotated MRI scans with corresponding unannotated iUS images, and 29 annotated iUS images. Image registration was performed to transfer the MRI annotations to the corresponding iUS images before training the nnU-Net model with different configurations of the data and label origins. The results showed no significant difference in Dice score for a model trained with only MRI annotated tumors compared to models trained with only iUS annotations and both, and to expert annotations, indicating that MRI tumor annotations can be used as a substitute for iUS tumor annotations to train a deep learning model for automatic brain tumor segmentation in iUS images. The best model obtained an average Dice score of $0.62\pm0.31$, compared to $0.67\pm0.25$ for an expert neurosurgeon, where the performance on larger tumors were similar, but lower for the models on smaller tumors. In addition, the results showed that removing smaller tumors from the training sets improved the results. The main models are available here: https://github.com/mathildefaanes/us_brain_tumor_segmentation/tree/main
URLs: https://github.com/mathildefaanes/us_brain_tumor_segmentation/tree/main
Authors: Youwei Zhou, Tianyang Xu, Cong Wu, Xiaojun Wu, Josef Kittler
Abstract: The shared topology of human skeletons motivated the recent investigation of graph convolutional network (GCN) solutions for action recognition. However, most of the existing GCNs rely on the binary connection of two neighboring vertices (joints) formed by an edge (bone), overlooking the potential of constructing multi-vertex convolution structures. Although some studies have attempted to utilize hyper-graphs to represent the topology, they rely on a fixed construction strategy, which limits their adaptivity in uncovering the intricate latent relationships within the action. In this paper, we address this oversight and explore the merits of an adaptive hyper-graph convolutional network (Hyper-GCN) to achieve the aggregation of rich semantic information conveyed by skeleton vertices. In particular, our Hyper-GCN adaptively optimises the hyper-graphs during training, revealing the action-driven multi-vertex relations. Besides, virtual connections are often designed to support efficient feature aggregation, implicitly extending the spectrum of dependencies within the skeleton. By injecting virtual connections into hyper-graphs, the semantic clues of diverse action categories can be highlighted. The results of experiments conducted on the NTU-60, NTU-120, and NW-UCLA datasets demonstrate the merits of our Hyper-GCN, compared to the state-of-the-art methods. The code is available at https://github.com/6UOOON9/Hyper-GCN.
Authors: Yifan Zhang
Abstract: The rapid advancement of large language models (LLMs) such as GPT-3, PaLM, and Llama has significantly transformed natural language processing, showcasing remarkable capabilities in understanding and generating language. However, a fundamental bottleneck persists: these models often struggle with tasks requiring complex, multi-step reasoning, particularly in mathematical problem-solving. This deficiency stems from the critical scarcity of large-scale, high-quality, domain-specific datasets necessary for cultivating sophisticated reasoning abilities. To overcome this challenge, we introduce Template-based Data Generation (TDG), a novel and scalable paradigm that harnesses frontier LLMs (GPT-4) to automatically generate parameterized meta-templates, which in turn synthesize a virtually infinite stream of high-quality problems and solutions. Using this paradigm, we create TemplateMath Part I: TemplateGSM, a foundational dataset of over 7 million synthetically generated grade school math problems. Each problem is accompanied by a programmatically verifiable solution, offering an unprecedented level of quality at scale. This resource not only resolves the data scarcity issue for supervised fine-tuning but also provides a robust mechanism for model alignment through Reinforcement Learning with Verifiable Rewards (RLVR). Our approach elevates data augmentation by employing GPT-4 for meta-template creation, guaranteeing diverse and complex problem structures. By providing a scalable solution to the data and verification bottleneck, TDG and TemplateGSM pave the way for a new generation of LLMs with powerful, reliable reasoning skills. The code and data are available at https://github.com/iiis-ai/TemplateMath.
Authors: Yaofo Chen, Zeng You, Shuhai Zhang, Haokun Li, Yirui Li, Yaowei Wang, Mingkui Tan
Abstract: Transformer-based Large Language Models (LLMs) have exhibited remarkable success in extensive tasks primarily attributed to self-attention mechanism, which requires a token to consider all preceding tokens as its context to compute attention. However, when the context length L becomes very large (e.g., 128K), the amount of potentially redundant information in the context tends to increase. The redundant context not only hampers the modeling representation performance but also incurs unnecessary computational and storage overhead. In this paper, we propose a plug-and-play Core Context Aware (CCA) Attention for efficient long-context modeling, comprising two complementary modules: 1) Globality-aware pooling module groups input tokens and dynamically compresses each group into one core token based on their significance. In this way, our method automatically focuses and strengthens core context while diminishing redundancy during the learning process, leading to effective long-term dependency modeling. 2) Locality-preserving module incorporates neighboring tokens to preserve local context for detailed representation. Notably, our CCA-Attention is able to replace the self-attention module in existing LLMs with minimal fine-tuning cost. Extensive experimental results show the superiority of our method in both long-context modeling and computational efficiency over state-of-the-art methods.
Authors: Mohsen Sadr, Hossein Gorji
Abstract: Inspired by the Boltzmann kinetics, we propose a collision-based dynamics with a Monte Carlo solution algorithm that approximates the solution of the multi-marginal optimal transport problem via randomized pairwise swapping of sample indices. The computational complexity and memory usage of the proposed method scale linearly with the number of samples, making it highly attractive for high-dimensional settings. In several examples, we demonstrate the efficiency of the proposed method compared to the state-of-the-art methods.
Authors: Ivan Zakazov, Mikolaj Boronski, Lorenzo Drudi, Robert West
Abstract: The ongoing revolution in language modeling has led to various novel applications, some of which rely on the emerging social abilities of large language models (LLMs). Already, many turn to the new cyber friends for advice during the pivotal moments of their lives and trust them with the deepest secrets, implying that accurate shaping of the LLM's personality is paramount. To this end, state-of-the-art approaches exploit a vast variety of training data, and prompt the model to adopt a particular personality. We ask (i) if personality-prompted models behave (i.e., make decisions when presented with a social situation) in line with the ascribed personality (ii) if their behavior can be finely controlled. We use classic psychological experiments, the Milgram experiment and the Ultimatum Game, as social interaction testbeds and apply personality prompting to open- and closed-source LLMs from 4 different vendors. Our experiments reveal failure modes of the prompt-based modulation of the models' behavior that are shared across all models tested and persist under prompt perturbations. These findings challenge the optimistic sentiment toward personality prompting generally held in the community.
Authors: Xiaohao Liu, Xiaobo Xia, Zhuo Huang, See-Kiong Ng, Tat-Seng Chua
Abstract: Multi-modal learning has achieved remarkable success by integrating information from various modalities, achieving superior performance in tasks like recognition and retrieval compared to uni-modal approaches. However, real-world scenarios often present novel modalities that are unseen during training due to resource and privacy constraints, a challenge current methods struggle to address. This paper introduces Modality Generalization (MG), which focuses on enabling models to generalize to unseen modalities. We define two cases: Weak MG, where both seen and unseen modalities can be mapped into a joint embedding space via existing perceptors, and Strong MG, where no such mappings exist. To facilitate progress, we propose a comprehensive benchmark featuring multi-modal algorithms and adapt existing methods that focus on generalization. Extensive experiments highlight the complexity of MG, exposing the limitations of existing methods and identifying key directions for future research. Our work provides a foundation for advancing robust and adaptable multi-modal models, enabling them to handle unseen modalities in realistic scenarios.
Authors: Nadav Z. Cohen, Oron Nir, Ariel Shamir
Abstract: Balancing content fidelity and artistic style is a pivotal challenge in image generation. While traditional style transfer methods and modern Denoising Diffusion Probabilistic Models (DDPMs) strive to achieve this balance, they often struggle to do so without sacrificing either style, content, or sometimes both. This work addresses this challenge by analyzing the ability of DDPMs to maintain content and style equilibrium. We introduce a novel method to identify sensitivities within the DDPM attention layers, identifying specific layers that correspond to different stylistic aspects. By directing conditional inputs only to these sensitive layers, our approach enables fine-grained control over style and content, significantly reducing issues arising from over-constrained inputs. Our findings demonstrate that this method enhances recent stylization techniques by better aligning style and content, ultimately improving the quality of generated visual content.
Authors: Xudong Liao, Yijun Sun, Han Tian, Xinchen Wan, Yilun Jin, Zilong Wang, Zhenghang Ren, Xinyang Huang, Wenxue Li, Kin Fai Tse, Zhizhen Zhong, Guyue Liu, Ying Zhang, Xiaofeng Ye, Yiming Zhang, Kai Chen
Abstract: Mixture-of-Expert (MoE) models outperform conventional models by selectively activating different subnets, named \emph{experts}, on a per-token basis. This gated computation generates dynamic communications that cannot be determined beforehand, challenging the existing GPU interconnects that remain \emph{static} during the distributed training process. In this paper, we advocate for a first-of-its-kind system, called mFabric, that unlocks topology reconfiguration \emph{during} distributed MoE training. Towards this vision, we first perform a production measurement study and show that the MoE dynamic communication pattern has \emph{strong locality}, alleviating the requirement of global reconfiguration. Based on this, we design and implement a \emph{regionally reconfigurable high-bandwidth domain} on top of existing electrical interconnects using optical circuit switching (OCS), achieving scalability while maintaining rapid adaptability. We have built a fully functional mFabric prototype with commodity hardware and a customized collective communication runtime that trains state-of-the-art MoE models with \emph{in-training} topology reconfiguration across 32 A100 GPUs. Large-scale packet-level simulations show that mFabric delivers comparable performance as the non-blocking fat-tree fabric while boosting the training cost efficiency (e.g., performance per dollar) of four representative MoE models by 1.2$\times$--1.5$\times$ and 1.9$\times$--2.3$\times$ at 100 Gbps and 400 Gbps link bandwidths, respectively.
Authors: Zhengyang Tang, Ziniu Li, Zhenyang Xiao, Tian Ding, Ruoyu Sun, Benyou Wang, Dayiheng Liu, Fei Huang, Tianyu Liu, Bowen Yu, Junyang Lin
Abstract: Despite their remarkable performance, Large Language Models (LLMs) face a critical challenge: providing feedback for tasks where human evaluation is difficult or where LLMs potentially outperform humans. In such scenarios, leveraging the critique ability of LLMs themselves - identifying and correcting flaws - shows considerable promise. This paper explores enhancing critique abilities of LLMs, noting that current approaches rely on human annotations or more powerful models, leaving the challenge of improving critique abilities without external supervision unresolved. We introduce SCRIT (Self-evolving CRITic), a framework that trains LLMs with self-generated data to evolve their critique abilities. To address the low quality of naively generated data, we propose a contrastive-critic approach that uses reference solutions during data synthesis to enhance the model's understanding of key concepts, and incorporates a self-validation scheme to ensure data quality. The final trained model operates without any reference solutions at inference time. Implemented with Qwen2.5-72B-Instruct, a leading LLM, SCRIT demonstrates consistent improvements across a wide range of benchmarks spanning both mathematical and scientific reasoning: achieving a 10.0\% relative gain in critique-correction accuracy and a 19.0\% relative improvement in error identification F1-score. Our analysis reveals that SCRIT's performance scales positively with data and model size and enables continuous improvement through multi-round iterations.
Authors: Darin Tsui, Kunal Talreja, Amirali Aghazadeh
Abstract: Computing the Fourier transform of a $q$-ary function $f:\mathbb{Z}_{q}^n\rightarrow \mathbb{R}$, which maps $q$-ary sequences to real numbers, is an important problem in mathematics with wide-ranging applications in biology, signal processing, and machine learning. Previous studies have shown that, under the sparsity assumption, the Fourier transform can be computed efficiently using fast and sample-efficient algorithms. However, in most practical settings, the function is defined over a more general space -- the space of generalized $q$-ary sequences $\mathbb{Z}_{q_1} \times \mathbb{Z}_{q_2} \times \cdots \times \mathbb{Z}_{q_n}$ -- where each $\mathbb{Z}_{q_i}$ corresponds to integers modulo $q_i$. Herein, we develop GFast, a coding theoretic algorithm that computes the $S$-sparse Fourier transform of $f$ with a sample complexity of $O(Sn)$, computational complexity of $O(Sn \log N)$, and a failure probability that approaches zero as $N=\prod_{i=1}^n q_i \rightarrow \infty$ with $S = N^\delta$ for some $0 \leq \delta < 1$. We show that a noise-robust version of GFast computes the transform with a sample complexity of $O(Sn^2)$ and computational complexity of $O(Sn^2 \log N)$ under the same high probability guarantees. Additionally, we demonstrate that GFast computes the sparse Fourier transform of generalized $q$-ary functions $8\times$ faster using $16\times$ fewer samples on synthetic experiments, and enables explaining real-world heart disease diagnosis and protein fitness models using up to $13\times$ fewer samples compared to existing Fourier algorithms applied to the most efficient parameterization of the models as $q$-ary functions.
Authors: Tuan-Dung Bui, Thanh Trong Vu, Thu-Trang Nguyen, Son Nguyen, Hieu Dinh Vo
Abstract: Ensuring the correctness of code generated by Large Language Models (LLMs) presents a significant challenge in AI-driven software development. Existing approaches predominantly rely on black-box (closed-box) approaches that evaluate correctness post-generation, failing to utilize the rich insights embedded in the LLMs' internal states during code generation. In this paper, we introduce OPENIA, a novel white-box (open-box) framework that leverages these internal representations to assess the correctness of LLM-generated code. OPENIA systematically analyzes the intermediate states of representative open-source LLMs specialized for code, including DeepSeek-Coder, CodeLlama, and MagicCoder, across diverse code generation benchmarks. Our empirical analysis reveals that these internal representations encode latent information, which strongly correlates with the correctness of the generated code. Building on these insights, OPENIA uses a white-box/open-box approach to make informed predictions about code correctness, offering significant advantages in adaptability and robustness over traditional classification-based methods and zero-shot approaches. Experimental results demonstrate that OPENIA consistently outperforms baseline models, achieving higher accuracy, precision, recall, and F1-Scores with up to a 2X improvement in standalone code generation and a 46% enhancement in repository-specific scenarios. By unlocking the potential of in-process signals, OPENIA paves the way for more proactive and efficient quality assurance mechanisms in LLM-assisted code generation.
Authors: Mohamad Qadri, Gokul Swamy, Jonathan Francis, Michael Kaess, Andrea Bajcsy
Abstract: Inverse Constraint Learning (ICL) is the problem of inferring constraints from safe (i.e., constraint-satisfying) demonstrations. The hope is that these inferred constraints can then be used downstream to search for safe policies for new tasks and, potentially, under different dynamics. Our paper explores the question of what mathematical entity ICL recovers. Somewhat surprisingly, we show that both in theory and in practice, ICL recovers the set of states where failure is inevitable, rather than the set of states where failure has already happened. In the language of safe control, this means we recover a backwards reachable tube (BRT) rather than a failure set. In contrast to the failure set, the BRT depends on the dynamics of the data collection system. We discuss the implications of the dynamics-conditionedness of the recovered constraint on both the sample-efficiency of policy search and the transferability of learned constraints.
Authors: Wataru Masaka, Mitsuki Sakamoto, Kenshi Abe, Kaito Ariu, Tuomas Sandholm, Atsushi Iwasaki
Abstract: We investigate how perturbation does and does not improve the Follow-the-Regularized-Leader (FTRL) algorithm in solving imperfect-information extensive-form games under sampling, where payoffs are estimated from sampled trajectories. While optimistic algorithms are effective under full feedback, they often become unstable in the presence of sampling noise. Payoff perturbation offers a promising alternative for stabilizing learning and achieving \textit{last-iterate convergence}. We present a unified framework for \textit{Perturbed FTRL} algorithms and study two variants: PFTRL-KL (standard KL divergence) and PFTRL-RKL (Reverse KL divergence), the latter featuring an estimator with both unbiasedness and conditional zero variance. While PFTRL-KL generally achieves equivalent or better performance across benchmark games, PFTRL-RKL consistently outperforms it in Leduc poker, whose structure is more asymmetric than the other games in a sense. Given the modest advantage of PFTRL-RKL, we design the second experiment to isolate the effect of conditional zero variance, showing that the variance-reduction property of RKL improve last-iterate performance.
Authors: Yixuan He, Aaron Sandel, David Wipf, Mihai Cucuringu, John Mitani, Gesine Reinert
Abstract: How can we identify groups of primate individuals which could be conjectured to drive social structure? To address this question, one of us has collected a time series of data for social interactions between chimpanzees. Here we use a network representation, leading to the task of combining these data into a time series of a single weighted network per time stamp, where different proximities should be given different weights reflecting their relative importance. We optimize these proximity-type weights in a principled way, using an innovative loss function which rewards structural consistency for consecutive time steps. The approach is empirically validated by carefully designed synthetic data. Using statistical tests, we provide a way of identifying groups of individuals that stay related for a significant length of time. Applying the approach to the chimpanzee data set, we detect cliques in the animal social network time series, which can be validated by real-world intuition from prior research and qualitative observations by chimpanzee experts.
Authors: Aytijhya Saha, Aaditya Ramdas
Abstract: This paper addresses a fundamental but largely unexplored challenge in sequential changepoint analysis: conducting inference following a detected change. We develop a very general framework to construct confidence sets for the unknown changepoint using only the data observed up to a data-dependent stopping time at which an arbitrary sequential detection algorithm declares a change. Our framework is nonparametric, making no assumption on the composite post-change class, the observation space, or the sequential detection procedure used, and is nonasymptotically valid. We also extend it to handle composite pre-change classes under a suitable assumption, and also derive confidence sets for the change magnitude in parametric settings. Extensive simulations demonstrate that the produced sets have reasonable size, and slightly conservative coverage. In summary, we present the first general method for sequential changepoint localization, which is theoretically sound and broadly applicable in practice.
Authors: Zhichen Dong, Zhanhui Zhou, Zhixuan Liu, Chao Yang, Chaochao Lu
Abstract: In this work, we argue that large language models (LLMs), though trained to predict only the next token, exhibit emergent planning behaviors: $\textbf{their hidden representations encode future outputs beyond the next token}$. Through simple probing, we demonstrate that LLM prompt representations encode global attributes of their entire responses, including $\textit{structure attributes}$ (e.g., response length, reasoning steps), $\textit{content attributes}$ (e.g., character choices in storywriting, multiple-choice answers at the end of response), and $\textit{behavior attributes}$ (e.g., answer confidence, factual consistency). In addition to identifying response planning, we explore how it scales with model size across tasks and how it evolves during generation. The findings that LLMs plan ahead for the future in their hidden representations suggest potential applications for improving transparency and generation control.
Authors: Vittorio Fra
Abstract: Hyperparameter optimization (HPO) is of paramount importance in the development of high-performance, specialized artificial intelligence (AI) models, ranging from well-established machine learning (ML) solutions to the deep learning (DL) domain and the field of spiking neural networks (SNNs). The latter introduce further complexity due to the neuronal computational units and their additional hyperparameters, whose inadequate setting can dramatically impact the final model performance. At the cost of possible reduced generalization capabilities, the most suitable strategy to fully disclose the power of SNNs is to adopt an application-oriented approach and perform extensive HPO experiments. To facilitate these operations, automatic pipelines are fundamental, and their configuration is crucial. In this document, the Neural Network Intelligence (NNI) toolkit is used as reference framework to present one such solution, with a use case example providing evidence of the corresponding results. In addition, a summary of published works employing the presented pipeline is reported as possible source of insights into application-oriented HPO experiments for SNN prototyping.
Authors: Allen M. Wang, Alessandro Pau, Cristina Rea, Oswin So, Charles Dawson, Olivier Sauter, Mark D. Boyer, Anna Vu, Cristian Galperti, Chuchu Fan, Antoine Merle, Yoeri Poels, Cristina Venturini, Stefano Marchioni, the TCV Team
Abstract: The rampdown phase of a tokamak pulse is difficult to simulate and often exacerbates multiple plasma instabilities. To reduce the risk of disrupting operations, we leverage advances in Scientific Machine Learning (SciML) to combine physics with data-driven models, developing a neural state-space model (NSSM) that predicts plasma dynamics during Tokamak \`a Configuration Variable (TCV) rampdowns. The NSSM efficiently learns dynamics from a modest dataset of 311 pulses with only five pulses in a reactor-relevant high-performance regime. The NSSM is parallelized across uncertainties, and reinforcement learning (RL) is applied to design trajectories that avoid instability limits. High-performance experiments at TCV show statistically significant improvements in relevant metrics. A predict-first experiment, increasing plasma current by 20% from baseline, demonstrates the NSSM's ability to make small extrapolations. The developed approach paves the way for designing tokamak controls with robustness to considerable uncertainty and demonstrates the relevance of SciML for fusion experiments.
Authors: Giorgio Franceschelli, Mirco Musolesi
Abstract: Despite the increasing use of large language models for creative tasks, their outputs often lack diversity. Common solutions, such as sampling at higher temperatures, can compromise the quality of the results. Dealing with this trade-off is still an open challenge in designing AI systems for creativity. Drawing on information theory, we propose a context-based score to quantitatively evaluate value and originality. This score incentivizes accuracy and adherence to the request while fostering divergence from the learned distribution. We show that our score can be used as a reward in a reinforcement learning framework to fine-tune large language models for maximum performance. We validate our strategy through experiments considering a variety of creative tasks, such as poetry generation and math problem solving, demonstrating that it enhances the value and originality of the generated solutions.
Authors: Giorgio Franceschelli, Mirco Musolesi
Abstract: Despite their growing capabilities, language models still frequently reproduce content from their training data, generate repetitive text, and favor common grammatical patterns and vocabulary. A possible cause is the decoding strategy: the most common strategies either consider only the most probable tokens, which reduces output diversity, or increase the likelihood of unlikely tokens, compromising output accuracy and correctness. In this paper, we propose DiffSampling, a new decoding method that leverages a mathematical analysis of the token probability distribution to ensure the generation of contextually appropriate text. In particular, the difference between consecutive, sorted probabilities can be used to truncate incorrect tokens. In addition, we also propose two variations of the proposed method that aim to correct the subtle inconsistencies of common sampling strategies. Experiments involving four different text-generation tasks demonstrate that our approach consistently performs at least on par with the existing methods it builds upon in terms of quality, while potentially improving output diversity.
Authors: Kaiwen Yan, Hongcheng Guo, Xuanqing Shi, Shaosheng Cao, Donglin Di, Zhoujun Li
Abstract: With the rapid advancement of Large Language Models (LLMs), the demand for robust instruction-following capabilities in code generation tasks has grown significantly. Code generation not only facilitates faster prototyping and automated testing, but also augments developer efficiency through improved maintainability and reusability of code. In this paper, we introduce CodeIF, the first benchmark specifically designed to assess the abilities of LLMs to adhere to task-oriented instructions within diverse code generation scenarios. CodeIF encompasses a broad range of tasks, including function synthesis, error debugging, algorithmic refactoring, and code explanation, thereby providing a comprehensive suite to evaluate model performance across varying complexity levels and programming domains. We conduct extensive experiments with LLMs, analyzing their strengths and limitations in meeting the demands of these tasks. The experimental results offer valuable insights into how well current models align with human instructions, as well as the extent to which they can generate consistent, maintainable, and contextually relevant code. Our findings not only underscore the critical role that instruction-following LLMs can play in modern software development, but also illuminate pathways for future research aimed at enhancing their adaptability, reliability, and overall effectiveness in automated code generation. CodeIF data and code are publicly available: https://github.com/lin-rany/codeIF
Authors: Sakshi Arya, Hyebin Song
Abstract: The multi-armed bandits (MAB) framework is a widely used approach for sequential decision-making, where a decision-maker selects an arm in each round with the goal of maximizing long-term rewards. Moreover, in many practical applications, such as personalized medicine and recommendation systems, feedback is provided in batches, contextual information is available at the time of decision-making, and rewards from different arms are related rather than independent. We propose a novel semi-parametric framework for batched bandits with covariates and a shared parameter across arms, leveraging the single-index regression (SIR) model to capture relationships between arm rewards while balancing interpretability and flexibility. Our algorithm, Batched single-Index Dynamic binning and Successive arm elimination (BIDS), employs a batched successive arm elimination strategy with a dynamic binning mechanism guided by the single-index direction. We consider two settings: one where a pilot direction is available and another where the direction is estimated from data, deriving theoretical regret bounds for both cases. When a pilot direction is available with sufficient accuracy, our approach achieves minimax-optimal rates (with $d = 1$) for nonparametric batched bandits, circumventing the curse of dimensionality. Extensive experiments on simulated and real-world datasets demonstrate the effectiveness of our algorithm compared to the nonparametric batched bandit method introduced by \cite{jiang2024batched}.
Authors: Tobias J\"ulg, Wolfram Burgard, Florian Walter
Abstract: Vision-Language-Action Models (VLAs) have demonstrated remarkable generalization capabilities in real-world experiments. However, their success rates are often not on par with expert policies, and they require fine-tuning when the setup changes. In this work, we introduce Refined Policy Distillation (RPD), a novel Reinforcement Learning (RL)-based policy refinement method that bridges this performance gap through a combination of on-policy RL with behavioral cloning. The core idea of RPD is to distill and refine VLAs into compact, high-performing expert policies by guiding the student policy during RL exploration using the actions of a teacher VLA, resulting in increased sample efficiency and faster convergence. We complement our method by fine-tuned versions of Octo and OpenVLA for ManiSkill3 to evaluate RPD in simulation. While this is a key requirement for applying RL, it also yields new insights beyond existing studies on VLA performance in real-world settings. Our experimental results across various manipulation tasks show that RPD enables the RL student to learn expert policies that outperform the VLA teacher in both dense and sparse reward settings, while also achieving faster convergence than the RL baseline. Our approach is even robust to changes in camera perspective and can generalize to task variations that the underlying VLA cannot solve. Our code, dataset, VLA checkpoints, and videos are available at https://refined-policy-distillation.github.io
Authors: AgiBot-World-Contributors, Qingwen Bu, Jisong Cai, Li Chen, Xiuqi Cui, Yan Ding, Siyuan Feng, Shenyuan Gao, Xindong He, Xuan Hu, Xu Huang, Shu Jiang, Yuxin Jiang, Cheng Jing, Hongyang Li, Jialu Li, Chiming Liu, Yi Liu, Yuxiang Lu, Jianlan Luo, Ping Luo, Yao Mu, Yuehan Niu, Yixuan Pan, Jiangmiao Pang, Yu Qiao, Guanghui Ren, Cheng Ruan, Jiaqi Shan, Yongjian Shen, Chengshi Shi, Mingkang Shi, Modi Shi, Chonghao Sima, Jianheng Song, Huijie Wang, Wenhao Wang, Dafeng Wei, Chengen Xie, Guo Xu, Junchi Yan, Cunbiao Yang, Lei Yang, Shukai Yang, Maoqing Yao, Jia Zeng, Chi Zhang, Qinglin Zhang, Bin Zhao, Chengyue Zhao, Jiaqi Zhao, Jianchao Zhu
Abstract: We explore how scalable robot data can address real-world challenges for generalized robotic manipulation. Introducing AgiBot World, a large-scale platform comprising over 1 million trajectories across 217 tasks in five deployment scenarios, we achieve an order-of-magnitude increase in data scale compared to existing datasets. Accelerated by a standardized collection pipeline with human-in-the-loop verification, AgiBot World guarantees high-quality and diverse data distribution. It is extensible from grippers to dexterous hands and visuo-tactile sensors for fine-grained skill acquisition. Building on top of data, we introduce Genie Operator-1 (GO-1), a novel generalist policy that leverages latent action representations to maximize data utilization, demonstrating predictable performance scaling with increased data volume. Policies pre-trained on our dataset achieve an average performance improvement of 30% over those trained on Open X-Embodiment, both in in-domain and out-of-distribution scenarios. GO-1 exhibits exceptional capability in real-world dexterous and long-horizon tasks, achieving over 60% success rate on complex tasks and outperforming prior RDT approach by 32%. By open-sourcing the dataset, tools, and models, we aim to democratize access to large-scale, high-quality robot data, advancing the pursuit of scalable and general-purpose intelligence.
Authors: Dikai Liu, Tianwei Zhang, Jianxiong Yin, Simon See
Abstract: Quadrupeds have gained rapid advancement in their capability of traversing across complex terrains. The adoption of deep Reinforcement Learning (RL), transformers and various knowledge transfer techniques can greatly reduce the sim-to-real gap. However, the classical teacher-student framework commonly used in existing locomotion policies requires a pre-trained teacher and leverages the privilege information to guide the student policy. With the implementation of large-scale models in robotics controllers, especially transformers-based ones, this knowledge distillation technique starts to show its weakness in efficiency, due to the requirement of multiple supervised stages. In this paper, we propose Unified Locomotion Transformer (ULT), a new transformer-based framework to unify the processes of knowledge transfer and policy optimization in a single network while still taking advantage of privilege information. The policies are optimized with reinforcement learning, next state-action prediction, and action imitation, all in just one training stage, to achieve zero-shot deployment. Evaluation results demonstrate that with ULT, optimal teacher and student policies can be obtained at the same time, greatly easing the difficulty in knowledge transfer, even with complex transformer-based models.
Authors: Jingyi Zhang, Jiaxing Huang, Huanjin Yao, Shunyu Liu, Xikun Zhang, Shijian Lu, Dacheng Tao
Abstract: Recent studies generally enhance MLLMs' reasoning capabilities via supervised fine-tuning on high-quality chain-of-thought reasoning data, which often leads models to merely imitate successful reasoning paths without understanding what the wrong reasoning paths are. In this work, we aim to enhance the MLLMs' reasoning ability beyond passively imitating positive reasoning paths. To this end, we design Step-wise Group Relative Policy Optimization (StepGRPO), a new online reinforcement learning framework that enables MLLMs to self-improve reasoning ability via simple, effective and dense step-wise rewarding. Specifically, StepGRPO introduces two novel rule-based reasoning rewards: Step-wise Reasoning Accuracy Reward (StepRAR) and Step-wise Reasoning Validity Reward (StepRVR). StepRAR rewards the reasoning paths that contain necessary intermediate reasoning steps via a soft key-step matching technique, while StepRAR rewards reasoning paths that follow a well-structured and logically consistent reasoning process through a reasoning completeness and logic evaluation strategy. With the proposed StepGRPO, we introduce R1-VL, a series of MLLMs with outstanding capabilities in step-by-step reasoning. Extensive experiments over 8 benchmarks demonstrate the superiority of our methods.
Authors: George Panagopoulos, Johannes F. Lutzeyer, Sofiane Ennadir, Jun Pang
Abstract: Genomic studies, including CRISPR-based Perturb-seq analyses, face a vast hypothesis space, while gene perturbations remain costly and time-consuming. Gene perturbation models based on graph neural networks are trained to predict the outcomes of gene perturbations to facilitate such experiments. Due to the cost of genomic experiments, active learning is often employed to train these models, alternating between wet-lab experiments and model updates. However, the operational constraints of the wet-lab and the iterative nature of active learning significantly increase the total training time. Furthermore, the inherent sensitivity to model initialization can lead to markedly different sets of gene perturbations across runs, which undermines the reproducibility, interpretability, and reusability of the method. To this end, we propose a graph-based data filtering method that, unlike active learning, selects the gene perturbations in one shot and in a model-free manner. The method optimizes a criterion that maximizes the supervision signal from the graph neural network to enhance generalization. The criterion is defined over the input graph and is optimized with submodular maximization. We compare it empirically to active learning, and the results demonstrate that despite yielding months of acceleration, it also improves the stability of the selected perturbation experiments while achieving comparable test error.
Authors: Liming Zheng, Feng Yan, Fanfan Liu, Chengjian Feng, Yufeng Zhong, Lin Ma
Abstract: The growing adoption of Vision-Language-Action (VLA) models in embodied AI intensifies the demand for diverse manipulation demonstrations. However, high costs associated with data collection often result in insufficient data coverage across all scenarios, which limits the performance of the models. It is observed that the spatial reasoning phase (SRP) in large workspace dominates the failure cases. Fortunately, this data can be collected with low cost, underscoring the potential of leveraging inexpensive data to improve model performance. In this paper, we introduce the RoboTron-Craft, a stage-divided and cost-effective pipeline for realistic manipulation generation. Base on this, the RoboTron-Platter method is introduced, a framework that decouples training trajectories into distinct task stages and leverages abundant easily collectible SRP data to enhance VLA model's generalization. Through analysis we demonstrate that sub-task-specific training with additional SRP data with proper proportion can act as a performance catalyst for robot manipulation, maximizing the utilization of costly physical interaction phase (PIP) data. Experiments show that through introducing large proportion of cost-effective SRP trajectories into a limited set of PIP data, we can achieve a maximum improvement of 41\% on success rate in zero-shot scenes, while with the ability to transfer manipulation skill to novel targets. Project available at https://github.com/ notFoundThisPerson/RoboTron-Craft.
URLs: https://github.com/
Authors: Dianxin Luan, John Thompson
Abstract: Optical Wireless Communication (OWC) has gained significant attention due to its high-speed data transmission and throughput. Optical wireless channels are often assumed to be flat, but we evaluate frequency selective channels to consider high data rate optical wireless or very dispersive environments. To address this for optical scenarios, this paper presents a robust channel estimation framework with low-complexity to mitigate frequency-selective effects, then to improve system reliability and performance. This channel estimation framework contains a neural network that can estimate general optical wireless channels without prior channel information about the environment. Based on this estimate and the corresponding delay spread, one of several candidate offline-trained neural networks will be activated to predict this channel. Simulation results demonstrate that the proposed method has improved and robust normalized mean square error (NMSE) and bit error rate (BER) performance compared to conventional estimation methods while maintaining computational efficiency. These findings highlight the potential of neural network solutions in enhancing the performance of OWC systems under indoor channel conditions.
Authors: Anja Surina, Amin Mansouri, Lars Quaedvlieg, Amal Seddas, Maryna Viazovska, Emmanuel Abbe, Caglar Gulcehre
Abstract: Discovering efficient algorithms for solving complex problems has been an outstanding challenge in mathematics and computer science, requiring substantial human expertise over the years. Recent advancements in evolutionary search with large language models (LLMs) have shown promise in accelerating the discovery of algorithms across various domains, particularly in mathematics and optimization. However, existing approaches treat the LLM as a static generator, missing the opportunity to update the model with the signal obtained from evolutionary exploration. In this work, we propose to augment LLM-based evolutionary search by continuously refining the search operator - the LLM - through reinforcement learning (RL) fine-tuning. Our method leverages evolutionary search as an exploration strategy to discover improved algorithms, while RL optimizes the LLM policy based on these discoveries. Our experiments on combinatorial optimization tasks demonstrate that integrating RL with evolutionary search accelerates the discovery of superior algorithms, showcasing the potential of RL-enhanced evolutionary strategies for algorithm design.
Authors: Taibiao Zhao, Xiaobing Chen, Mingxuan Sun
Abstract: The adaptation of large language models (LLMs) to time series forecasting poses unique challenges, as time series data is continuous in nature, while LLMs operate on discrete tokens. Despite the success of LLMs in natural language processing (NLP) and other structured domains, aligning time series data with language-based representations while maintaining both predictive accuracy and interpretability remains a significant hurdle. Existing methods have attempted to reprogram time series data into text-based forms, but these often fall short in delivering meaningful, interpretable results. In this paper, we propose a multi-level text alignment framework for time series forecasting using LLMs that not only improves prediction accuracy but also enhances the interpretability of time series representations. Our method decomposes time series into trend, seasonal, and residual components, which are then reprogrammed into component-specific text representations. We introduce a multi-level alignment mechanism, where component-specific embeddings are aligned with pre-trained word tokens, enabling more interpretable forecasts. Experiments on multiple datasets demonstrate that our method outperforms state-of-the-art models in accuracy while providing good interpretability.
Authors: Takahiro Misawa, Ai Koizumi, Ryo Tamura, Kazuyoshi Yoshimi
Abstract: Generative AI has recently had a profound impact on various fields, including daily life, research, and education. To explore its efficient utilization in data-driven materials science, we organized a hackathon -- AIMHack2024 -- in July 2024. In this hackathon, researchers from fields such as materials science, information science, bioinformatics, and condensed matter physics worked together to explore how generative AI can facilitate research and education. Based on the results of the hackathon, this paper presents topics related to (1) conducting AI-assisted software trials, (2) building AI tutors for software, and (3) developing GUI applications for software. While generative AI continues to evolve rapidly, this paper provides an early record of its application in data-driven materials science and highlights strategies for integrating AI into research and education.
Authors: Maria-Florina Balcan, Kiriaki Fragkia, Keegan Harris
Abstract: We study structured Stackelberg games, in which both players (the leader and the follower) observe contextual information about the state of the world at time of play. The leader plays against one of a finite number of followers, but the follower's type is not known until after the game has ended. Importantly, we assume a fixed relationship between the contextual information and the follower's type, thereby allowing the leader to leverage this additional structure when deciding her strategy. Under this setting, we find that standard learning theoretic measures of complexity do not characterize the difficulty of the leader's learning task. Instead, we introduce a new notion of dimension, the Stackelberg-Littlestone dimension, which we show characterizes the instance-optimal regret of the leader in the online setting. Based on this, we also provide a provably optimal learning algorithm. We extend our results to the distributional setting, where we use two new notions of dimension, the $\gamma$-Stackelberg-Natarajan dimension and $\gamma$-Stackelberg-Graph dimension. We prove that these control the sample complexity lower and upper bounds respectively, and we design a simple, improper algorithm that achieves the upper bound.
Authors: Nusrat Zahan, Imranur Rahman, Laurie Williams
Abstract: Practitioners often struggle with the overwhelming number of security practices outlined in cybersecurity frameworks for risk mitigation. Given the limited budget, time, and resources, practitioners want to prioritize the adoption of security practices based on empirical evidence. The goal of this study is to assist practitioners and policymakers in making informed decisions on which security practices to adopt by evaluating the relationship between software security practices adoption and security outcome metrics. To do this, we analyzed the adoption of security practices and their impact on security outcome metrics across 145K npm packages. We selected the OpenSSF Scorecard metrics to automatically measure the adoption of security practices in npm GitHub repositories. We also investigated project-level security outcome metrics: the number of open vulnerabilities (Vul_Count)), mean time to remediate (MTTR) vulnerabilities in dependencies, and mean time to update (MTTU) dependencies. We conducted regression and causal analysis using 11 Scorecard metrics and the aggregated Scorecard score (computed by aggregating individual security practice scores) as predictors and Vul_Count), MTTR, and MTTU as target variables. Our findings reveal that aggregated adoption of security practices is associated with 5.2 fewer vulnerabilities, 216.8 days faster MTTR, and 52.3 days faster MTTU. Repository characteristics have an impact on security practice effectiveness: repositories with high security practice adoptions, especially those that are mature, actively maintained, large in size, have many contributors, few dependencies, and high download volumes, tend to exhibit better outcomes compared to smaller or inactive repositories.
Authors: An T. Le, Khai Nguyen, Minh Nhat Vu, Jo\~ao Carvalho, Jan Peters
Abstract: Sampling-based model predictive control (MPC) offers strong performance in nonlinear and contact-rich robotic tasks, yet often suffers from poor exploration due to locally greedy sampling schemes. We propose \emph{Model Tensor Planning} (MTP), a novel sampling-based MPC framework that introduces high-entropy control trajectory generation through structured tensor sampling. By sampling over randomized multipartite graphs and interpolating control trajectories with B-splines and Akima splines, MTP ensures smooth and globally diverse control candidates. We further propose a simple $\beta$-mixing strategy that blends local exploitative and global exploratory samples within the modified Cross-Entropy Method (CEM) update, balancing control refinement and exploration. Theoretically, we show that MTP achieves asymptotic path coverage and maximum entropy in the control trajectory space in the limit of infinite tensor depth and width. Our implementation is fully vectorized using JAX and compatible with MuJoCo XLA, supporting \emph{Just-in-time} (JIT) compilation and batched rollouts for real-time control with online domain randomization. Through experiments on various challenging robotic tasks, ranging from dexterous in-hand manipulation to humanoid locomotion, we demonstrate that MTP outperforms standard MPC and evolutionary strategy baselines in task success and control robustness. Design and sensitivity ablations confirm the effectiveness of MTP tensor sampling structure, spline interpolation choices, and mixing strategy. Altogether, MTP offers a scalable framework for robust exploration in model-based planning and control.
Authors: Yang Lyu, Tan Minh Nguyen, Yuchun Qian, Xin T. Tong
Abstract: Diffusion models are popular tools for generating new data samples, using a forward process that adds noise to data and a reverse process to denoise and produce samples. However, when the data distribution consists of n points, empirical diffusion models tend to reproduce existing data points, a phenomenon known as the memorization effect. Current literature often addresses this with complex machine learning techniques. This work shows that the memorization issue can be solved simply by applying an inertia update at the end of the empirical diffusion simulation. Our inertial diffusion model requires only the empirical score function and no additional training. We demonstrate that the distribution of samples from this model approximates the true data distribution on a $C^2$ manifold of dimension $d$, within a Wasserstein-1 distance of order $O(n^{-\frac{2}{d+4}})$. This bound significantly shrinks the Wasserstein distance between the population and empirical distributions, confirming that the inertial diffusion model produces new and diverse samples. Remarkably, this estimate is independent of the ambient space dimension, as no further training is needed. Our analysis shows that the inertial diffusion samples resemble Gaussian kernel density estimations on the manifold, revealing a novel connection between diffusion models and manifold learning.
Authors: Jack O'Hagan, Andrew Keane, Andrew Flynn
Abstract: Artificial Intelligence has advanced significantly in recent years thanks to innovations in the design and training of artificial neural networks (ANNs). Despite these advancements, we still understand relatively little about how elementary forms of ANNs learn, fail to learn, and generate false information without the intent to deceive, a phenomenon known as `confabulation'. To provide some foundational insight, in this paper we analyse how confabulation occurs in reservoir computers (RCs): a dynamical system in the form of an ANN. RCs are particularly useful to study as they are known to confabulate in a well-defined way: when RCs are trained to reconstruct the dynamics of a given attractor, they sometimes construct an attractor that they were not trained to construct, a so-called `untrained attractor' (UA). This paper sheds light on the role played by UAs when reconstruction fails and their influence when modelling transitions between reconstructed attractors. Based on our results, we conclude that UAs are an intrinsic feature of learning systems whose state spaces are bounded, and that this means of confabulation may be present in systems beyond RCs.
Authors: Jaehyun Jeon, Min Soo Kim, Jang Han Yoon, Sumin Shim, Yejin Choi, Hanbin Kim, Youngjae Yu
Abstract: User interface (UI) design goes beyond visuals, guiding user behavior and overall user experience (UX). Strategically crafted interfaces, for example, can boost sign-ups and drive business sales, underscoring the shift toward UI/UX as a unified design concept. While recent studies have explored UI quality evaluation using Multimodal Large Language Models (MLLMs), they largely focus on surface-level features, overlooking behavior-oriented aspects. To fill this gap, we introduce WiserUI-Bench, a novel benchmark for assessing models' multimodal understanding of UI/UX design. It includes 300 diverse real-world UI image pairs, each consisting of two design variants A/B-tested at scale by actual companies, where one was empirically validated to steer more user actions than the other. Each pair is accompanied one or more of 684 expert-curated rationales that capture key factors behind each winning design's effectiveness, spanning diverse cognitive dimensions of UX. Our benchmark supports two core tasks: (1) selecting the more effective UI/UX design by predicting the A/B test verified winner and (2) assessing how well a model, given the winner, can explain its effectiveness in alignment with expert reasoning. Experiments across several MLLMs show that current models exhibit limited nuanced reasoning about UI/UX design and its behavioral impact. We believe our work will foster research in UI/UX understanding and enable broader applications such as behavior-aware interface optimization.
Authors: Yexiang Liu, Zekun Li, Zhi Fang, Nan Xu, Ran He, Tieniu Tan
Abstract: Recently, scaling test-time compute on Large Language Models (LLM) has garnered wide attention. However, there has been limited investigation of how various reasoning prompting strategies perform as scaling. In this paper, we focus on a standard and realistic scaling setting: majority voting. We systematically conduct experiments on 6 LLMs $\times$ 8 prompting strategies $\times$ 6 benchmarks. Experiment results consistently show that as the sampling time and computational overhead increase, complicated prompting strategies with superior initial performance gradually fall behind simple Chain-of-Thought. We analyze this phenomenon and provide theoretical proofs. Additionally, we propose a probabilistic method to efficiently predict scaling performance and identify the best prompting strategy under large sampling times, eliminating the need for resource-intensive inference processes in practical applications. Furthermore, we introduce two ways derived from our theoretical analysis to significantly improve the scaling performance. We hope that our research can promote to re-examine the role of complicated prompting, unleash the potential of simple prompting strategies, and provide new insights for enhancing test-time scaling performance. Code is available at https://github.com/MraDonkey/rethinking_prompting.
Authors: Andrey Alexandrov, Giovanni Acampora, Giovanni De Lellis, Antonia Di Crescenzo, Chiara Errico, Daria Morozova, Valeri Tioukov, Autilia Vittiello
Abstract: Accurately tracking particles and determining their coordinate along the optical axis is a major challenge in optical microscopy, especially when extremely high precision is needed. In this study, we introduce a deep learning approach using convolutional neural networks (CNNs) that can determine axial coordinates from dual-focal-plane images without relying on predefined models. Our method achieves an axial localization precision of 40 nanometers-six times better than traditional single-focal-plane techniques. The model's simple design and strong performance make it suitable for a wide range of uses, including dark matter detection, proton therapy for cancer, and radiation protection in space. It also shows promise in fields like biological imaging, materials science, and environmental monitoring. This work highlights how machine learning can turn complex image data into reliable, precise information, offering a flexible and powerful tool for many scientific applications.
Authors: Tim Tsz-Kit Lau, Qi Long, Weijie Su
Abstract: The ever-growing scale of deep learning models and datasets underscores the critical importance of efficient optimization methods. While preconditioned gradient methods such as Adam and AdamW are the de facto optimizers for training neural networks and large language models, structure-aware preconditioned optimizers like Shampoo and Muon, which utilize the matrix structure of gradients, have demonstrated promising evidence of faster convergence. In this paper, we introduce a unifying framework for analyzing "matrix-aware" preconditioned methods, which not only sheds light on the effectiveness of Muon and related optimizers but also leads to a class of new structure-aware preconditioned methods. A key contribution of this framework is its precise distinction between preconditioning strategies that treat neural network weights as vectors (addressing curvature anisotropy) versus those that consider their matrix structure (addressing gradient anisotropy). This perspective provides new insights into several empirical phenomena in language model pre-training, including Adam's training instabilities, Muon's accelerated convergence, and the necessity of learning rate warmup for Adam. Building upon this framework, we introduce PolarGrad, a new class of preconditioned optimization methods based on the polar decomposition of matrix-valued gradients. As a special instance, PolarGrad includes Muon with updates scaled by the nuclear norm of the gradients. We provide numerical implementations of these methods, leveraging efficient numerical polar decomposition algorithms for enhanced convergence. Our extensive evaluations across diverse matrix optimization problems and language model pre-training tasks demonstrate that PolarGrad outperforms both Adam and Muon.
Authors: Junyan Liu, Arnab Maiti, Artin Tajdini, Kevin Jamieson, Lillian J. Ratliff
Abstract: We initiate the study of a repeated principal-agent problem over a finite horizon $T$, where a principal sequentially interacts with $K\geq 2$ types of agents arriving in an adversarial order. At each round, the principal strategically chooses one of the $N$ arms to incentivize for an arriving agent of unknown type. The agent then chooses an arm based on its own utility and the provided incentive, and the principal receives a corresponding reward. The objective is to minimize regret against the best incentive in hindsight. Without prior knowledge of agent behavior, we show that the problem becomes intractable, leading to linear regret. We analyze two key settings where sublinear regret is achievable. In the first setting, the principal knows the arm each agent type would select greedily for any given incentive. Under this setting, we propose an algorithm that achieves a regret bound of $O(\min\{\sqrt{KT\log N},K\sqrt{T}\})$ and provide a matching lower bound up to a $\log K$ factor. In the second setting, an agent's response varies smoothly with the incentive and is governed by a Lipschitz constant $L\geq 1$. Under this setting, we show that there is an algorithm with a regret bound of $\tilde{O}((LN)^{1/3}T^{2/3})$ and establish a matching lower bound up to logarithmic factors. Finally, we extend our algorithmic results for both settings by allowing the principal to incentivize multiple arms simultaneously in each round.
Authors: Xinxu Wei, Kanhao Zhao, Yong Jiao, Lifang He, Yu Zhang
Abstract: As large language models (LLMs) continue to revolutionize AI research, there is a growing interest in building large-scale brain foundation models to advance neuroscience. While most existing brain foundation models are pre-trained on time-series signals or connectome features, we propose a novel graph-based pre-training paradigm for constructing a brain graph foundation model. In this paper, we introduce the Brain Graph Foundation Model, termed BrainGFM, a unified framework that leverages graph contrastive learning and graph masked autoencoders for large-scale fMRI-based pre-training. BrainGFM is pre-trained on a diverse mixture of brain atlases with varying parcellations, significantly expanding the pre-training corpus and enhancing the model's ability to generalize across heterogeneous fMRI-derived brain representations. To support efficient and versatile downstream transfer, we integrate both graph prompts and language prompts into the model design, enabling BrainGFM to flexibly adapt to a wide range of atlases, neurological and psychiatric disorders, and task settings. Furthermore, we employ meta-learning to optimize the graph prompts, facilitating strong generalization to previously unseen disorders under both few-shot and zero-shot learning conditions via language-guided prompting. BrainGFM is pre-trained on 27 neuroimaging datasets spanning 25 common neurological and psychiatric disorders, encompassing 2 types of brain atlases (functional and anatomical) across 8 widely-used parcellations, and covering over 25,000 subjects, 60,000 fMRI scans, and a total of 400,000 graph samples aggregated across all atlases and parcellations. The code is available at: https://github.com/weixinxu666/BrainGFM
Authors: Ziv Scully, Alexander Terenin
Abstract: The Gittins index is a tool that optimally solves a variety of decision-making problems involving uncertainty, including multi-armed bandit problems, minimizing mean latency in queues, and search problems like the Pandora's box model. However, despite the above examples and later extensions thereof, the space of problems that the Gittins index can solve perfectly optimally is limited, and its definition is rather subtle compared to those of other multi-armed bandit algorithms. As a result, the Gittins index is often regarded as being primarily a concept of theoretical importance, rather than a practical tool for solving decision-making problems. The aim of this tutorial is to demonstrate that the Gittins index can be fruitfully applied to practical problems. We start by giving an example-driven introduction to the Gittins index, then walk through several examples of problems it solves - some optimally, some suboptimally but still with excellent performance. Two practical highlights in the latter category are applying the Gittins index to Bayesian optimization, and applying the Gittins index to minimizing tail latency in queues.
Authors: Jaehong Chung, Michael Manga, Timothy Kneafsey, Tapan Mukerji, Mengsu Hu
Abstract: Microearthquakes (MEQs) generated by subsurface fluid injection record the evolving stress state and permeability of reservoirs. Forecasting their full spatiotemporal evolution is therefore critical for applications such as enhanced geothermal systems (EGS), CO$_2$ sequestration and other geo-engineering applications. We present a transformer-based deep learning model that ingests hydraulic stimulation history and prior MEQ observations to forecast four key quantities: cumulative MEQ count, cumulative logarithmic seismic moment, and the 50th- and 95th-percentile extents ($P_{50}, P_{95}$) of the MEQ cloud. Applied to the EGS Collab Experiment 1 dataset, the model achieves $R^2 >0.98$ for the 1-second forecast horizon and $R^2 >0.88$ for the 15-second forecast horizon across all targets, and supplies uncertainty estimates through a learned standard deviation term. These accurate, uncertainty-quantified forecasts enable real-time inference of fracture propagation and permeability evolution, demonstrating the strong potential of deep-learning approaches to improve seismic-risk assessment and guide mitigation strategies in future fluid-injection operations.
Authors: Ryan Thompson, Matt P. Wand, Joanna J. J. Wang
Abstract: Linear mixed models (LMMs), which incorporate fixed and random effects, are key tools for analyzing heterogeneous data, such as in personalized medicine. Nowadays, this type of data is increasingly wide, sometimes containing thousands of candidate predictors, necessitating sparsity for prediction and interpretation. However, existing sparse learning methods for LMMs do not scale well beyond tens or hundreds of predictors, leaving a large gap compared with sparse methods for linear models, which ignore random effects. This paper closes the gap with a new $\ell_0$ regularized method for LMM subset selection that can run on datasets containing thousands of predictors in seconds to minutes. On the computational front, we develop a coordinate descent algorithm as our main workhorse and provide a guarantee of its convergence. We also develop a local search algorithm to help traverse the nonconvex optimization surface. Both algorithms readily extend to subset selection in generalized LMMs via a penalized quasi-likelihood approximation. On the statistical front, we provide a finite-sample bound on the Kullback-Leibler divergence of the new method. We then demonstrate its excellent performance in experiments involving synthetic and real datasets.
Authors: Guan Wang, Jin Li, Yuhao Sun, Xing Chen, Changling Liu, Yue Wu, Meng Lu, Sen Song, Yasin Abbasi Yadkori
Abstract: Reasoning, the process of devising and executing complex goal-oriented action sequences, remains a critical challenge in AI. Current large language models (LLMs) primarily employ Chain-of-Thought (CoT) techniques, which suffer from brittle task decomposition, extensive data requirements, and high latency. Inspired by the hierarchical and multi-timescale processing in the human brain, we propose the Hierarchical Reasoning Model (HRM), a novel recurrent architecture that attains significant computational depth while maintaining both training stability and efficiency. HRM executes sequential reasoning tasks in a single forward pass without explicit supervision of the intermediate process, through two interdependent recurrent modules: a high-level module responsible for slow, abstract planning, and a low-level module handling rapid, detailed computations. With only 27 million parameters, HRM achieves exceptional performance on complex reasoning tasks using only 1000 training samples. The model operates without pre-training or CoT data, yet achieves nearly perfect performance on challenging tasks including complex Sudoku puzzles and optimal path finding in large mazes. Furthermore, HRM outperforms much larger models with significantly longer context windows on the Abstraction and Reasoning Corpus (ARC), a key benchmark for measuring artificial general intelligence capabilities. These results underscore HRM's potential as a transformative advancement toward universal computation and general-purpose reasoning systems.
Authors: Amir Aghdam, Vincent Tao Hu, Bj\"orn Ommer
Abstract: We address the task of zero-shot video classification for extremely fine-grained actions (e.g., Windmill Dunk in basketball), where no video examples or temporal annotations are available for unseen classes. While image-language models (e.g., CLIP, SigLIP) show strong open-set recognition, they lack temporal modeling needed for video understanding. We propose ActAlign, a truly zero-shot, training-free method that formulates video classification as a sequence alignment problem, preserving the generalization strength of pretrained image-language models. For each class, a large language model (LLM) generates an ordered sequence of sub-actions, which we align with video frames using Dynamic Time Warping (DTW) in a shared embedding space. Without any video-text supervision or fine-tuning, ActAlign achieves 30.5% accuracy on ActionAtlas--the most diverse benchmark of fine-grained actions across multiple sports--where human performance is only 61.6%. ActAlign outperforms billion-parameter video-language models while using 8x fewer parameters. Our approach is model-agnostic and domain-general, demonstrating that structured language priors combined with classical alignment methods can unlock the open-set recognition potential of image-language models for fine-grained video understanding.
Authors: Yifan Liu, Yu Fang, Zhouhan Lin
Abstract: Conventional voice conversion modifies voice characteristics from a source speaker to a target speaker, relying on audio input from both sides. However, this process becomes infeasible when clean audio is unavailable, such as in silent videos or noisy environments. In this work, we focus on the task of Silent Face-based Voice Conversion (SFVC), which does voice conversion entirely from visual inputs. i.e., given images of a target speaker and a silent video of a source speaker containing lip motion, SFVC generates speech aligning the identity of the target speaker while preserving the speech content in the source silent video. As this task requires generating intelligible speech and converting identity using only visual cues, it is particularly challenging. To address this, we introduce MuteSwap, a novel framework that employs contrastive learning to align cross-modality identities and minimize mutual information to separate shared visual features. Experimental results show that MuteSwap achieves impressive performance in both speech synthesis and identity conversion, especially under noisy conditions where methods dependent on audio input fail to produce intelligible results, demonstrating both the effectiveness of our training approach and the feasibility of SFVC.
Authors: Quentin Le Roux, Yannick Teglia, Teddy Furon, Philippe Loubet-Moundi, Eric Bourbao
Abstract: The widespread use of deep learning face recognition raises several security concerns. Although prior works point at existing vulnerabilities, DNN backdoor attacks against real-life, unconstrained systems dealing with images captured in the wild remain a blind spot of the literature. This paper conducts the first system-level study of backdoors in deep learning-based face recognition systems. This paper yields four contributions by exploring the feasibility of DNN backdoors on these pipelines in a holistic fashion. We demonstrate for the first time two backdoor attacks on the face detection task: face generation and face landmark shift attacks. We then show that face feature extractors trained with large margin losses also fall victim to backdoor attacks. Combining our models, we then show using 20 possible pipeline configurations and 15 attack cases that a single backdoor enables an attacker to bypass the entire function of a system. Finally, we provide stakeholders with several best practices and countermeasures.
Authors: Ashutosh Hathidara, Julien Yu, Sebastian Schreiber
Abstract: Large language models (LLMs) are increasingly tasked with invoking enterprise APIs, yet they routinely falter when near-duplicate tools vie for the same user intent or when required arguments are left underspecified. We introduce DiaFORGE (Dialogue Framework for Organic Response Generation & Evaluation), a disambiguation-centric, three-stage pipeline that (i) synthesizes persona-driven, multi-turn dialogues in which the assistant must distinguish among highly similar tools, (ii) performs supervised fine-tuning of open-source models with reasoning traces across 3B - 70B parameters, and (iii) evaluates real-world readiness via a dynamic suite that redeploys each model in a live agentic loop and reports end-to-end goal completion alongside conventional static metrics. On our dynamic benchmark DiaBENCH, models trained with DiaFORGE raise tool-invocation success by 27 pp over GPT-4o and by 49 pp over Claude-3.5-Sonnet, both under optimized prompting. To spur further research, we release an open corpus of 5000 production-grade enterprise API specifications paired with rigorously validated, disambiguation-focused dialogues, offering a practical blueprint for building reliable, enterprise-ready tool-calling agents.
Authors: Toluwani Aremu, Noor Hussein, Munachiso Nwadike, Samuele Poppi, Jie Zhang, Karthik Nandakumar, Neil Gong, Nils Lukas
Abstract: Watermarking offers a promising solution for GenAI providers to establish the provenance of their generated content. A watermark is a hidden signal embedded in the generated content, whose presence can later be verified using a secret watermarking key. A security threat to GenAI providers are \emph{forgery attacks}, where malicious users insert the provider's watermark into generated content that was \emph{not} produced by the provider's models, potentially damaging their reputation and undermining trust. One potential defense to resist forgery is using multiple keys to watermark generated content. However, it has been shown that forgery attacks remain successful when adversaries can collect sufficiently many watermarked samples. We propose an improved multi-key watermarking method that resists all surveyed forgery attacks and scales independently of the number of watermarked samples collected by the adversary. Our method accepts content as genuinely watermarked only if \emph{exactly} one watermark is detected. We focus on the image and text modalities, but our detection method is modality-agnostic, since it treats the underlying watermarking method as a black-box. We derive theoretical bounds on forgery-resistance and empirically validate them using Mistral-7B. Our results show a decrease in forgery success from up to $100\%$ using single-key baselines to only $2\%$. While our method resists all surveyed attacks, we find that highly capable, adaptive attackers can still achieve success rates of up to $65\%$ if watermarked content generated using different keys is easily separable.
Authors: Chen Feng, Yicheng Lin, Shaojie Zhuo, Chenzheng Su, Ramchalam Kinattinkara Ramakrishnan, Zhaocong Yuan, Xiaopeng Zhang
Abstract: Recent advances in Automatic Speech Recognition (ASR) have demonstrated remarkable accuracy and robustness in diverse audio applications, such as live transcription and voice command processing. However, deploying these models on resource-constrained edge devices (e.g., IoT device, wearables) still presents substantial challenges due to strict limits on memory, compute and power. Quantization, particularly Post-Training Quantization (PTQ), offers an effective way to reduce model size and inference cost without retraining. Despite its importance, the performance implications of various advanced quantization methods and bit-width configurations on ASR models remain unclear. In this work, we present a comprehensive benchmark of eight state-of-the-art (SOTA) PTQ methods applied to two leading edge-ASR model families, Whisper and Moonshine. We systematically evaluate model performances (i.e., accuracy, memory I/O and bit operations) across seven diverse datasets from the open ASR leader-board, analyzing the impact of quantization and various configurations on both weights and activations. Built on an extension of the LLM compression toolkit, our framework integrates edge-ASR models, diverse advanced quantization algorithms, a unified calibration and evaluation data pipeline, with detailed analysis tools. Our results characterize the trade-offs between efficiency and accuracy, demonstrating that even $3$-bit quantization can succeed on high capacity models when using advanced PTQ techniques. These findings provide valuable insights for optimizing ASR models on low-power, always-on edge devices.
Authors: Wonung Kim, Yubin Lee, Yoonsung Kim, Jinwoo Hwang, Seongryong Oh, Jiyong Jung, Aziz Huseynov, Woong Gyu Park, Chang Hyun Park, Divya Mahajan, Jongse Park
Abstract: Transformers are the driving force behind today's Large Language Models (LLMs), serving as the foundation for their performance and versatility. Yet, their compute and memory costs grow with sequence length, posing scalability challenges for long-context inferencing. In response, the algorithm community is exploring alternative architectures, such as state space models (SSMs), linear attention, and recurrent neural networks (RNNs), which we refer to as post-transformers. This shift presents a key challenge: building a serving system that efficiently supports both transformer and post-transformer LLMs within a unified framework. To address this challenge, we analyze the performance characteristics of transformer and post-transformer LLMs. Despite their algorithmic differences, both are fundamentally limited by memory bandwidth under batched inference due to attention in transformers and state updates in post-transformers. Further analyses suggest two additional insights: (1) state update operations, unlike attention, incur high hardware cost, making per-bank PIM acceleration inefficient, and (2) different low-precision arithmetic methods offer varying accuracy-area tradeoffs, while we identify Microsoft's MX as the Pareto-optimal choice. Building on these insights, we design Pimba as an array of State-update Processing Units (SPUs), each shared between two banks to enable interleaved access to PIM. Each SPU includes a State-update Processing Engine (SPE) that comprises element-wise multipliers and adders using MX-based quantized arithmetic, enabling efficient execution of state update and attention operations. Our evaluation shows that, compared to LLM-optimized GPU and GPU+PIM systems, Pimba achieves up to 4.1x and 2.1x higher token generation throughput, respectively.
Authors: Yen-Linh Vu, Dinh-Thang Duong, Truong-Binh Duong, Anh-Khoi Nguyen, Thanh-Huy Nguyen, Le Thien Phuc Nguyen, Jianhua Xing, Xingjian Li, Tianyang Wang, Ulas Bagci, Min Xu
Abstract: Recent progress has been made in region-aware vision-language modeling, particularly with the emergence of the Describe Anything Model (DAM). DAM is capable of generating detailed descriptions of any specific image areas or objects without the need for additional localized image-text alignment supervision. We hypothesize that such region-level descriptive capability is beneficial for the task of Visual Question Answering (VQA), especially in challenging scenarios involving images with dense text. In such settings, the fine-grained extraction of textual information is crucial to producing correct answers. Motivated by this, we introduce DAM-QA, a framework with a tailored evaluation protocol, developed to investigate and harness the region-aware capabilities from DAM for the text-rich VQA problem that requires reasoning over text-based information within images. DAM-QA incorporates a mechanism that aggregates answers from multiple regional views of image content, enabling more effective identification of evidence that may be tied to text-related elements. Experiments on six VQA benchmarks show that our approach consistently outperforms the baseline DAM, with a notable 7+ point gain on DocVQA. DAM-QA also achieves the best overall performance among region-aware models with fewer parameters, significantly narrowing the gap with strong generalist VLMs. These results highlight the potential of DAM-like models for text-rich and broader VQA tasks when paired with efficient usage and integration strategies. Our code is publicly available at https://github.com/Linvyl/DAM-QA.git.
Authors: Danilo Avola, Emad Emam, Dario Montagnini, Daniele Pannone, Amedeo Ranaldi
Abstract: Person Re-Identification is a key and challenging task in video surveillance. While traditional methods rely on visual data, issues like poor lighting, occlusion, and suboptimal angles often hinder performance. To address these challenges, we introduce WhoFi, a novel pipeline that utilizes Wi-Fi signals for person re-identification. Biometric features are extracted from Channel State Information (CSI) and processed through a modular Deep Neural Network (DNN) featuring a Transformer-based encoder. The network is trained using an in-batch negative loss function to learn robust and generalizable biometric signatures. Experiments on the NTU-Fi dataset show that our approach achieves competitive results compared to state-of-the-art methods, confirming its effectiveness in identifying individuals via Wi-Fi signals.
Authors: Xiaoya Li, Xiaofei Sun, Albert Wang, Jiwei Li, Chris Shum
Abstract: The exponential growth in demand for GPU computing resources has created an urgent need for automated CUDA optimization strategies. While recent advances in LLMs show promise for code generation, current SOTA models achieve low success rates in improving CUDA speed. In this paper, we introduce CUDA-L1, an automated reinforcement learning framework for CUDA optimization that employs a novel contrastive RL algorithm. CUDA-L1 achieves significant performance improvements on the CUDA optimization task: trained on NVIDIA A100, it delivers an average speedup of x3.12 with a median speedup of x1.42 across all 250 CUDA kernels of KernelBench, with peak speedups reaching x120. Furthermore, the model also demonstrates portability across GPU architectures, achieving average speedups of x3.12 on L40, x2.50 on RTX 3090, x2.39 on H100, and x2.37 on H20 despite being optimized specifically for A100. The capabilities of CUDA-L1 demonstrate that, RL can transform an initially poor-performing LLM into an effective CUDA optimizer through speedup-based reward signals alone, without human expertise or domain knowledge. This paradigm opens possibilities for automated optimization of CUDA operations, and holds promise to substantially promote GPU efficiency and alleviate the rising pressure on GPU computing resources. We also identify important challenges posed by training RL models for tasks like CUDA development, where RL often learns to exploit loopholes in reward functions rather than solve the intended optimization problems. By identifying these failure modes and analyzing their root causes, we develop practical methods for creating more robust training procedures that prevent reward hacking.
Authors: Sebastian Pokutta
Abstract: We consider the problem of minimizing a difference of (smooth) convex functions over a compact convex feasible region $P$, i.e., $\min_{x \in P} f(x) - g(x)$, with smooth $f$ and Lipschitz continuous $g$. This computational study builds upon and complements the framework of Maskan et al. [2025] by integrating advanced Frank-Wolfe variants to reduce computational overhead. We empirically show that constrained DC problems can be efficiently solved using a combination of the Blended Pairwise Conditional Gradients (BPCG) algorithm [Tsuji et al., 2022] with warm-starting and the adaptive error bound from Maskan et al. [2025]. The result is a highly efficient and scalable projection-free algorithm for constrained DC optimization.
Authors: Antonio Tudisco, Deborah Volpe, Giacomo Orlandi, Giovanna Turvani
Abstract: The growing variety of quantum hardware technologies, each with unique peculiarities such as connectivity and native gate sets, creates challenges when selecting the best platform for executing a specific quantum circuit. This selection process usually involves a brute-force approach: compiling the circuit on various devices and evaluating performance based on factors such as circuit depth and gate fidelity. However, this method is computationally expensive and does not scale well as the number of available quantum processors increases. In this work, we propose a Graph Neural Network (GNN)-based predictor that automates hardware selection by analyzing the Directed Acyclic Graph (DAG) representation of a quantum circuit. Our study evaluates 498 quantum circuits (up to 27 qubits) from the MQT Bench dataset, compiled using Qiskit on four devices: three superconducting quantum processors (IBM-Kyiv, IBM-Brisbane, IBM-Sherbrooke) and one trapped-ion processor (IONQ-Forte). Performance is estimated using a metric that integrates circuit depth and gate fidelity, resulting in a dataset where 93 circuits are optimally compiled on the trapped-ion device, while the remaining circuits prefer superconducting platforms. By exploiting graph-based machine learning, our approach avoids extracting the circuit features for the model evaluation but directly embeds it as a graph, significantly accelerating the optimal target decision-making process and maintaining all the information. Experimental results prove 94.4% accuracy and an 85.5% F1 score for the minority class, effectively predicting the best compilation target. The developed code is publicly available on GitHub (https://github.com/antotu/GNN-Model-Quantum-Predictor).
URLs: https://github.com/antotu/GNN-Model-Quantum-Predictor).
Authors: Michael Amir, Guang Yang, Zhan Gao, Keisuke Okumura, Heedo Woo, Amanda Prorok
Abstract: Constraint-based optimization is a cornerstone of robotics, enabling the design of controllers that reliably encode task and safety requirements such as collision avoidance or formation adherence. However, handcrafted constraints can fail in multi-agent settings that demand complex coordination. We introduce ReCoDe--Reinforcement-based Constraint Design--a decentralized, hybrid framework that merges the reliability of optimization-based controllers with the adaptability of multi-agent reinforcement learning. Rather than discarding expert controllers, ReCoDe improves them by learning additional, dynamic constraints that capture subtler behaviors, for example, by constraining agent movements to prevent congestion in cluttered scenarios. Through local communication, agents collectively constrain their allowed actions to coordinate more effectively under changing conditions. In this work, we focus on applications of ReCoDe to multi-agent navigation tasks requiring intricate, context-based movements and consensus, where we show that it outperforms purely handcrafted controllers, other hybrid approaches, and standard MARL baselines. We give empirical (real robot) and theoretical evidence that retaining a user-defined controller, even when it is imperfect, is more efficient than learning from scratch, especially because ReCoDe can dynamically change the degree to which it relies on this controller.
Authors: David Bauer, Qi Wu, Hamid Gadirov, Kwan-Liu Ma
Abstract: Real-time path tracing is rapidly becoming the standard for rendering in entertainment and professional applications. In scientific visualization, volume rendering plays a crucial role in helping researchers analyze and interpret complex 3D data. Recently, photorealistic rendering techniques have gained popularity in scientific visualization, yet they face significant challenges. One of the most prominent issues is slow rendering performance and high pixel variance caused by Monte Carlo integration. In this work, we introduce a novel radiance caching approach for path-traced volume rendering. Our method leverages advances in volumetric scene representation and adapts 3D Gaussian splatting to function as a multi-level, path-space radiance cache. This cache is designed to be trainable on the fly, dynamically adapting to changes in scene parameters such as lighting configurations and transfer functions. By incorporating our cache, we achieve less noisy, higher-quality images without increasing rendering costs. To evaluate our approach, we compare it against a baseline path tracer that supports uniform sampling and next-event estimation and the state-of-the-art for neural radiance caching. Through both quantitative and qualitative analyses, we demonstrate that our path-space radiance cache is a robust solution that is easy to integrate and significantly enhances the rendering quality of volumetric visualization applications while maintaining comparable computational efficiency.
Authors: Hussein A. Ammar, Raviraj Adve, Shahram Shahbazpanahi, Gary Boudreau, Israfil Bahceci
Abstract: In the user-centric cell-free massive MIMO (UC-mMIMO) network scheme, user mobility necessitates updating the set of serving access points to maintain the user-centric clustering. Such updates are typically performed through handoff (HO) operations; however, frequent HOs lead to overheads associated with the allocation and release of resources. This paper presents a deep reinforcement learning (DRL)-based solution to predict and manage these connections for mobile users. Our solution employs the Soft Actor-Critic algorithm, with continuous action space representation, to train a deep neural network to serve as the HO policy. We present a novel proposition for a reward function that integrates a HO penalty in order to balance the attainable rate and the associated overhead related to HOs. We develop two variants of our system; the first one uses mobility direction-assisted (DA) observations that are based on the user movement pattern, while the second one uses history-assisted (HA) observations that are based on the history of the large-scale fading (LSF). Simulation results show that our DRL-based continuous action space approach is more scalable than discrete space counterpart, and that our derived HO policy automatically learns to gather HOs in specific time slots to minimize the overhead of initiating HOs. Our solution can also operate in real time with a response time less than 0.4 ms.
Authors: T. Aleksandra Ma, Sile Yin, Li-Chia Yang, Shuo Zhang
Abstract: Speech enhancement in audio-only settings remains challenging, particularly in the presence of interfering speakers. This paper presents a simple yet effective real-time audio-visual speech enhancement (AVSE) system, RAVEN, which isolates and enhances the on-screen target speaker while suppressing interfering speakers and background noise. We investigate how visual embeddings learned from audio-visual speech recognition (AVSR) and active speaker detection (ASD) contribute to AVSE across different SNR conditions and numbers of interfering speakers. Our results show concatenating embeddings from AVSR and ASD models provides the greatest improvement in low-SNR, multi-speaker environments, while AVSR embeddings alone perform best in noise-only scenarios. In addition, we develop a real-time streaming system that operates on a computer CPU and we provide a video demonstration and code repository. To our knowledge, this is the first open-source implementation of a real-time AVSE system.
Authors: Viacheslav Pirogov, Maksim Artemev
Abstract: Deepfakes powered by advanced machine learning models present a significant and evolving threat to identity verification and the authenticity of digital media. Although numerous detectors have been developed to address this problem, their effectiveness has yet to be tested when applied to real-world data. In this work we evaluate modern deepfake detectors, introducing a novel testing procedure designed to mimic real-world scenarios for deepfake detection. Using state-of-the-art deepfake generation methods, we create a comprehensive dataset containing more than 500,000 high-quality deepfake images. Our analysis shows that detecting deepfakes still remains a challenging task. The evaluation shows that in fewer than half of the deepfake detectors tested achieved an AUC score greater than 60%, with the lowest being 50%. We demonstrate that basic image manipulations, such as JPEG compression or image enhancement, can significantly reduce model performance. All code and data are publicly available at https://github.com/SumSubstance/Deepfake-Detectors-in-the-Wild.
URLs: https://github.com/SumSubstance/Deepfake-Detectors-in-the-Wild.
Authors: Harry Shomer, Jiejun Xu
Abstract: Label placement is a critical aspect of map design, serving as a form of spatial annotation that directly impacts clarity and interpretability. Despite its importance, label placement remains largely manual and difficult to scale, as existing automated systems struggle to integrate cartographic conventions, adapt to context, or interpret labeling instructions. In this work, we introduce a new paradigm for automatic label placement (ALP) that formulates the task as a data editing problem and leverages large language models (LLMs) for context-aware spatial annotation. To support this direction, we curate MAPLE, the first known benchmarking dataset for evaluating ALP on real-world maps, encompassing diverse landmark types and label placement annotations from open-source data. Our method retrieves labeling guidelines relevant to each landmark type leveraging retrieval-augmented generation (RAG), integrates them into prompts, and employs instruction-tuned LLMs to generate ideal label coordinates. We evaluate four open-source LLMs on MAPLE, analyzing both overall performance and generalization across different types of landmarks. This includes both zero-shot and instruction-tuned performance. Our results demonstrate that LLMs, when guided by structured prompts and domain-specific retrieval, can learn to perform accurate spatial edits, aligning the generated outputs with expert cartographic standards. Overall, our work presents a scalable framework for AI-assisted map finishing and demonstrates the potential of foundation models in structured data editing tasks. The code and data can be found at https://github.com/HarryShomer/MAPLE.