Authors: Yulison Herry Chrisnanto, Julian Evan Chrisnanto
Abstract: Cut order planning (COP) is a critical challenge in the textile industry, directly impacting fabric utilization and production costs. Conventional methods based on static heuristics and catalog-based estimations often struggle to adapt to dynamic production environments, resulting in suboptimal solutions and increased waste. In response, we propose a novel Quantum-Inspired Deep Reinforcement Learning (QI-DRL) framework that integrates Long Short-Term Memory (LSTM) networks with Ornstein-Uhlenbeck noise. This hybrid approach is designed to explicitly address key research questions regarding the benefits of quantum-inspired probabilistic representations, the role of LSTM-based memory in capturing sequential dependencies, and the effectiveness of OU noise in facilitating smooth exploration and faster convergence. Extensive training over 1000 episodes demonstrates robust performance, with an average reward of 0.81 (-+0.03) and a steady decrease in prediction loss to 0.15 (-+0.02). A comparative analysis reveals that the proposed approach achieves fabric cost savings of up to 13% compared to conventional methods. Furthermore, statistical evaluations indicate low variability and stable convergence. Despite the fact that the simulation model makes several simplifying assumptions, these promising results underscore the potential of the scalable and adaptive framework to enhance manufacturing efficiency and pave the way for future innovations in COP optimization.
Authors: Xiaohan Yi, Guikun Xu, Xi Xiao, Zhong Zhang, Liu Liu, Yatao Bian, Peilin Zhao
Abstract: We present CrystalDiT, a diffusion transformer for crystal structure generation that achieves state-of-the-art performance by challenging the trend of architectural complexity. Instead of intricate, multi-stream designs, CrystalDiT employs a unified transformer that imposes a powerful inductive bias: treating lattice and atomic properties as a single, interdependent system. Combined with a periodic table-based atomic representation and a balanced training strategy, our approach achieves 9.62% SUN (Stable, Unique, Novel) rate on MP-20, substantially outperforming recent methods including FlowMM (4.38%) and MatterGen (3.42%). Notably, CrystalDiT generates 63.28% unique and novel structures while maintaining comparable stability rates, demonstrating that architectural simplicity can be more effective than complexity for materials discovery. Our results suggest that in data-limited scientific domains, carefully designed simple architectures outperform sophisticated alternatives that are prone to overfitting.
Authors: K\'evin Ducharlet, Louise Trav\'e-Massuy\`es, Jean-Bernard Lasserre, Marie-V\'eronique Le Lann, Youssef Miloudi
Abstract: Outlier detection holds significant importance in the realm of data mining, particularly with the growing pervasiveness of data acquisition methods. The ability to identify outliers in data streams is essential for maintaining data quality and detecting faults. However, dealing with data streams presents challenges due to the non-stationary nature of distributions and the ever-increasing data volume. While numerous methods have been proposed to tackle this challenge, a common drawback is the lack of straightforward parameterization in many of them. This article introduces two novel methods: DyCF and DyCG. DyCF leverages the Christoffel function from the theory of approximation and orthogonal polynomials. Conversely, DyCG capitalizes on the growth properties of the Christoffel function, eliminating the need for tuning parameters. Both approaches are firmly rooted in a well-defined algebraic framework, meeting crucial demands for data stream processing, with a specific focus on addressing low-dimensional aspects and maintaining data history without memory cost. A comprehensive comparison between DyCF, DyCG, and state-of-the-art methods is presented, using both synthetic and real industrial data streams. The results show that DyCF outperforms fine-tuning methods, offering superior performance in terms of execution time and memory usage. DyCG performs less well, but has the considerable advantage of requiring no tuning at all.
Authors: Bangchao Deng, Lianhua Ji, Chunhua Chen, Xin Jing, Ling Ding, Bingqing QU, Pengyang Wang, Dingqi Yang
Abstract: Next location prediction is a critical task in human mobility modeling, enabling applications like travel planning and urban mobility management. Existing methods mainly rely on historical spatiotemporal trajectory data to train sequence models that directly forecast future locations. However, they often overlook the importance of the future spatiotemporal contexts, which are highly informative for the future locations. For example, knowing how much time and distance a user will travel could serve as a critical clue for predicting the user's next location. Against this background, we propose \textbf{STRelay}, a universal \textbf{\underline{S}}patio\textbf{\underline{T}}emporal \textbf{\underline{Relay}}ing framework explicitly modeling the future spatiotemporal context given a human trajectory, to boost the performance of different location prediction models. Specifically, STRelay models future spatiotemporal contexts in a relaying manner, which is subsequently integrated with the encoded historical representation from a base location prediction model, enabling multi-task learning by simultaneously predicting the next time interval, next moving distance interval, and finally the next location. We evaluate STRelay integrated with four state-of-the-art location prediction base models on four real-world trajectory datasets. Results demonstrate that STRelay consistently improves prediction performance across all cases by 3.19\%-11.56\%. Additionally, we find that the future spatiotemporal contexts are particularly helpful for entertainment-related locations and also for user groups who prefer traveling longer distances. The performance gain on such non-daily-routine activities, which often suffer from higher uncertainty, is indeed complementary to the base location prediction models that often excel at modeling regular daily routine patterns.
Authors: Weilin Ruan, Xilin Dang, Ziyu Zhou, Sisuo Lyu, Yuxuan Liang
Abstract: Traffic prediction is a cornerstone of modern intelligent transportation systems and a critical task in spatio-temporal forecasting. Although advanced Spatio-temporal Graph Neural Networks (STGNNs) and pre-trained models have achieved significant progress in traffic prediction, two key challenges remain: (i) limited contextual capacity when modeling complex spatio-temporal dependencies, and (ii) low predictability at fine-grained spatio-temporal points due to heterogeneous patterns. Inspired by Retrieval-Augmented Generation (RAG), we propose RAST, a universal framework that integrates retrieval-augmented mechanisms with spatio-temporal modeling to address these challenges. Our framework consists of three key designs: 1) Decoupled Encoder and Query Generator to capture decoupled spatial and temporal features and construct a fusion query via residual fusion; 2) Spatio-temporal Retrieval Store and Retrievers to maintain and retrieve vectorized fine-grained patterns; and 3) Universal Backbone Predictor that flexibly accommodates pre-trained STGNNs or simple MLP predictors. Extensive experiments on six real-world traffic networks, including large-scale datasets, demonstrate that RAST achieves superior performance while maintaining computational efficiency.
Authors: Zeyu Zhang, Quanyu Dai, Rui Li, Xiaohe Bo, Xu Chen, Zhenhua Dong
Abstract: LLM-based agents have been extensively applied across various domains, where memory stands out as one of their most essential capabilities. Previous memory mechanisms of LLM-based agents are manually predefined by human experts, leading to higher labor costs and suboptimal performance. In addition, these methods overlook the memory cycle effect in interactive scenarios, which is critical to optimizing LLM-based agents for specific environments. To address these challenges, in this paper, we propose to optimize LLM-based agents with an adaptive and data-driven memory framework by modeling memory cycles. Specifically, we design an MoE gate function to facilitate memory retrieval, propose a learnable aggregation process to improve memory utilization, and develop task-specific reflection to adapt memory storage. Our memory framework empowers LLM-based agents to learn how to memorize information effectively in specific environments, with both off-policy and on-policy optimization. In order to evaluate the effectiveness of our proposed methods, we conduct comprehensive experiments across multiple aspects. To benefit the research community in this area, we release our project at https://github.com/nuster1128/learn_to_memorize.
Authors: Yifu Han, Louis J. Durlofsky
Abstract: Many subsurface formations, including some of those under consideration for large-scale geological carbon storage, include extensive faults that can strongly impact fluid flow. In this study, we develop a new recurrent transformer U-Net surrogate model to provide very fast predictions for pressure and CO2 saturation in realistic faulted subsurface aquifer systems. The geomodel includes a target aquifer (into which supercritical CO2 is injected), surrounding regions, caprock, two extensive faults, and two overlying aquifers. The faults can act as leakage pathways between the three aquifers. The heterogeneous property fields in the target aquifer are characterized by hierarchical uncertainty, meaning both the geological metaparameters (e.g., mean and standard deviation of log-permeability) and the detailed cell properties of each realization, are uncertain. Fault permeabilities are also treated as uncertain. The model is trained with simulation results for (up to) 4000 randomly sampled realizations. Error assessments show that this model is more accurate than a previous recurrent residual U-Net, and that it maintains accuracy for qualitatively different leakage scenarios. The new surrogate is then used for global sensitivity analysis and data assimilation. A hierarchical Markov chain Monte Carlo data assimilation procedure is applied. Different monitoring strategies, corresponding to different amounts and types of observed data collected at monitoring wells, are considered for three synthetic true models. Detailed results demonstrate the degree of uncertainty reduction achieved with the various monitoring strategies. Posterior results for 3D saturation plumes and leakage volumes indicate the benefits of measuring pressure and saturation in all three aquifers.
Authors: Krisanu Sarkar
Abstract: The persistent challenge of catastrophic forgetting in neural networks has motivated extensive research in continual learning . This work presents a novel continual learning framework that integrates Fisher-weighted asymmetric regularization of parameter variances within a variational learning paradigm. Our method dynamically modulates regularization intensity according to parameter uncertainty, achieving enhanced stability and performance. Comprehensive evaluations on standard continual learning benchmarks including SplitMNIST, PermutedMNIST, and SplitFashionMNIST demonstrate substantial improvements over existing approaches such as Variational Continual Learning and Elastic Weight Consolidation . The asymmetric variance penalty mechanism proves particularly effective in maintaining knowledge across sequential tasks while improving model accuracy. Experimental results show our approach not only boosts immediate task performance but also significantly mitigates knowledge degradation over time, effectively addressing the fundamental challenge of catastrophic forgetting in neural networks
Authors: Yunyan Zheng, Zhichao Zhang, Wei Yao
Abstract: Graph signal processing has become an essential tool for analyzing data structured on irregular domains. While conventional graph shift operators (GSOs) are effective for certain tasks, they inherently lack flexibility in modeling dependencies between non-adjacent nodes, limiting their ability to represent complex graph structures. To address this limitation, this paper proposes the unified extended matrix (UEM) framework, which integrates the extended-adjacency matrix and the unified graph representation matrix through parametric design, so as to be able to flexibly adapt to different graph structures and reveal more graph signal information. Theoretical analysis of the UEM is conducted, demonstrating positive semi-definiteness and eigenvalue monotonicity under specific conditions. Then, we propose graph Fourier transform based on UEM (UEM-GFT), which can adaptively tune spectral properties to enhance signal processing performance. Experimental results on synthetic and real-world datasets demonstrate that the UEM-GFT outperforms existing GSO-based methods in anomaly detection tasks, achieving superior performance across varying network topologies.
Authors: Zhendong Yang, Jie Wang, Liansong Zong, Xiaorong Liu, Quan Qian, Shiqian Chen
Abstract: Few-Shot Class-Incremental Fault Diagnosis (FSC-FD), which aims to continuously learn from new fault classes with only a few samples without forgetting old ones, is critical for real-world industrial systems. However, this challenging task severely amplifies the issues of catastrophic forgetting of old knowledge and overfitting on scarce new data. To address these challenges, this paper proposes a novel framework built upon Dual-Granularity Representations, termed the Dual-Granularity Guidance Network (DGGN). Our DGGN explicitly decouples feature learning into two parallel streams: 1) a fine-grained representation stream, which utilizes a novel Multi-Order Interaction Aggregation module to capture discriminative, class-specific features from the limited new samples. 2) a coarse-grained representation stream, designed to model and preserve general, class-agnostic knowledge shared across all fault types. These two representations are dynamically fused by a multi-semantic cross-attention mechanism, where the stable coarse-grained knowledge guides the learning of fine-grained features, preventing overfitting and alleviating feature conflicts. To further mitigate catastrophic forgetting, we design a Boundary-Aware Exemplar Prioritization strategy. Moreover, a decoupled Balanced Random Forest classifier is employed to counter the decision boundary bias caused by data imbalance. Extensive experiments on the TEP benchmark and a real-world MFF dataset demonstrate that our proposed DGGN achieves superior diagnostic performance and stability compared to state-of-the-art FSC-FD approaches. Our code is publicly available at https://github.com/MentaY/DGGN
Authors: Dhruv D. Modi, Rong Pan
Abstract: Time series foundation models (TSFMs) such as Lag-Llama, TimeGPT, Chronos, MOMENT, UniTS, and TimesFM have shown strong generalization and zero-shot capabilities for time series forecasting, anomaly detection, classification, and imputation. Despite these advantages, their predictions still suffer from variance, domain-specific bias, and limited uncertainty quantification when deployed on real operational data. This paper investigates a suite of statistical and ensemble-based enhancement techniques, including bootstrap-based bagging, regression-based stacking, prediction interval construction, statistical residual modeling, and iterative error feedback, to improve robustness and accuracy. Using the Belgium Electricity Short-Term Load Forecasting dataset as a case study, we demonstrate that the proposed hybrids consistently outperform standalone foundation models across multiple horizons. Regression-based ensembles achieve the lowest mean squared error; bootstrap aggregation markedly reduces long-context errors; residual modeling corrects systematic bias; and the resulting prediction intervals achieve near nominal coverage with widths shrinking as context length increases. The results indicate that integrating statistical reasoning with modern foundation models yields measurable gains in accuracy, reliability, and interpretability for real-world time series applications.
Authors: Tianhua Chen
Abstract: From large language models to multi-modal agents, Generative Artificial Intelligence (AI) now underpins state-of-the-art systems. Despite their varied architectures, many share a common foundation in probabilistic latent variable models (PLVMs), where hidden variables explain observed data for density estimation, latent reasoning, and structured inference. This paper presents a unified perspective by framing both classical and modern generative methods within the PLVM paradigm. We trace the progression from classical flat models such as probabilistic PCA, Gaussian mixture models, latent class analysis, item response theory, and latent Dirichlet allocation, through their sequential extensions including Hidden Markov Models, Gaussian HMMs, and Linear Dynamical Systems, to contemporary deep architectures: Variational Autoencoders as Deep PLVMs, Normalizing Flows as Tractable PLVMs, Diffusion Models as Sequential PLVMs, Autoregressive Models as Explicit Generative Models, and Generative Adversarial Networks as Implicit PLVMs. Viewing these architectures under a common probabilistic taxonomy reveals shared principles, distinct inference strategies, and the representational trade-offs that shape their strengths. We offer a conceptual roadmap that consolidates generative AI's theoretical foundations, clarifies methodological lineages, and guides future innovation by grounding emerging architectures in their probabilistic heritage.
Authors: Boran Zhao, Hetian Liu, Zihang Yuan, Li Zhu, Fan Yang, Lina Xie Tian Xia, Wenzhe Zhao, Pengju Ren
Abstract: Training deep neural networks (DNNs) directly on edge devices has attracted increasing attention, as it offers promising solutions to challenges such as domain adaptation and privacy preservation. However, conventional DNN training typically requires large-scale datasets, which imposes prohibitive overhead on edge devices-particularly for emerging large language model (LLM) tasks. To address this challenge, a DNN-free method (ie., dataset sampling without DNN), named NMS (Near-Memory Sampling), has been introduced. By first conducting dimensionality reduction of the dataset and then performing exemplar sampling in the reduced space, NMS avoids the architectural bias inherent in DNN-based methods and thus achieves better generalization. However, The state-of-the-art, NMS, suffers from two limitations: (1) The mismatch between the search method and the non-monotonic property of the perplexity error function leads to the emergence of outliers in the reduced representation; (2) Key parameter (ie., target perplexity) is selected empirically, introducing arbitrariness and leading to uneven sampling. These two issues lead to representative bias of examplars, resulting in degraded accuracy. To address these issues, we propose AdapSNE, which integrates an efficient non-monotonic search method-namely, the Fireworks Algorithm (FWA)-to suppress outliers, and employs entropy-guided optimization to enforce uniform sampling, thereby ensuring representative training samples and consequently boosting training accuracy. To cut the edge-side cost arising from the iterative computations of FWA search and entropy-guided optimization, we design an accelerator with custom dataflow and time-multiplexing markedly reducing on-device training energy and area.
Authors: Junle Liu, Chang Liu, Yanyu Ke, Qiuxiang Huang, Jiachen Zhao, Wenliang Chen, K. T. Tse, Gang Hu
Abstract: Acquiring temporally high-frequency and spatially high-resolution turbulent wake flow fields in particle image velocimetry (PIV) experiments remains a significant challenge due to hardware limitations and measurement noise. In contrast, temporal high-frequency measurements of spatially sparse wall pressure are more readily accessible in wind tunnel experiments. In this study, we propose a novel cross-modal temporal upscaling framework, LatentFlow, which reconstructs high-frequency (512 Hz) turbulent wake flow fields by fusing synchronized low-frequency (15 Hz) flow field and pressure data during training, and high-frequency wall pressure signals during inference. The first stage involves training a pressure-conditioned $\beta$-variation autoencoder ($p$C-$\beta$-VAE) to learn a compact latent representation that captures the intrinsic dynamics of the wake flow. A secondary network maps synchronized low-frequency wall pressure signals into the latent space, enabling reconstruction of the wake flow field solely from sparse wall pressure. Once trained, the model utilizes high-frequency, spatially sparse wall pressure inputs to generate corresponding high-frequency flow fields via the $p$C-$\beta$-VAE decoder. By decoupling the spatial encoding of flow dynamics from temporal pressure measurements, LatentFlow provides a scalable and robust solution for reconstructing high-frequency turbulent wake flows in data-constrained experimental settings.
Authors: Kushal Kapoor, Wyatt Mackey, Yiannis Aloimonos, Xiaomin Lin
Abstract: We propose HiCL, a novel hippocampal-inspired dual-memory continual learning architecture designed to mitigate catastrophic forgetting by using elements inspired by the hippocampal circuitry. Our system encodes inputs through a grid-cell-like layer, followed by sparse pattern separation using a dentate gyrus-inspired module with top-k sparsity. Episodic memory traces are maintained in a CA3-like autoassociative memory. Task-specific processing is dynamically managed via a DG-gated mixture-of-experts mechanism, wherein inputs are routed to experts based on cosine similarity between their normalized sparse DG representations and learned task-specific DG prototypes computed through online exponential moving averages. This biologically grounded yet mathematically principled gating strategy enables differentiable, scalable task-routing without relying on a separate gating network, and enhances the model's adaptability and efficiency in learning multiple sequential tasks. Cortical outputs are consolidated using Elastic Weight Consolidation weighted by inter-task similarity. Crucially, we incorporate prioritized replay of stored patterns to reinforce essential past experiences. Evaluations on standard continual learning benchmarks demonstrate the effectiveness of our architecture in reducing task interference, achieving near state-of-the-art results in continual learning tasks at lower computational costs.
Authors: Andrei Mateescu, Ioana Hadarau, Ionut Anghel, Tudor Cioara, Ovidiu Anchidin, Ancuta Nemes
Abstract: With the advent of wearable Internet of Things (IoT) devices, remote patient monitoring (RPM) emerged as a promising solution for managing heart failure. However, the heart rate can fluctuate significantly due to various factors, and without correlating it to the patient's actual physical activity, it becomes difficult to assess whether changes are significant. Although Artificial Intelligence (AI) models may enhance the accuracy and contextual understanding of remote heart rate monitoring, the integration of activity data is still rarely addressed. In this paper, we propose a Transformer model combined with a Laplace diffusion technique to model heart rate fluctuations driven by physical activity of the patient. Unlike prior models that treat activity as secondary, our approach conditions the entire modeling process on activity context using specialized embeddings and attention mechanisms to prioritize activity specific historical patents. The model captures both long-term patterns and activity-specific heart rate dynamics by incorporating contextualized embeddings and dedicated encoder. The Transformer model was validated on a real-world dataset collected from 29 patients over a 4-month period. Experimental results show that our model outperforms current state-of-the-art methods, achieving a 43% reduction in mean absolute error compared to the considered baseline models. Moreover, the coefficient of determination R2 is 0.97 indicating the model predicted heart rate is in strong agreement with actual heart rate values. These findings suggest that the proposed model is a practical and effective tool for supporting both healthcare providers and remote patient monitoring systems.
Authors: Miru Kim, Mugon Joe, Minhae Kwon
Abstract: The expansion of machine learning into dynamic environments presents challenges in handling open-world problems where label shift, covariate shift, and unknown classes emerge. Post-training methods have been explored to address these challenges, adapting models to newly emerging data. However, these methods struggle when the initial pre-training is performed on class-imbalanced datasets, limiting generalization to minority classes. To address this, we propose a method that effectively handles open-world problems even when pre-training is conducted on imbalanced data. Our contrastive-based pre-training approach enhances classification performance, particularly for underrepresented classes. Our post-training mechanism generates reliable pseudo-labels, improving model robustness against open-world problems. We also introduce selective activation criteria to optimize the post-training process, reducing unnecessary computation. Extensive experiments demonstrate that our method significantly outperforms state-of-the-art adaptation techniques in both accuracy and efficiency across diverse open-world scenarios.
Authors: Jiacheng Li, Jianchao Tan, Zhidong Yang, Pingwei Sun, Feiye Huo, Jiayu Qin, Yerui Sun, Yuchen Xie, Xunliang Cai, Xiangyu Zhang, Maoxin He, Guangming Tan, Weile Jia, Tong Zhao
Abstract: Transformer architecture gradually dominates the LLM field. Recent advances in training optimization for Transformer-based large language models (LLMs) primarily focus on architectural modifications or optimizer adjustments. However, these approaches lack systematic optimization of weight patterns during training. Weight pattern refers to the distribution and relative magnitudes of weight parameters in a neural network. To address this issue, we propose a Weight Scaling method called WISCA to enhance training efficiency and model quality by strategically improving neural network weight patterns without changing network structures. By rescaling weights while preserving model outputs, WISCA indirectly optimizes the model's training trajectory. Experiments demonstrate that WISCA significantly improves convergence quality (measured by generalization capability and loss reduction), particularly in LLMs with Grouped Query Attention (GQA) architectures and LoRA fine-tuning tasks. Empirical results show 5.6% average improvement on zero-shot validation tasks and 2.12% average reduction in training perplexity across multiple architectures.
Authors: Zhong Guan, Likang Wu, Hongke Zhao, Jiahui Wang, Le Wu
Abstract: Many existing studies have achieved significant improvements in the reasoning capabilities of large language models (LLMs) through reinforcement learning with verifiable rewards (RLVR), while the enhancement of reasoning abilities in small language models (SLMs) has not yet been sufficiently explored. Combining distilled data from larger models with RLVR on small models themselves is a natural approach, but it still faces various challenges and issues. Therefore, we propose \textit{\underline{R}}ecall-\textit{\underline{E}}xtend \textit{\underline{D}}ynamics(RED): Enhancing Small Language Models through Controlled Exploration and Refined Offline Integration. In this paper, we explore the perspective of varying exploration spaces, balancing offline distillation with online reinforcement learning. Simultaneously, we specifically design and optimize for the insertion problem within offline data. By monitoring the ratio of entropy changes in the model concerning offline and online data, we regulate the weight of offline-SFT, thereby addressing the issues of insufficient exploration space in small models and the redundancy and complexity during the distillation process. Furthermore, to tackle the distribution discrepancies between offline data and the current policy, we design a sample-accuracy-based policy shift mechanism that dynamically chooses between imitating offline distilled data and learning from its own policy.
Authors: Muchammad Daniyal Kautsar, Afra Majida Hariono, Widyawan, Syukron Abu Ishaq Alfarozi, Kuntpong Wararatpanya
Abstract: Large Language Models (LLMs) present significant deployment challenges due to their immense size and computational requirements. Model compression techniques are essential for making these models practical for resource-constrained environments. A prominent compression strategy is low-rank factorization via Singular Value Decomposition (SVD) to reduce model parameters by approximating weight matrices. However, standard SVD focuses on minimizing matrix reconstruction error, often leading to a substantial loss of the model's functional performance. This performance degradation occurs because existing methods do not adequately correct for the functional information lost during compression. To address this gap, we introduce Corrective Adaptive Low-Rank Decomposition (CALR), a two-component compression approach. CALR combines a primary path of SVD-compressed layers with a parallel, learnable, low-rank corrective module that is explicitly trained to recover the functional residual error. Our experimental evaluation on SmolLM2-135M, Qwen3-0.6B, and Llama-3.2-1B, demonstrates that CALR can reduce parameter counts by 26.93% to 51.77% while retaining 59.45% to 90.42% of the original model's performance, consistently outperforming LaCo, ShortGPT, and LoSparse. CALR's success shows that treating functional information loss as a learnable signal is a highly effective compression paradigm. This approach enables the creation of significantly smaller, more efficient LLMs, advancing their accessibility and practical deployment in real-world applications.
Authors: Zhuding Liang, Jianxun Cui, Qingshuang Zeng, Feng Liu, Nenad Filipovic, Tijana Geroski
Abstract: Accurate and timely traffic flow forecasting is crucial for intelligent transportation systems. This paper presents a novel deep learning model, the Spatial-Temporal Unified Graph Attention Network (STGAtt). By leveraging a unified graph representation and an attention mechanism, STGAtt effectively captures complex spatial-temporal dependencies. Unlike methods relying on separate spatial and temporal dependency modeling modules, STGAtt directly models correlations within a Spatial-Temporal Unified Graph, dynamically weighing connections across both dimensions. To further enhance its capabilities, STGAtt partitions traffic flow observation signal into neighborhood subsets and employs a novel exchanging mechanism, enabling effective capture of both short-range and long-range correlations. Extensive experiments on the PEMS-BAY and SHMetro datasets demonstrate STGAtt's superior performance compared to state-of-the-art baselines across various prediction horizons. Visualization of attention weights confirms STGAtt's ability to adapt to dynamic traffic patterns and capture long-range dependencies, highlighting its potential for real-world traffic flow forecasting applications.
Authors: Harrison J. Goldwyn, Mitchell Krock, Johann Rudi, Daniel Getter, Julie Bessac
Abstract: Accurate quantification of uncertainty in neural network predictions remains a central challenge for scientific applications involving high-dimensional, correlated data. While existing methods capture either aleatoric or epistemic uncertainty, few offer closed-form, multidimensional distributions that preserve spatial correlation while remaining computationally tractable. In this work, we present a framework for training neural networks with a multidimensional Gaussian loss, generating closed-form predictive distributions over outputs with non-identically distributed and heteroscedastic structure. Our approach captures aleatoric uncertainty by iteratively estimating the means and covariance matrices, and is demonstrated on a super-resolution example. We leverage a Fourier representation of the covariance matrix to stabilize network training and preserve spatial correlation. We introduce a novel regularization strategy -- referred to as information sharing -- that interpolates between image-specific and global covariance estimates, enabling convergence of the super-resolution downscaling network trained on image-specific distributional loss functions. This framework allows for efficient sampling, explicit correlation modeling, and extensions to more complex distribution families all without disrupting prediction performance. We demonstrate the method on a surface wind speed downscaling task and discuss its broader applicability to uncertainty-aware prediction in scientific models.
Authors: Gabriel Moreira, Zita Marinho, Manuel Marques, Jo\~ao Paulo Costeira, Chenyan Xiong
Abstract: Traditional neural embeddings represent concepts as points, excelling at similarity but struggling with higher-level reasoning and asymmetric relationships. We introduce a novel paradigm: embedding concepts as linear subspaces. This framework inherently models generality via subspace dimensionality and hierarchy through subspace inclusion. It naturally supports set-theoretic operations like intersection (conjunction), linear sum (disjunction) and orthogonal complements (negations), aligning with classical formal semantics. To enable differentiable learning, we propose a smooth relaxation of orthogonal projection operators, allowing for the learning of both subspace orientation and dimension. Our method achieves state-of-the-art results in reconstruction and link prediction on WordNet. Furthermore, on natural language inference benchmarks, our subspace embeddings surpass bi-encoder baselines, offering an interpretable formulation of entailment that is both geometrically grounded and amenable to logical operations.
Authors: Shanhao Yuan, Yanqin Liu, Runfa Zhang, Limei Yan, Shunjun Wu, Libo Feng
Abstract: In this study, we firstly propose an auxiliary equation neural networks method (AENNM), an innovative analytical method that integrates neural networks (NNs) models with the auxiliary equation method to obtain exact solutions of nonlinear partial differential equations (NLPDEs). A key novelty of this method is the introduction of a novel activation function derived from the solutions of the Riccati equation, establishing a new mathematical link between differential equations theory and deep learning. By combining the strong approximation capability of NNs with the high precision of symbolic computation, AENNM significantly enhances computational efficiency and accuracy. To demonstrate the effectiveness of the AENNM in solving NLPDEs, three numerical examples are investigated, including the nonlinear evolution equation, the Korteweg-de Vries-Burgers equation, and the (2+1)-dimensional Boussinesq equation. Furthermore, some new trial functions are constructed by setting specific activation functions within the "2-2-2-1" and "3-2-2-1" NNs models. By embedding the auxiliary equation method into the NNs framework, we derive previously unreported solutions. The exact analytical solutions are expressed in terms of hyperbolic functions, trigonometric functions, and rational functions. Finally, three-dimensional plots, contour plots, and density plots are presented to illustrate the dynamic characteristics of the obtained solutions. This research provides a novel methodological framework for addressing NLPDEs, with broad applicability across scientific and engineering fields.
Authors: Dmitrii Feoktistov, Igor Ignashin, Andrey Veprikov, Nikita Borovko, Alexander Bogdanov, Savelii Chezhegov, Aleksandr Beznosikov
Abstract: While traditional Deep Learning (DL) optimization methods treat all training samples equally, Distributionally Robust Optimization (DRO) adaptively assigns importance weights to different samples. However, a significant gap exists between DRO and current DL practices. Modern DL optimizers require adaptivity and the ability to handle stochastic gradients, as these methods demonstrate superior performance. Additionally, for practical applications, a method should allow weight assignment not only to individual samples, but also to groups of objects (for example, all samples of the same class). This paper aims to bridge this gap by introducing ALSO $\unicode{x2013}$ Adaptive Loss Scaling Optimizer $\unicode{x2013}$ an adaptive algorithm for a modified DRO objective that can handle weight assignment to sample groups. We prove the convergence of our proposed algorithm for non-convex objectives, which is the typical case for DL models. Empirical evaluation across diverse Deep Learning tasks, from Tabular DL to Split Learning tasks, demonstrates that ALSO outperforms both traditional optimizers and existing DRO methods.
Authors: Yanlin Qu, Jose Blanchet, Peter Glynn
Abstract: Lyapunov functions are fundamental to establishing the stability of Markovian models, yet their construction typically demands substantial creativity and analytical effort. In this paper, we show that deep learning can automate this process by training neural networks to satisfy integral equations derived from first-transition analysis. Beyond stability analysis, our approach can be adapted to solve Poisson's equation and estimate stationary distributions. While neural networks are inherently function approximators on compact domains, it turns out that our approach remains effective when applied to Markov chains on non-compact state spaces. We demonstrate the effectiveness of this methodology through several examples from queueing theory and beyond.
Authors: Haosen Ge, Shuo Li, Lianghuan Huang
Abstract: Effective prompt engineering remains a challenging task for many applications. We introduce Weak-to-Strong Transfer (WST), an automatic prompt engineering framework where a small "Teacher" model generates instructions that enhance the performance of a much larger "Student" model. Unlike prior work, WST requires only a weak teacher, making it efficient and broadly applicable in settings where large models are closed-source or difficult to fine-tune. Using reinforcement learning, the Teacher Model's instructions are iteratively improved based on the Student Model's outcomes, yielding substantial gains across reasoning (MATH-500, GSM8K) and alignment (HH-RLHF) benchmarks - 98% on MATH-500 and 134% on HH-RLHF - and surpassing baselines such as GPT-4o-mini and Llama-70B. These results demonstrate that small models can reliably scaffold larger ones, unlocking latent capabilities while avoiding misleading prompts that stronger teachers may introduce, establishing WST as a scalable solution for efficient and safe LLM prompt refinement.
Authors: ZeMing Gong, Chuanqi Tang, Xiaoliang Huo, Nicholas Pellegrino, Austin T. Wang, Graham W. Taylor, Angel X. Chang, Scott C. Lowe, Joakim Bruslund Haurum
Abstract: Taxonomic classification in biodiversity research involves organizing biological specimens into structured hierarchies based on evidence, which can come from multiple modalities such as images and genetic information. We investigate whether hyperbolic networks can provide a better embedding space for such hierarchical models. Our method embeds multimodal inputs into a shared hyperbolic space using contrastive and a novel stacked entailment-based objective. Experiments on the BIOSCAN-1M dataset show that hyperbolic embedding achieves competitive performance with Euclidean baselines, and outperforms all other models on unseen species classification using DNA barcodes. However, fine-grained classification and open-world generalization remain challenging. Our framework offers a structure-aware foundation for biodiversity modelling, with potential applications to species discovery, ecological monitoring, and conservation efforts.
Authors: Ivan Rodkin, Daniil Orel, Konstantin Smirnov, Arman Bolatov, Bilal Elbouardi, Besher Hassan, Yuri Kuratov, Aydar Bulatov, Preslav Nakov, Timothy Baldwin, Artem Shelmanov, Mikhail Burtsev
Abstract: Reasoning is a core capability of large language models, yet understanding how they learn and perform multi-step reasoning remains an open problem. In this study, we explore how different architectures and training methods affect model multi-step reasoning capabilities within a cellular automata framework. By training on state sequences generated with random Boolean functions for random initial conditions to exclude memorization, we demonstrate that most neural architectures learn to abstract the underlying rules. While models achieve high accuracy in next-state prediction, their performance declines sharply if multi-step reasoning is required. We confirm that increasing model depth plays a crucial role for sequential computations. We demonstrate that an extension of the effective model depth with recurrence, memory, and test-time compute scaling substantially enhances reasoning capabilities.
Authors: Jiaee Cheong, Abtin Mogharabin, Paul Liang, Hatice Gunes, Sinan Kalkan
Abstract: Early efforts on leveraging self-supervised learning (SSL) to improve machine learning (ML) fairness has proven promising. However, such an approach has yet to be explored within a multimodal context. Prior work has shown that, within a multimodal setting, different modalities contain modality-unique information that can complement information of other modalities. Leveraging on this, we propose a novel subject-level loss function to learn fairer representations via the following three mechanisms, adapting the variance-invariance-covariance regularization (VICReg) method: (i) the variance term, which reduces reliance on the protected attribute as a trivial solution; (ii) the invariance term, which ensures consistent predictions for similar individuals; and (iii) the covariance term, which minimizes correlational dependence on the protected attribute. Consequently, our loss function, coined as FAIRWELL, aims to obtain subject-independent representations, enforcing fairness in multimodal prediction tasks. We evaluate our method on three challenging real-world heterogeneous healthcare datasets (i.e. D-Vlog, MIMIC and MODMA) which contain different modalities of varying length and different prediction tasks. Our findings indicate that our framework improves overall fairness performance with minimal reduction in classification performance and significantly improves on the performance-fairness Pareto frontier.
Authors: Yuebo Luo, Shiyang Li, Junran Tao, Kiran Thorat, Xi Xie, Hongwu Peng, Nuo Xu, Caiwen Ding, Shaoyi Huang
Abstract: The increasing scale and complexity of integrated circuit design have led to increased challenges in Electronic Design Automation (EDA). Graph Neural Networks (GNNs) have emerged as a promising approach to assist EDA design as circuits can be naturally represented as graphs. While GNNs offer a foundation for circuit analysis, they often fail to capture the full complexity of EDA designs. Heterogeneous Graph Neural Networks (HGNNs) can better interpret EDA circuit graphs as they capture both topological relationships and geometric features. However, the improved representation capability comes at the cost of even higher computational complexity and processing cost due to their serial module-wise message-passing scheme, creating a significant performance bottleneck. In this paper, we propose DR-CircuitGNN, a fast GPU kernel design by leveraging row-wise sparsity-aware Dynamic-ReLU and optimizing SpMM kernels during heterogeneous message-passing to accelerate HGNNs training on EDA-related circuit graph datasets. To further enhance performance, we propose a parallel optimization strategy that maximizes CPU-GPU concurrency by concurrently processing independent subgraphs using multi-threaded CPU initialization and GPU kernel execution via multiple cudaStreams. Our experiments show that on three representative CircuitNet designs (small, medium, large), the proposed method can achieve up to 3.51x and 4.09x speedup compared to the SOTA for forward and backward propagation, respectively. On full-size CircuitNet and sampled Mini-CircuitNet, our parallel design enables up to 2.71x speed up over the official DGL implementation cuSPARSE with negligible impact on correlation scores and error rates.
Authors: Nathan X. Kodama, Kenneth A. Loparo
Abstract: Inferring temporal interaction graphs and higher-order structure from neural signals is a key problem in building generative models for systems neuroscience. Foundation models for large-scale neural data represent shared latent structures of neural signals. However, extracting interpretable latent graph representations in foundation models remains challenging and unsolved. Here we explore latent graph learning in generative models of neural signals. By testing against numerical simulations of neural circuits with known ground-truth connectivity, we evaluate several hypotheses for explaining learned model weights. We discover modest alignment between extracted network representations and the underlying directed graphs and strong alignment in the co-input graph representations. These findings motivate paths towards incorporating graph-based geometric constraints in the construction of large-scale foundation models for neural data.
Authors: Manpreet Singh, Hassan Sajjad
Abstract: Quantization offers a practical solution to deploy LLMs in resource-constraint environments. However, its impact on internal representations remains understudied, raising questions about the reliability of quantized models. In this study, we employ a range of interpretability techniques to investigate how quantization affects model and neuron behavior. We analyze multiple LLMs under 4-bit and 8-bit quantization. Our findings reveal that the impact of quantization on model calibration is generally minor. Analysis of neuron activations indicates that the number of dead neurons, i.e., those with activation values close to 0 across the dataset, remains consistent regardless of quantization. In terms of neuron contribution to predictions, we observe that smaller full precision models exhibit fewer salient neurons, whereas larger models tend to have more, with the exception of Llama-2-7B. The effect of quantization on neuron redundancy varies across models. Overall, our findings suggest that effect of quantization may vary by model and tasks, however, we did not observe any drastic change which may discourage the use of quantization as a reliable model compression technique.
Authors: Baozhuo Su, Zhengxian Qu
Abstract: Regression under uncertainty is fundamental across science and engineering. We present an Anchored Mixture of Experts (Anchor-MoE), a model that handles both probabilistic and point regression. For simplicity, we use a tuned gradient-boosting model to furnish the anchor mean; however, any off-the-shelf point regressor can serve as the anchor. The anchor prediction is projected into a latent space, where a learnable metric-window kernel scores locality and a soft router dispatches each sample to a small set of mixture-density-network experts; the experts produce a heteroscedastic correction and predictive variance. We train by minimizing negative log-likelihood, and on a disjoint calibration split fit a post-hoc linear map on predicted means to improve point accuracy. On the theory side, assuming a H\"older smooth regression function of order~$\alpha$ and fixed Lipschitz partition-of-unity weights with bounded overlap, we show that Anchor-MoE attains the minimax-optimal $L^2$ risk rate $O\!\big(N^{-2\alpha/(2\alpha+d)}\big)$. In addition, the CRPS test generalization gap scales as $\widetilde{O}\!\Big(\sqrt{(\log(Mh)+P+K)/N}\Big)$; it is logarithmic in $Mh$ and scales as the square root in $P$ and $K$. Under bounded-overlap routing, $K$ can be replaced by $k$, and any dependence on a latent dimension is absorbed into $P$. Under uniformly bounded means and variances, an analogous $\widetilde{O}\!\big(\sqrt{(\log(Mh)+P+K)/N}\big)$ scaling holds for the test NLL up to constants. Empirically, across standard UCI regressions, Anchor-MoE consistently matches or surpasses the strong NGBoost baseline in RMSE and NLL; on several datasets it achieves new state-of-the-art probabilistic regression results on our benchmark suite. Code is available at https://github.com/BaozhuoSU/Probabilistic_Regression.
URLs: https://github.com/BaozhuoSU/Probabilistic_Regression.
Authors: Hadi Jahanshahi, Zheng H. Zhu
Abstract: This paper introduces Uncertainty Propagation Network (UPN), a novel family of neural differential equations that naturally incorporate uncertainty quantification into continuous-time modeling. Unlike existing neural ODEs that predict only state trajectories, UPN simultaneously model both state evolution and its associated uncertainty by parameterizing coupled differential equations for mean and covariance dynamics. The architecture efficiently propagates uncertainty through nonlinear dynamics without discretization artifacts by solving coupled ODEs for state and covariance evolution while enabling state-dependent, learnable process noise. The continuous-depth formulation adapts its evaluation strategy to each input's complexity, provides principled uncertainty quantification, and handles irregularly-sampled observations naturally. Experimental results demonstrate UPN's effectiveness across multiple domains: continuous normalizing flows (CNFs) with uncertainty quantification, time-series forecasting with well-calibrated confidence intervals, and robust trajectory prediction in both stable and chaotic dynamical systems.
Authors: Junhyun Lee, Veronika Thost, Bumsoo Kim, Jaewoo Kang, Tengfei Ma
Abstract: Message Passing Neural Networks (MPNNs) hold a key position in machine learning on graphs, but they struggle with unintended behaviors, such as over-smoothing and over-squashing, due to irregular data structures. The observation and formulation of these limitations have become foundational in constructing more informative graph representations. In this paper, we delve into the limitations of MPNNs, focusing on aspects that have previously been overlooked. Our observations reveal that even within a single layer, the information specific to an individual node can become significantly diluted. To delve into this phenomenon in depth, we present the concept of Over-dilution and formulate it with two dilution factors: intra-node dilution for attribute-level and inter-node dilution for node-level representations. We also introduce a transformer-based solution that alleviates over-dilution and complements existing node embedding methods like MPNNs. Our findings provide new insights and contribute to the development of informative representations. The implementation and supplementary materials are publicly available at https://github.com/LeeJunHyun/NATR.
Authors: Yannik Hahn, Jan Voets, Antonin Koenigsfeld, Hasan Tercan, Tobias Meisen
Abstract: Modern manufacturing relies heavily on fusion welding processes, including gas metal arc welding (GMAW). Despite significant advances in machine learning-based quality prediction, current models exhibit critical limitations when confronted with the inherent distribution shifts that occur in dynamic manufacturing environments. In this work, we extend the VQ-VAE Transformer architecture - previously demonstrating state-of-the-art performance in weld quality prediction - by leveraging its autoregressive loss as a reliable out-of-distribution (OOD) detection mechanism. Our approach exhibits superior performance compared to conventional reconstruction methods, embedding error-based techniques, and other established baselines. By integrating OOD detection with continual learning strategies, we optimize model adaptation, triggering updates only when necessary and thereby minimizing costly labeling requirements. We introduce a novel quantitative metric that simultaneously evaluates OOD detection capability while interpreting in-distribution performance. Experimental validation in real-world welding scenarios demonstrates that our framework effectively maintains robust quality prediction capabilities across significant distribution shifts, addressing critical challenges in dynamic manufacturing environments where process parameters frequently change. This research makes a substantial contribution to applied artificial intelligence by providing an explainable and at the same time adaptive solution for quality assurance in dynamic manufacturing processes - a crucial step towards robust, practical AI systems in the industrial environment.
Authors: Bicheng Wang, Junping Wang, Yibo Xue
Abstract: Industrial chain plays an increasingly important role in the sustainable development of national economy. However, as a typical complex network, data-driven deep learning is still in its infancy in describing and analyzing the resilience of complex networks, and its core is the lack of a theoretical framework to describe the system dynamics. In this paper, we propose a physically informative neural symbolic approach to describe the evolutionary dynamics of complex networks for resilient prediction. The core idea is to learn the dynamics of the activity state of physical entities and integrate it into the multi-layer spatiotemporal co-evolution network, and use the physical information method to realize the joint learning of physical symbol dynamics and spatiotemporal co-evolution topology, so as to predict the industrial chain resilience. The experimental results show that the model can obtain better results and predict the elasticity of the industry chain more accurately and effectively, which has certain practical significance for the development of the industry.
Authors: Guangyu Nie, Yang Jiao, Yi Ren
Abstract: Effective properties of composite materials are defined as the ensemble average of property-specific PDE solutions over the underlying microstructure distributions. Traditionally, predicting such properties can be done by solving PDEs derived from microstructure samples or building data-driven models that directly map microstructure samples to properties. The former has a higher running cost, but provides explainable sensitivity information that may guide material design; the latter could be more cost-effective if the data overhead is amortized, but its learned sensitivities are often less explainable. With a focus on properties governed by linear self-adjoint PDEs (e.g., Laplace, Helmholtz, and Maxwell curl-curl) defined on bi-phase microstructures, we propose a structure-property model that is both cost-effective and explainable. Our method is built on top of the strong contrast expansion (SCE) formalism, which analytically maps $N$-point correlations of an unbounded random field to its effective properties. Since real-world material samples have finite sizes and analytical PDE kernels are not always available, we propose Neural Contrast Expansion (NCE), an SCE-inspired architecture to learn surrogate PDE kernels from structure-property data. For static conduction and electromagnetic wave propagation cases, we show that NCE models reveal accurate and insightful sensitivity information useful for material design. Compared with other PDE kernel learning methods, our method does not require measurements about the PDE solution fields, but rather only requires macroscopic property measurements that are more accessible in material development contexts.
Authors: Chaolong Ying, Yinan Zhang, Lei Zhang, Jiazhuang Wang, Shujun Jia, Tianshu Yu
Abstract: Map-to-map matching is a critical task for aligning spatial data across heterogeneous sources, yet it remains challenging due to the lack of ground truth correspondences, sparse node features, and scalability demands. In this paper, we propose an unsupervised graph-based framework that addresses these challenges through three key innovations. First, our method is an unsupervised learning approach that requires no training data, which is crucial for large-scale map data where obtaining labeled training samples is challenging. Second, we introduce pseudo coordinates that capture the relative spatial layout of nodes within each map, which enhances feature discriminability and enables scale-invariant learning. Third, we design an mechanism to adaptively balance feature and geometric similarity, as well as a geometric-consistent loss function, ensuring robustness to noisy or incomplete coordinate data. At the implementation level, to handle large-scale maps, we develop a tile-based post-processing pipeline with overlapping regions and majority voting, which enables parallel processing while preserving boundary coherence. Experiments on real-world datasets demonstrate that our method achieves state-of-the-art accuracy in matching tasks, surpassing existing methods by a large margin, particularly in high-noise and large-scale scenarios. Our framework provides a scalable and practical solution for map alignment, offering a robust and efficient alternative to traditional approaches.
Authors: Cody Grogan, Som Dhulipala, Mauricio Tano, Izabela Gutowska, Som Dutta
Abstract: Neural-Network (NN) based turbulence closures have been developed for being used as pre-trained surrogates for traditional turbulence closures, with the aim to increase computational efficiency and prediction accuracy of CFD simulations. The bottleneck to the widespread adaptation of these ML-based closures is the relative lack of uncertainty quantification (UQ) for these models. Especially, quantifying uncertainties associated with out-of-training inputs, that is when the ML-based turbulence closures are queried on inputs outside their training data regime. In the current paper, a published algebraic turbulence closure1 has been utilized to compare the quality of epistemic UQ between three NN-based methods and Gaussian Process (GP). The three NN-based methods explored are Deep Ensembles (DE), Monte-Carlo Dropout (MCD), and Stochastic Variational Inference (SVI). In the in-training results, we find the exact GP performs the best in accuracy with a Root Mean Squared Error (RMSE) of $2.14 \cdot 10^{-5}$ followed by the DE with an RMSE of $4.59 \cdot 10^{-4}$. Next, the paper discusses the performance of the four methods for quantifying out-of-training uncertainties. For performance, the Exact GP yet again is the best in performance, but has similar performance to the DE in the out-of-training regions. In UQ accuracy for the out-of-training case, SVI and DE hold the best miscalibration error for one of the cases. However, the DE performs the best in Negative Log-Likelihood for both out-of-training cases. We observe that for the current problem, in terms of accuracy GP > DE > SV I > MCD. The DE results are relatively robust and provide intuitive UQ estimates, despite performing naive ensembling. In terms of computational cost, the GP is significantly higher than the NN-based methods with a $O(n^3)$ computational complexity for each training step
Authors: Mohsen Sheibanian, Pouya Shaeri, Alimohammad Beigi, Ryan T. Woo, Aryan Keluskar
Abstract: Deep neural networks are increasingly bottlenecked by the cost of optimization, both in terms of GPU memory and compute time. Existing acceleration techniques, such as mixed precision, second-order methods, and batch size scaling, are typically used in isolation. We present Tri-Accel, a unified optimization framework that co-adapts three acceleration strategies along with adaptive parameters during training: (1) Precision-Adaptive Updates that dynamically assign mixed-precision levels to layers based on curvature and gradient variance; (2) Sparse Second-Order Signals that exploit Hessian/Fisher sparsity patterns to guide precision and step size decisions; and (3) Memory-Elastic Batch Scaling that adjusts batch size in real time according to VRAM availability. On CIFAR-10 with ResNet-18 and EfficientNet-B0, Tri-Accel achieves up to 9.9% reduction in training time and 13.3% lower memory usage, while improving accuracy by +1.1 percentage points over FP32 baselines. Tested on CIFAR-10/100, our approach demonstrates adaptive learning behavior, with efficiency gradually improving over the course of training as the system learns to allocate resources more effectively. Compared to static mixed-precision training, Tri-Accel maintains 78.1% accuracy while reducing memory footprint from 0.35GB to 0.31GB on standard hardware. The framework is implemented with custom Triton kernels, whose hardware-aware adaptation enables automatic optimization without manual hyperparameter tuning, making it practical for deployment across diverse computational environments. This work demonstrates how algorithmic adaptivity and hardware awareness can be combined to improve scalability in resource-constrained settings, paving the way for more efficient neural network training on edge devices and cost-sensitive cloud deployments.
Authors: Sadman Mohammad Nasif, Md Abrar Jahin, M. F. Mridha
Abstract: The growing adoption of home banking systems has heightened the risk of cyberfraud, necessitating fraud detection mechanisms that are not only accurate but also fair and explainable. While AI models have shown promise in this domain, they face key limitations, including computational inefficiency, the interpretability challenges of spiking neural networks (SNNs), and the complexity and convergence instability of hyper-heuristic reinforcement learning (RL)-based hyperparameter optimization. To address these issues, we propose a novel framework that integrates a Cortical Spiking Network with Population Coding (CSNPC) and a Reinforcement-Guided Hyper-Heuristic Optimizer for Spiking Systems (RHOSS). The CSNPC, a biologically inspired SNN, employs population coding for robust classification, while RHOSS uses Q-learning to dynamically select low-level heuristics for hyperparameter optimization under fairness and recall constraints. Embedded within the Modular Supervisory Framework for Spiking Network Training and Interpretation (MoSSTI), the system incorporates explainable AI (XAI) techniques, specifically, saliency-based attribution and spike activity profiling, to increase transparency. Evaluated on the Bank Account Fraud (BAF) dataset suite, our model achieves a $90.8\%$ recall at a strict $5\%$ false positive rate (FPR), outperforming state-of-the-art spiking and non-spiking models while maintaining over $98\%$ predictive equality across key demographic attributes. The explainability module further confirms that saliency attributions align with spiking dynamics, validating interpretability. These results demonstrate the potential of combining population-coded SNNs with reinforcement-guided hyper-heuristics for fair, transparent, and high-performance fraud detection in real-world financial applications.
Authors: Junxuan Wang, Xuyang Ge, Wentao Shu, Zhengfu He, Xipeng Qiu
Abstract: While transformer models are widely believed to operate in high-dimensional hidden spaces, we show that attention outputs are confined to a surprisingly low-dimensional subspace, where about 60\% of the directions account for 99\% of the variance--a phenomenon that is induced by the attention output projection matrix and consistently observed across diverse model families and datasets. Critically, we find this low-rank structure as a fundamental cause of the prevalent dead feature problem in sparse dictionary learning, where it creates a mismatch between randomly initialized features and the intrinsic geometry of the activation space. Building on this insight, we propose a subspace-constrained training method for sparse autoencoders (SAEs), initializing feature directions into the active subspace of activations. Our approach reduces dead features from 87\% to below 1\% in Attention Output SAEs with 1M features, and can further extend to other sparse dictionary learning methods. Our findings provide both new insights into the geometry of attention and practical tools for improving sparse dictionary learning in large language models.
Authors: Tao Liu, Xuehe Wang
Abstract: Handling data staleness remains a significant challenge in federated learning with highly time-sensitive tasks, where data is generated continuously and data staleness largely affects model performance. Although recent works attempt to optimize data staleness by determining local data update frequency or client selection strategy, none of them explore taking both data staleness and data volume into consideration. In this paper, we propose DUFL(Data Updating in Federated Learning), an incentive mechanism featuring an innovative local data update scheme manipulated by three knobs: the server's payment, outdated data conservation rate, and clients' fresh data collection volume, to coordinate staleness and volume of local data for best utilities. To this end, we introduce a novel metric called DoS(the Degree of Staleness) to quantify data staleness and conduct a theoretic analysis illustrating the quantitative relationship between DoS and model performance. We model DUFL as a two-stage Stackelberg game with dynamic constraint, deriving the optimal local data update strategy for each client in closed-form and the approximately optimal strategy for the server. Experimental results on real-world datasets demonstrate the significant performance of our approach.
Authors: Lei Jiang, Wen Ge, Niels Cariou-Kotlarek, Mingxuan Yi, Po-Yu Chen, Lingyi Yang, Francois Buet-Golfouse, Gaurav Mittal, Hao Ni
Abstract: Diffusion models have achieved state-of-the-art results in generative modelling but remain computationally intensive at inference time, often requiring thousands of discretization steps. To this end, we propose Sig-DEG (Signature-based Differential Equation Generator), a novel generator for distilling pre-trained diffusion models, which can universally approximate the backward diffusion process at a coarse temporal resolution. Inspired by high-order approximations of stochastic differential equations (SDEs), Sig-DEG leverages partial signatures to efficiently summarize Brownian motion over sub-intervals and adopts a recurrent structure to enable accurate global approximation of the SDE solution. Distillation is formulated as a supervised learning task, where Sig-DEG is trained to match the outputs of a fine-resolution diffusion model on a coarse time grid. During inference, Sig-DEG enables fast generation, as the partial signature terms can be simulated exactly without requiring fine-grained Brownian paths. Experiments demonstrate that Sig-DEG achieves competitive generation quality while reducing the number of inference steps by an order of magnitude. Our results highlight the effectiveness of signature-based approximations for efficient generative modeling.
Authors: Yang Zhou, Sunzhu Li, Shunyu Liu, Wenkai Fang, Jiale Zhao, Jingwen Yang, Jianwei Lv, Kongcheng Zhang, Yihe Zhou, Hengtong Lu, Wei Chen, Yan Xie, Mingli Song
Abstract: Recent advances in Large Language Models (LLMs) have underscored the potential of Reinforcement Learning (RL) to facilitate the emergence of reasoning capabilities. Despite the encouraging results, a fundamental dilemma persists as RL improvement relies on learning from high-quality samples, yet the exploration for such samples remains bounded by the inherent limitations of LLMs. This, in effect, creates an undesirable cycle in which what cannot be explored cannot be learned. In this work, we propose Rubric-Scaffolded Reinforcement Learning (RuscaRL), a novel instructional scaffolding framework designed to break the exploration bottleneck for general LLM reasoning. Specifically, RuscaRL introduces checklist-style rubrics as (1) explicit scaffolding for exploration during rollout generation, where different rubrics are provided as external guidance within task instructions to steer diverse high-quality responses. This guidance is gradually decayed over time, encouraging the model to internalize the underlying reasoning patterns; (2) verifiable rewards for exploitation during model training, where we can obtain robust LLM-as-a-Judge scores using rubrics as references, enabling effective RL on general reasoning tasks. Extensive experiments demonstrate the superiority of the proposed RuscaRL across various benchmarks, effectively expanding reasoning boundaries under the best-of-N evaluation. Notably, RuscaRL significantly boosts Qwen-2.5-7B-Instruct from 23.6 to 50.3 on HealthBench-500, surpassing GPT-4.1. Furthermore, our fine-tuned variant on Qwen3-30B-A3B-Instruct achieves 61.1 on HealthBench-500, outperforming leading LLMs including OpenAI-o3.
Authors: Manan Gupta, Dhruv Kumar
Abstract: Neural networks often contain polysemantic neurons that respond to multiple, sometimes unrelated, features, complicating mechanistic interpretability. We introduce the Polysemanticity Index (PSI), a null-calibrated metric that quantifies when a neuron's top activations decompose into semantically distinct clusters. PSI multiplies three independently calibrated components: geometric cluster quality (S), alignment to labeled categories (Q), and open-vocabulary semantic distinctness via CLIP (D). On a pretrained ResNet-50 evaluated with Tiny-ImageNet images, PSI identifies neurons whose activation sets split into coherent, nameable prototypes, and reveals strong depth trends: later layers exhibit substantially higher PSI than earlier layers. We validate our approach with robustness checks (varying hyperparameters, random seeds, and cross-encoder text heads), breadth analyses (comparing class-only vs. open-vocabulary concepts), and causal patch-swap interventions. In particular, aligned patch replacements increase target-neuron activation significantly more than non-aligned, random, shuffled-position, or ablate-elsewhere controls. PSI thus offers a principled and practical lever for discovering, quantifying, and studying polysemantic units in neural networks.
Authors: Fu-Chieh Chang, Yu-Ting Lee, Pei-Yuan Wu
Abstract: Reflection, the ability of large language models (LLMs) to evaluate and revise their own reasoning, has been widely used to improve performance on complex reasoning tasks. Yet, most prior work emphasizes designing reflective prompting strategies or reinforcement learning objectives, leaving the inner mechanisms of reflection underexplored. In this paper, we investigate reflection through the lens of latent directions in model activations. We propose a methodology based on activation steering to characterize how instructions with different reflective intentions: no reflection, intrinsic reflection, and triggered reflection. By constructing steering vectors between these reflection levels, we demonstrate that (1) new reflection-inducing instructions can be systematically identified, (2) reflective behavior can be directly enhanced or suppressed through activation interventions, and (3) suppressing reflection is considerably easier than stimulating it. Experiments on GSM8k-adv with Qwen2.5-3B and Gemma3-4B reveal clear stratification across reflection levels, and steering interventions confirm the controllability of reflection. Our findings highlight both opportunities (e.g., reflection-enhancing defenses) and risks (e.g., adversarial inhibition of reflection in jailbreak attacks). This work opens a path toward mechanistic understanding of reflective reasoning in LLMs.
Authors: Dhruv Sarkar, Samrat Mukhopadhyay, Abhishek Sinha
Abstract: We study an online learning problem with long-term budget constraints in the adversarial setting. In this problem, at each round $t$, the learner selects an action from a convex decision set, after which the adversary reveals a cost function $f_t$ and a resource consumption function $g_t$. The cost and consumption functions are assumed to be $\alpha$-approximately convex - a broad class that generalizes convexity and encompasses many common non-convex optimization problems, including DR-submodular maximization, Online Vertex Cover, and Regularized Phase Retrieval. The goal is to design an online algorithm that minimizes cumulative cost over a horizon of length $T$ while approximately satisfying a long-term budget constraint of $B_T$. We propose an efficient first-order online algorithm that guarantees $O(\sqrt{T})$ $\alpha$-regret against the optimal fixed feasible benchmark while consuming at most $O(B_T \log T)+ \tilde{O}(\sqrt{T})$ resources in both full-information and bandit feedback settings. In the bandit feedback setting, our approach yields an efficient solution for the $\texttt{Adversarial Bandits with Knapsacks}$ problem with improved guarantees. We also prove matching lower bounds, demonstrating the tightness of our results. Finally, we characterize the class of $\alpha$-approximately convex functions and show that our results apply to a broad family of problems.
Authors: Maurizio Diaz
Abstract: A bottleneck for long-context LLM inference is the linearly growing KV cache. Recent work has proposed CARTRIDGES, an approach which leverages offline compute to train a much smaller KV cache than is typically required for a full document (up to 40x less memory usage at inference time). In this paper, we present the first mechanistic exploration of the learned CARTRIDGE key-value cache structure. In particular, we propose that (1) CARTRIDGE keys act as stable, shareable retrieval routers for the compressed corpora and (2) most of the learned compression occurs within the CARTRIDGE value vectors. We present empirical evidence of our routing theory across tasks, model families, and model sizes; for example, we can ablate the learned CARTRIDGE key vectors between tasks with little performance loss. Finally, we propose a slight improvement in initialization called Sampled Chunk Initialization (SCI). We suggest that SCI can lead to faster CARTRIDGE convergence than previously demonstrated in the literature. Our findings lay the groundwork for broader empirical study of CARTRIDGE training optimization which may be crucial for further scaling.
Authors: Kiran Madhusudhanan, Vijaya Krishna Yalavarthi, Jonas Sonntag, Maximilian Stubbemann, Lars Schmidt-Thieme
Abstract: Tabular regression is a well-studied problem with numerous industrial applications, yet most existing approaches focus on point estimation, often leading to overconfident predictions. This issue is particularly critical in industrial automation, where trustworthy decision-making is essential. Probabilistic regression models address this challenge by modeling prediction uncertainty. However, many conventional methods assume a fixed-shape distribution (typically Gaussian), and resort to estimating distribution parameters. This assumption is often restrictive, as real-world target distributions can be highly complex. To overcome this limitation, we introduce TabResFlow, a Normalizing Spline Flow model designed specifically for univariate tabular regression, where commonly used simple flow networks like RealNVP and Masked Autoregressive Flow (MAF) are unsuitable. TabResFlow consists of three key components: (1) An MLP encoder for each numerical feature. (2) A fully connected ResNet backbone for expressive feature extraction. (3) A conditional spline-based normalizing flow for flexible and tractable density estimation. We evaluate TabResFlow on nine public benchmark datasets, demonstrating that it consistently surpasses existing probabilistic regression models on likelihood scores. Our results demonstrate 9.64% improvement compared to the strongest probabilistic regression model (TreeFlow), and on average 5.6 times speed-up in inference time compared to the strongest deep learning alternative (NodeFlow). Additionally, we validate the practical applicability of TabResFlow in a real-world used car price prediction task under selective regression. To measure performance in this setting, we introduce a novel Area Under Risk Coverage (AURC) metric and show that TabResFlow achieves superior results across this metric.
Authors: Prashant Gupta
Abstract: This thesis develops computational methods in similarity-preserving hashing, classification, and cancer genomics. Standard space partitioning-based hashing relies on Binary Search Trees (BSTs), but their exponential growth and sparsity hinder efficiency. To overcome this, we introduce Compressed BST of Inverted hash tables (ComBI), which enables fast approximate nearest-neighbor search with reduced memory. On datasets of up to one billion samples, ComBI achieves 0.90 precision with 4X-296X speed-ups over Multi-Index Hashing, and also outperforms Cellfishing.jl on single-cell RNA-seq searches with 2X-13X gains. Building on hashing structures, we propose Guided Random Forest (GRAF), a tree-based ensemble classifier that integrates global and local partitioning, bridging decision trees and boosting while reducing generalization error. Across 115 datasets, GRAF delivers competitive or superior accuracy, and its unsupervised variant (uGRAF) supports guided hashing and importance sampling. We show that GRAF and ComBI can be used to estimate per-sample classifiability, which enables scalable prediction of cancer patient survival. To address challenges in interpreting mutations, we introduce Continuous Representation of Codon Switches (CRCS), a deep learning framework that embeds genetic changes into numerical vectors. CRCS allows identification of somatic mutations without matched normals, discovery of driver genes, and scoring of tumor mutations, with survival prediction validated in bladder, liver, and brain cancers. Together, these methods provide efficient, scalable, and interpretable tools for large-scale data analysis and biomedical applications.
Authors: Haitao Tian, Argyrios Zolotas, Miguel Arana-Catania
Abstract: In this study, we explore the use of Convolutional Neural Networks for improving train speed estimation accuracy, addressing the complex challenges of modern railway systems. We investigate three CNN architectures - single-branch 2D, single-branch 1D, and multiple-branch models - and compare them with the Adaptive Kalman Filter. We analyse their performance using simulated train operation datasets with and without Wheel Slide Protection activation. Our results reveal that CNN-based approaches, especially the multiple-branch model, demonstrate superior accuracy and robustness compared to traditional methods, particularly under challenging operational conditions. These findings highlight the potential of deep learning techniques to enhance railway safety and operational efficiency by more effectively capturing intricate patterns in complex transportation datasets.
Authors: Lingkai Kong, Haotian Sun, Yuchen Zhuang, Haorui Wang, Wenhao Mu, Chao Zhang
Abstract: Graph neural networks (GNNs) are powerful tools on graph data. However, their predictions are mis-calibrated and lack interpretability, limiting their adoption in critical applications. To address this issue, we propose a new uncertainty-aware and interpretable graph classification model that combines graph functional neural process and graph generative model. The core of our method is to assume a set of latent rationales which can be mapped to a probabilistic embedding space; the predictive distribution of the classifier is conditioned on such rationale embeddings by learning a stochastic correlation matrix. The graph generator serves to decode the graph structure of the rationales from the embedding space for model interpretability. For efficient model training, we adopt an alternating optimization procedure which mimics the well known Expectation-Maximization (EM) algorithm. The proposed method is general and can be applied to any existing GNN architecture. Extensive experiments on five graph classification datasets demonstrate that our framework outperforms state-of-the-art methods in both uncertainty quantification and GNN interpretability. We also conduct case studies to show that the decoded rationale structure can provide meaningful explanations.
Authors: Diksha Gupta, Nirupam Gupta, Chuan Xu, Giovanni Neglia
Abstract: Distributed learning (DL) enables scalable model training over decentralized data, but remains challenged by Byzantine faults and high communication costs. While both issues have been studied extensively in isolation, their interaction is less explored. Prior work shows that naively combining communication compression with Byzantine-robust aggregation degrades resilience to faulty nodes (or workers). The state-of-the-art algorithm, namely Byz-DASHA-PAGE [29], makes use of the momentum variance reduction scheme to mitigate the detrimental impact of compression noise on Byzantine-robustness. We propose a new algorithm, named RoSDHB, that integrates the classic Polyak's momentum with a new coordinated compression mechanism. We show that RoSDHB performs comparably to Byz-DASHA-PAGE under the standard (G, B)-gradient dissimilarity heterogeneity model, while it relies on fewer assumptions. In particular, we only assume Lipschitz smoothness of the average loss function of the honest workers, in contrast to [29]that additionally assumes a special smoothness of bounded global Hessian variance. Empirical results on benchmark image classification task show that RoSDHB achieves strong robustness with significant communication savings.
Authors: Nishant Gavhane, Arush Mehrotra, Rohit Chawla, Peter Proenca
Abstract: The deployment of large-scale Mixture-of-Experts (MoE) models on edge devices presents significant challenges due to memory constraints. While MoE architectures enable efficient utilization of computational resources by activating only a subset of experts per inference, they require careful memory management to operate efficiently in resource-constrained environments. Traditional heuristic-based expert caching strategies such as MoE-Infinity struggle to maintain high cache hit rates as models parameters scale. In this work, we introduce MoE-Beyond, a learning-based expert activation predictor trained to predict expert activations during autoregressive decoding. By framing the task as a multi-label sequence prediction problem, we train a lightweight transformer model on 66 million expert activation traces extracted from LDJnr-Puffin dataset [5] using DeepSeek-V2-Chat-Lite MoE. Our predictor generalizes effectively across unseen prompts from WebGLM-QA dataset [6], achieving 97.5% accuracy and an 86.6% F1-score. Simulation results show that MoE-Beyond improves GPU cache hit rate from 17% to 72% when only 10% of experts fit in GPU cache, outperforming heuristic baselines.
Authors: Nanfei Jiang, Hoi-To Wai, Mahnoosh Alizadeh
Abstract: This paper considers a finite-sum optimization problem under first-order queries and investigates the benefits of strategic querying on stochastic gradient-based methods compared to uniform querying strategy. We first introduce Oracle Gradient Querying (OGQ), an idealized algorithm that selects one user's gradient yielding the largest possible expected improvement (EI) at each step. However, OGQ assumes oracle access to the gradients of all users to make such a selection, which is impractical in real-world scenarios. To address this limitation, we propose Strategic Gradient Querying (SGQ), a practical algorithm that has better transient-state performance than SGD while making only one query per iteration. For smooth objective functions satisfying the Polyak-Lojasiewicz condition, we show that under the assumption of EI heterogeneity, OGQ enhances transient-state performance and reduces steady-state variance, while SGQ improves transient-state performance over SGD. Our numerical experiments validate our theoretical findings.
Authors: Meysam Shirdel Bilehsavar, Razieh Ghaedi, Samira Seyed Taheri, Xinqi Fan, Christian O'Reilly
Abstract: Clustering algorithms are widely used in various applications, with density-based methods such as Density-Based Spatial Clustering of Applications with Noise (DBSCAN) being particularly prominent. These algorithms identify clusters in high-density regions while treating sparser areas as noise. However, reliance on user-defined parameters often poses optimization challenges that require domain expertise. This paper presents a novel density-based clustering method inspired by the concept of selective attention, which minimizes the need for user-defined parameters under standard conditions. Initially, the algorithm operates without requiring user-defined parameters. If parameter adjustment is needed, the method simplifies the process by introducing a single integer parameter that is straightforward to tune. The approach computes a threshold to filter out the most sparsely distributed points and outliers, forms a preliminary cluster structure, and then reintegrates the excluded points to finalize the results. Experimental evaluations on diverse data sets highlight the accessibility and robust performance of the method, providing an effective alternative for density-based clustering tasks.
Authors: Jack Youstra, Mohammed Mahfoud, Yang Yan, Henry Sleight, Ethan Perez, Mrinank Sharma
Abstract: Large language model fine-tuning APIs enable widespread model customization, yet pose significant safety risks. Recent work shows that adversaries can exploit access to these APIs to bypass model safety mechanisms by encoding harmful content in seemingly harmless fine-tuning data, evading both human monitoring and standard content filters. We formalize the fine-tuning API defense problem, and introduce the Cipher Fine-tuning Robustness benchmark (CIFR), a benchmark for evaluating defense strategies' ability to retain model safety in the face of cipher-enabled attackers while achieving the desired level of fine-tuning functionality. We include diverse cipher encodings and families, with some kept exclusively in the test set to evaluate for generalization across unseen ciphers and cipher families. We then evaluate different defenses on the benchmark and train probe monitors on model internal activations from multiple fine-tunes. We show that probe monitors achieve over 99% detection accuracy, generalize to unseen cipher variants and families, and compare favorably to state-of-the-art monitoring approaches. We open-source CIFR and the code to reproduce our experiments to facilitate further research in this critical area. Code and data are available online https://github.com/JackYoustra/safe-finetuning-api
Authors: Yajat Yadav, Jathin Korrapati, Patrick Mendoza
Abstract: Orthogonal gradient descent has emerged as a powerful method for continual learning tasks. However, its Euclidean projections overlook the underlying information-geometric structure of the space of distributions parametrized by neural networks, which can lead to suboptimal convergence in learning tasks. To counteract this, we combine it with the idea of the natural gradient and present ONG (Orthogonal Natural Gradient Descent). ONG preconditions each new task gradient with an efficient EKFAC approximation of the inverse Fisher information matrix, yielding updates that follow the steepest descent direction under a Riemannian metric. To preserve performance on previously learned tasks, ONG projects these natural gradients onto the orthogonal complement of prior task gradients. We provide a theoretical justification for this procedure, introduce the ONG algorithm, and benchmark its performance on the Permuted and Rotated MNIST datasets. All code for our experiments/reproducibility can be found at https://github.com/yajatyadav/orthogonal-natural-gradient.
URLs: https://github.com/yajatyadav/orthogonal-natural-gradient.
Authors: Jeng-Lin Li, Ming-Ching Chang, Wei-Chao Chen
Abstract: Out-of-distribution (OOD) detection ensures safe and reliable model deployment. Contemporary OOD algorithms using geometry projection can detect OOD or adversarial samples from clean in-distribution (ID) samples. However, this setting regards adversarial ID samples as OOD, leading to incorrect OOD predictions. Existing efforts on OOD detection with ID and OOD data under attacks are minimal. In this paper, we develop a robust OOD detection method that distinguishes adversarial ID samples from OOD ones. The sharp loss landscape created by adversarial training hinders model convergence, impacting the latent embedding quality for OOD score calculation. Therefore, we introduce a {\bf Sharpness-aware Geometric Defense (SaGD)} framework to smooth out the rugged adversarial loss landscape in the projected latent geometry. Enhanced geometric embedding convergence enables accurate ID data characterization, benefiting OOD detection against adversarial attacks. We use Jitter-based perturbation in adversarial training to extend the defense ability against unseen attacks. Our SaGD framework significantly improves FPR and AUC over the state-of-the-art defense approaches in differentiating CIFAR-100 from six other OOD datasets under various attacks. We further examine the effects of perturbations at various adversarial training levels, revealing the relationship between the sharp loss landscape and adversarial OOD detection.
Authors: Leon Dimitrov
Abstract: Graphs have become a central representation in machine learning for capturing relational and structured data across various domains. Traditional graph neural networks often struggle to capture long-range dependencies between nodes due to their local structure. Graph transformers overcome this by using attention mechanisms that allow nodes to exchange information globally. However, there are two types of attention in graph transformers: dense and sparse. In this paper, we compare these two attention mechanisms, analyze their trade-offs, and highlight when to use each. We also outline current challenges and problems in designing attention for graph transformers.
Authors: Hikaru Tsujimura, Arush Tagade
Abstract: Large Language Models (LLMs) often display overconfidence, presenting information with unwarranted certainty in high-stakes contexts. We investigate the internal basis of this behavior via mechanistic interpretability. Using open-sourced Llama 3.2 models fine-tuned on human annotated assertiveness datasets, we extract residual activations across all layers, and compute similarity metrics to localize assertive representations. Our analysis identifies layers most sensitive to assertiveness contrasts and reveals that high-assertive representations decompose into two orthogonal sub-components of emotional and logical clusters-paralleling the dual-route Elaboration Likelihood Model in Psychology. Steering vectors derived from these sub-components show distinct causal effects: emotional vectors broadly influence prediction accuracy, while logical vectors exert more localized effects. These findings provide mechanistic evidence for the multi-component structure of LLM assertiveness and highlight avenues for mitigating overconfident behavior.
Authors: Hao Wen, Xinrui Wu, Yi Sun, Feifei Zhang, Liye Chen, Jie Wang, Yunxin Liu, Ya-Qin Zhang, Yuanchun Li
Abstract: Recent advancements in Large Language Models (LLMs) have leveraged increased test-time computation to enhance reasoning capabilities, a strategy that, while effective, incurs significant latency and resource costs, limiting their applicability in real-world time-constrained or cost-sensitive scenarios. This paper introduces BudgetThinker, a novel framework designed to empower LLMs with budget-aware reasoning, enabling precise control over the length of their thought processes. We propose a methodology that periodically inserts special control tokens during inference to continuously inform the model of its remaining token budget. This approach is coupled with a comprehensive two-stage training pipeline, beginning with Supervised Fine-Tuning (SFT) to familiarize the model with budget constraints, followed by a curriculum-based Reinforcement Learning (RL) phase that utilizes a length-aware reward function to optimize for both accuracy and budget adherence. We demonstrate that BudgetThinker significantly surpasses strong baselines in maintaining performance across a variety of reasoning budgets on challenging mathematical benchmarks. Our method provides a scalable and effective solution for developing efficient and controllable LLM reasoning, making advanced models more practical for deployment in resource-constrained and real-time environments.
Authors: Kaiwen Zuo, Zelin Liu, Raman Dutt, Ziyang Wang, Zhongtian Sun, Yeming Wang, Fan Mo, Pietro Li\`o
Abstract: Large Vision-Language Models (LVLMs) augmented with Retrieval-Augmented Generation (RAG) are increasingly employed in medical AI to enhance factual grounding through external clinical image-text retrieval. However, this reliance creates a significant attack surface. We propose MedThreatRAG, a novel multimodal poisoning framework that systematically probes vulnerabilities in medical RAG systems by injecting adversarial image-text pairs. A key innovation of our approach is the construction of a simulated semi-open attack environment, mimicking real-world medical systems that permit periodic knowledge base updates via user or pipeline contributions. Within this setting, we introduce and emphasize Cross-Modal Conflict Injection (CMCI), which embeds subtle semantic contradictions between medical images and their paired reports. These mismatches degrade retrieval and generation by disrupting cross-modal alignment while remaining sufficiently plausible to evade conventional filters. While basic textual and visual attacks are included for completeness, CMCI demonstrates the most severe degradation. Evaluations on IU-Xray and MIMIC-CXR QA tasks show that MedThreatRAG reduces answer F1 scores by up to 27.66% and lowers LLaVA-Med-1.5 F1 rates to as low as 51.36%. Our findings expose fundamental security gaps in clinical RAG systems and highlight the urgent need for threat-aware design and robust multimodal consistency checks. Finally, we conclude with a concise set of guidelines to inform the safe development of future multimodal medical RAG systems.
Authors: Xing Wei, Yuqi Ouyang
Abstract: With the rapidly increased number of vehicles in urban areas, existing road infrastructure struggles to accommodate modern traffic demands, resulting in the issue of congestion. This highlights the importance of efficient path planning strategies. However, most recent navigation models focus solely on deterministic or time-dependent networks, while overlooking the correlations and the stochastic nature of traffic flows. In this work, we address the reliable shortest path problem within stochastic transportation networks under certain dependencies. We propose a path planning solution that integrates the decision Transformer with the Generalized Policy Gradient (GPG) framework. Based on the decision Transformer's capability to model long-term dependencies, our proposed solution improves the accuracy and stability of path decisions. Experimental results on the Sioux Falls Network (SFN) demonstrate that our approach outperforms previous baselines in terms of on-time arrival probability, providing more accurate path planning solutions.
Authors: Xiaomeng Fan, Yuwei Wu, Zhi Gao, Mehrtash Harandi, Yunde Jia
Abstract: Hyperbolic neural networks (HNNs) have demonstrated notable efficacy in representing real-world data with hierarchical structures via exploiting the geometric properties of hyperbolic spaces characterized by negative curvatures. Curvature plays a crucial role in optimizing HNNs. Inappropriate curvatures may cause HNNs to converge to suboptimal parameters, degrading overall performance. So far, the theoretical foundation of the effect of curvatures on HNNs has not been developed. In this paper, we derive a PAC-Bayesian generalization bound of HNNs, highlighting the role of curvatures in the generalization of HNNs via their effect on the smoothness of the loss landscape. Driven by the derived bound, we propose a sharpness-aware curvature learning method to smooth the loss landscape, thereby improving the generalization of HNNs. In our method, we design a scope sharpness measure for curvatures, which is minimized through a bi-level optimization process. Then, we introduce an implicit differentiation algorithm that efficiently solves the bi-level optimization by approximating gradients of curvatures. We present the approximation error and convergence analyses of the proposed method, showing that the approximation error is upper-bounded, and the proposed method can converge by bounding gradients of HNNs. Experiments on four settings: classification, learning from long-tailed data, learning from noisy data, and few-shot learning show that our method can improve the performance of HNNs.
Authors: Wenjie Bao, Jian Lou, Yuke Hu, Xiaochen Li, Zhihao Liu, Jiaqi Liu, Zhan Qin, Kui Ren
Abstract: Transformer has become fundamental to a vast series of pre-trained large models that have achieved remarkable success across diverse applications. Machine unlearning, which focuses on efficiently removing specific data influences to comply with privacy regulations, shows promise in restricting updates to influence-critical parameters. However, existing parameter-efficient unlearning methods are largely devised in a module-oblivious manner, which tends to inaccurately identify these parameters and leads to inferior unlearning performance for Transformers. In this paper, we propose {\tt MAPE-Unlearn}, a module-aware parameter-efficient machine unlearning approach that uses a learnable pair of masks to pinpoint influence-critical parameters in the heads and filters of Transformers. The learning objective of these masks is derived by desiderata of unlearning and optimized through an efficient algorithm featured by a greedy search with a warm start. Extensive experiments on various Transformer models and datasets demonstrate the effectiveness and robustness of {\tt MAPE-Unlearn} for unlearning.
Authors: Aviral Dhingra
Abstract: Deep neural networks often contain far more parameters than training examples, yet they still manage to generalize well in practice. Classical complexity measures such as VC-dimension or PAC-Bayes bounds usually become vacuous in this overparameterized regime, offering little explanation for the empirical success of models like Transformers. In this work, I explore an alternative notion of capacity for attention-based models, based on the effective rank of their attention matrices. The intuition is that, although the parameter count is enormous, the functional dimensionality of attention is often much lower. I show that this quantity leads to a generalization bound whose dependence on sample size matches empirical scaling laws observed in large language models, up to logarithmic factors. While the analysis is not a complete theory of overparameterized learning, it provides evidence that spectral properties of attention, rather than raw parameter counts, may be the right lens for understanding why these models generalize.
Authors: Mohammad Amin Esabat, Saeed Jaamei, Fatemeh Asadi
Abstract: . Predicting and calculating the aerodynamic coefficients of airfoils near the ground with CFD software requires much time. However, the availability of data from CFD simulation results and the development of new neural network methods have made it possible to present the simulation results using methods like VGG, a CCN neural network method. In this article, lift-to-drag coefficients of airfoils near the ground surface are predicted with the help of a neural network. This prediction can only be realized by providing data for training and learning the code that contains information on the lift-to-drag ratio of the primary data and images related to the airfoil cross-section, which are converted into a matrix. One advantage of the VGG method over other methods is that its results are more accurate than those of other CNN methods.
Authors: Joschka Beck, Arlene John
Abstract: Advancements in deep learning have enabled highly accurate arrhythmia detection from electrocardiogram (ECG) signals, but limited interpretability remains a barrier to clinical adoption. This study investigates the application of Explainable AI (XAI) techniques specifically adapted for time-series ECG analysis. Using the MIT-BIH arrhythmia dataset, a convolutional neural network-based model was developed for arrhythmia classification, with R-peak-based segmentation via the Pan-Tompkins algorithm. To increase the dataset size and to reduce class imbalance, an additional 12-lead ECG dataset was incorporated. A user needs assessment was carried out to identify what kind of explanation would be preferred by medical professionals. Medical professionals indicated a preference for saliency map-based explanations over counterfactual visualisations, citing clearer correspondence with ECG interpretation workflows. Four SHapley Additive exPlanations (SHAP)-based approaches: permutation importance, KernelSHAP, gradient-based methods, and Deep Learning Important FeaTures (DeepLIFT), were implemented and compared. The model achieved 98.3% validation accuracy on MIT-BIH but showed performance degradation on the combined dataset, underscoring dataset variability challenges. Permutation importance and KernelSHAP produced cluttered visual outputs, while gradient-based and DeepLIFT methods highlighted waveform regions consistent with clinical reasoning, but with variability across samples. Findings emphasize the need for domain-specific XAI adaptations in ECG analysis and highlight saliency mapping as a more clinically intuitive approach
Authors: Dhiraj S Kori, Abhinav Chandraker, Syed Abdur Rahman, Punit Rathore, Ankur Chauhan
Abstract: This study proposes a Physics-Informed Neural Network (PINN) framework to predict the low-cycle fatigue (LCF) life of irradiated austenitic and ferritic/martensitic (F/M) steels used in nuclear reactors. These materials experience cyclic loading and irradiation at elevated temperatures, causing complex degradation that traditional empirical models fail to capture accurately. The developed PINN model incorporates physical fatigue life constraints into its loss function, improving prediction accuracy and generalizability. Trained on 495 data points, including both irradiated and unirradiated conditions, the model outperforms traditional machine learning models like Random Forest, Gradient Boosting, eXtreme Gradient Boosting, and the conventional Neural Network. SHapley Additive exPlanations analysis identifies strain amplitude, irradiation dose, and testing temperature as dominant features, each inversely correlated with fatigue life, consistent with physical understanding. PINN captures saturation behaviour in fatigue life at higher strain amplitudes in F/M steels. Overall, the PINN framework offers a reliable and interpretable approach for predicting fatigue life in irradiated alloys, enabling informed alloy selection.
Authors: Yifei Yao, Mengnan Du
Abstract: Understanding the internal representations of large language models (LLMs) remains a central challenge for interpretability research. Sparse autoencoders (SAEs) offer a promising solution by decomposing activations into interpretable features, but existing approaches rely on fixed sparsity constraints that fail to account for input complexity. We propose Adaptive Top K Sparse Autoencoders (AdaptiveK), a novel framework that dynamically adjusts sparsity levels based on the semantic complexity of each input. Leveraging linear probes, we demonstrate that context complexity is linearly encoded in LLM representations, and we use this signal to guide feature allocation during training. Experiments across three language models (Pythia-70M, Pythia-160M, and Gemma-2-2B) demonstrate that this complexity-driven adaptation significantly outperforms fixed-sparsity approaches on reconstruction fidelity, explained variance, and cosine similarity metrics while eliminating the computational burden of extensive hyperparameter tuning.
Authors: Qijia Zhai
Abstract: We investigate the learning dynamics of shallow ReLU neural networks on the unit sphere \(S^2\subset\mathbb{R}^3\) in polar coordinates \((\tau,\phi)\), considering both fixed and trainable neuron directions \(\{w_i\}\). For fixed weights, spherical harmonic expansions reveal an intrinsic low-frequency preference with coefficients decaying as \(O(\ell^{5/2}/2^\ell)\), typically leading to the Frequency Principle (FP) of lower-frequency-first learning. However, this principle can be violated under specific initial conditions or error distributions. With trainable weights, an additional rotation term in the harmonic evolution equations preserves exponential decay with decay order \(O(\ell^{7/2}/2^\ell)\) factor, also leading to the FP of lower-frequency-first learning. But like fixed weights case, the principle can be violated under specific initial conditions or error distributions. Our numerical results demonstrate that trainable directions increase learning complexity and can either maintain a low-frequency advantage or enable faster high-frequency emergence. This analysis suggests the FP should be viewed as a tendency rather than a rule on curved domains like \(S^2\), providing insights into how direction updates and harmonic expansions shape frequency-dependent learning.
Authors: Muhammet Anil Yagiz, Zeynep Sude Cengiz, Polat Goktas
Abstract: The rapid expansion of immersive Metaverse applications introduces complex challenges at the intersection of performance, privacy, and environmental sustainability. Centralized architectures fall short in addressing these demands, often resulting in elevated energy consumption, latency, and privacy concerns. This paper proposes MetaFed, a decentralized federated learning (FL) framework that enables sustainable and intelligent resource orchestration for Metaverse environments. MetaFed integrates (i) multi-agent reinforcement learning for dynamic client selection, (ii) privacy-preserving FL using homomorphic encryption, and (iii) carbon-aware scheduling aligned with renewable energy availability. Evaluations on MNIST and CIFAR-10 using lightweight ResNet architectures demonstrate that MetaFed achieves up to 25\% reduction in carbon emissions compared to conventional approaches, while maintaining high accuracy and minimal communication overhead. These results highlight MetaFed as a scalable solution for building environmentally responsible and privacy-compliant Metaverse infrastructures.
Authors: Yuxuan Song, Zhe Zhang, Yu Pei, Jingjing Gong, Qiying Yu, Zheng Zhang, Mingxuan Wang, Hao Zhou, Jingjing Liu, Wei-Ying Ma
Abstract: Generative modeling of discrete variables is challenging yet crucial for applications in natural language processing and biological sequence design. We introduce the Shortlisting Model (SLM), a novel simplex-based diffusion model inspired by progressive candidate pruning. SLM operates on simplex centroids, reducing generation complexity and enhancing scalability. Additionally, SLM incorporates a flexible implementation of classifier-free guidance, enhancing unconditional generation performance. Extensive experiments on DNA promoter and enhancer design, protein design, character-level and large-vocabulary language modeling demonstrate the competitive performance and strong potential of SLM. Our code can be found at https://github.com/GenSI-THUAIR/SLM
Authors: Shir Bernstein, David Beste, Daniel Ayzenshteyn, Lea Schonherr, Yisroel Mirsky
Abstract: Large Language Models (LLMs) are increasingly trusted to perform automated code review and static analysis at scale, supporting tasks such as vulnerability detection, summarization, and refactoring. In this paper, we identify and exploit a critical vulnerability in LLM-based code analysis: an abstraction bias that causes models to overgeneralize familiar programming patterns and overlook small, meaningful bugs. Adversaries can exploit this blind spot to hijack the control flow of the LLM's interpretation with minimal edits and without affecting actual runtime behavior. We refer to this attack as a Familiar Pattern Attack (FPA). We develop a fully automated, black-box algorithm that discovers and injects FPAs into target code. Our evaluation shows that FPAs are not only effective, but also transferable across models (GPT-4o, Claude 3.5, Gemini 2.0) and universal across programming languages (Python, C, Rust, Go). Moreover, FPAs remain effective even when models are explicitly warned about the attack via robust system prompts. Finally, we explore positive, defensive uses of FPAs and discuss their broader implications for the reliability and safety of code-oriented LLMs.
Authors: Jiali Cui, Yan-Ying Chen, Yanxia Zhang, Matthew Klenk
Abstract: This paper presents a novel generative framework for learning shared latent representations across multimodal data. Many advanced multimodal methods focus on capturing all combinations of modality-specific details across inputs, which can inadvertently obscure the high-level semantic concepts that are shared across modalities. Notably, Multimodal VAEs with low-dimensional latent variables are designed to capture shared representations, enabling various tasks such as joint multimodal synthesis and cross-modal inference. However, multimodal VAEs often struggle to design expressive joint variational posteriors and suffer from low-quality synthesis. In this work, ShaLa addresses these challenges by integrating a novel architectural inference model and a second-stage expressive diffusion prior, which not only facilitates effective inference of shared latent representation but also significantly improves the quality of downstream multimodal synthesis. We validate ShaLa extensively across multiple benchmarks, demonstrating superior coherence and synthesis quality compared to state-of-the-art multimodal VAEs. Furthermore, ShaLa scales to many more modalities while prior multimodal VAEs have fallen short in capturing the increasing complexity of the shared latent space.
Authors: Omar Bekdache, Naresh Shanbhag
Abstract: Federated learning (FL) accelerates the deployment of deep learning models on edge devices while preserving data privacy. However, FL systems face challenges due to client-side constraints on computational resources, and from a lack of robustness to common corruptions such as noise, blur, and weather effects. Existing robust training methods are computationally expensive and unsuitable for resource-constrained clients. We propose FedERL, federated efficient and robust learning, as the first work to explicitly address corruption robustness under time and energy constraints on the client side. At its core, FedERL employs a novel data-agnostic robust training (DART) method on the server to enhance robustness without access to the training data. In doing so, FedERL ensures zero robustness overhead for clients. Extensive experiments demonstrate FedERL's ability to handle common corruptions at a fraction of the time and energy cost of traditional robust training methods. In scenarios with limited time and energy budgets, FedERL surpasses the performance of traditional robust training, establishing it as a practical and scalable solution for real-world FL applications.
Authors: Yicong Wu, Guangyue Lu, Yuan Zuo, Huarong Zhang, Junjie Wu
Abstract: Generalizing to unseen graph tasks without task-pecific supervision remains challenging. Graph Neural Networks (GNNs) are limited by fixed label spaces, while Large Language Models (LLMs) lack structural inductive biases. Recent advances in Large Reasoning Models (LRMs) provide a zero-shot alternative via explicit, long chain-of-thought reasoning. Inspired by this, we propose a GNN-free approach that reformulates graph tasks--node classification, link prediction, and graph classification--as textual reasoning problems solved by LRMs. We introduce the first datasets with detailed reasoning traces for these tasks and develop Graph-R1, a reinforcement learning framework that leverages task-specific rethink templates to guide reasoning over linearized graphs. Experiments demonstrate that Graph-R1 outperforms state-of-the-art baselines in zero-shot settings, producing interpretable and effective predictions. Our work highlights the promise of explicit reasoning for graph learning and provides new resources for future research.
Authors: Xiaoyang Lin, Runhao Jiang, Renchi Yang
Abstract: Multi-relational graphs (MRGs) are an expressive data structure for modeling diverse interactions/relations among real objects (i.e., nodes), which pervade extensive applications and scenarios. Given an MRG G with N nodes, partitioning the node set therein into K disjoint clusters (MRGC) is a fundamental task in analyzing MRGs, which has garnered considerable attention. However, the majority of existing solutions towards MRGC either yield severely compromised result quality by ineffective fusion of heterogeneous graph structures and attributes, or struggle to cope with sizable MRGs with millions of nodes and billions of edges due to the adoption of sophisticated and costly deep learning models. In this paper, we present DEMM and DEMM+, two effective MRGC approaches to address the limitations above. Specifically, our algorithms are built on novel two-stage optimization objectives, where the former seeks to derive high-caliber node feature vectors by optimizing the multi-relational Dirichlet energy specialized for MRGs, while the latter minimizes the Dirichlet energy of clustering results over the node affinity graph. In particular, DEMM+ achieves significantly higher scalability and efficiency over our based method DEMM through a suite of well-thought-out optimizations. Key technical contributions include (i) a highly efficient approximation solver for constructing node feature vectors, and (ii) a theoretically-grounded problem transformation with carefully-crafted techniques that enable linear-time clustering without explicitly materializing the NxN dense affinity matrix. Further, we extend DEMM+ to handle attribute-less MRGs through non-trivial adaptations. Extensive experiments, comparing DEMM+ against 20 baselines over 11 real MRGs, exhibit that DEMM+ is consistently superior in terms of clustering quality measured against ground-truth labels, while often being remarkably faster.
Authors: Jacob Portes, Connor Jennings, Erica Ji Yuen, Sasha Doubov, Michael Carbin
Abstract: How does retrieval performance scale with pretraining FLOPs? We benchmark retrieval performance across LLM model sizes from 125 million parameters to 7 billion parameters pretrained on datasets ranging from 1 billion tokens to more than 2 trillion tokens. We find that retrieval performance on zero-shot BEIR tasks predictably scales with LLM size, training duration, and estimated FLOPs. We also show that In-Context Learning scores are strongly correlated with retrieval scores across retrieval tasks. Finally, we highlight the implications this has for the development of LLM-based retrievers.
Authors: Yinsong Wang, Xiao Liu, Quan Zeng, Yu Ding
Abstract: Recent breakthroughs in autonomous experimentation have demonstrated remarkable physical capabilities, yet their cognitive control remains limited--often relying on static heuristics or classical optimization. A core limitation is the absence of a principled mechanism to detect and adapt to the unexpectedness. While traditional surprise measures--such as Shannon or Bayesian Surprise--offer momentary detection of deviation, they fail to capture whether a system is truly learning and adapting. In this work, we introduce Mutual Information Surprise (MIS), a new framework that redefines surprise not as anomaly detection, but as a signal of epistemic growth. MIS quantifies the impact of new observations on mutual information, enabling autonomous systems to reflect on their learning progression. We develop a statistical test sequence to detect meaningful shifts in estimated mutual information and propose a mutual information surprise reaction policy (MISRP) that dynamically governs system behavior through sampling adjustment and process forking. Empirical evaluations--on both synthetic domains and a dynamic pollution map estimation task--show that MISRP-governed strategies significantly outperform classical surprise-based approaches in stability, responsiveness, and predictive accuracy. By shifting surprise from reactive to reflective, MIS offers a path toward more self-aware and adaptive autonomous systems.
Authors: Avishag Shapira, Simon Shigol, Asaf Shabtai
Abstract: The widespread adoption of machine learning (ML) systems increased attention to their security and emergence of adversarial machine learning (AML) techniques that exploit fundamental vulnerabilities in ML systems, creating an urgent need for comprehensive risk assessment for ML-based systems. While traditional risk assessment frameworks evaluate conventional cybersecurity risks, they lack ability to address unique challenges posed by AML threats. Existing AML threat evaluation approaches focus primarily on technical attack robustness, overlooking crucial real-world factors like deployment environments, system dependencies, and attack feasibility. Attempts at comprehensive AML risk assessment have been limited to domain-specific solutions, preventing application across diverse systems. Addressing these limitations, we present FRAME, the first comprehensive and automated framework for assessing AML risks across diverse ML-based systems. FRAME includes a novel risk assessment method that quantifies AML risks by systematically evaluating three key dimensions: target system's deployment environment, characteristics of diverse AML techniques, and empirical insights from prior research. FRAME incorporates a feasibility scoring mechanism and LLM-based customization for system-specific assessments. Additionally, we developed a comprehensive structured dataset of AML attacks enabling context-aware risk assessment. From an engineering application perspective, FRAME delivers actionable results designed for direct use by system owners with only technical knowledge of their systems, without expertise in AML. We validated it across six diverse real-world applications. Our evaluation demonstrated exceptional accuracy and strong alignment with analysis by AML experts. FRAME enables organizations to prioritize AML risks, supporting secure AI deployment in real-world environments.
Authors: Dongseok Kim, Wonjun Jeong, Gisung Oh
Abstract: We propose Anti-regularization (AR), which adds a sign-reversed reward term to the loss to intentionally increase model expressivity in the small-sample regime, and then attenuates this intervention with a power-law decay as the sample size grows. We formalize spectral safety and trust-region conditions, and design a lightweight stability safeguard that combines a projection operator with gradient clipping, ensuring stable intervention under stated assumptions. Our analysis spans linear smoothers and the Neural Tangent Kernel (NTK) regime, providing practical guidance on selecting the decay exponent by balancing empirical risk against variance. Empirically, AR reduces underfitting while preserving generalization and improving calibration in both regression and classification. Ablation studies confirm that the decay schedule and the stability safeguard are critical to preventing overfitting and numerical instability. We further examine a degrees-of-freedom targeting schedule that keeps per-sample complexity approximately constant. AR is simple to implement and reproducible, integrating cleanly into standard empirical risk minimization pipelines. It enables robust learning in data- and resource-constrained settings by intervening only when beneficial and fading away when unnecessary.
Authors: Haochen You, Baojing Liu, Hongyang He
Abstract: One-step generative modeling seeks to generate high-quality data samples in a single function evaluation, significantly improving efficiency over traditional diffusion or flow-based models. In this work, we introduce Modular MeanFlow (MMF), a flexible and theoretically grounded approach for learning time-averaged velocity fields. Our method derives a family of loss functions based on a differential identity linking instantaneous and average velocities, and incorporates a gradient modulation mechanism that enables stable training without sacrificing expressiveness. We further propose a curriculum-style warmup schedule to smoothly transition from coarse supervision to fully differentiable training. The MMF formulation unifies and generalizes existing consistency-based and flow-matching methods, while avoiding expensive higher-order derivatives. Empirical results across image synthesis and trajectory modeling tasks demonstrate that MMF achieves competitive sample quality, robust convergence, and strong generalization, particularly under low-data or out-of-distribution settings.
Authors: Yizhi Li, Qingshui Gu, Zhoufutu Wen, Ziniu Li, Tianshun Xing, Shuyue Guo, Tianyu Zheng, Xin Zhou, Xingwei Qu, Wangchunshu Zhou, Zheng Zhang, Wei Shen, Qian Liu, Chenghua Lin, Jian Yang, Ge Zhang, Wenhao Huang
Abstract: Recent advancements in aligning large language models via reinforcement learning have achieved remarkable gains in solving complex reasoning problems, but at the cost of expensive on-policy rollouts and limited exploration of diverse reasoning paths. In this work, we introduce TreePO, involving a self-guided rollout algorithm that views sequence generation as a tree-structured searching process. Composed of dynamic tree sampling policy and fixed-length segment decoding, TreePO leverages local uncertainty to warrant additional branches. By amortizing computation across common prefixes and pruning low-value paths early, TreePO essentially reduces the per-update compute burden while preserving or enhancing exploration diversity. Key contributions include: (1) a segment-wise sampling algorithm that alleviates the KV cache burden through contiguous segments and spawns new branches along with an early-stop mechanism; (2) a tree-based segment-level advantage estimation that considers both global and local proximal policy optimization. and (3) analysis on the effectiveness of probability and quality-driven dynamic divergence and fallback strategy. We empirically validate the performance gain of TreePO on a set reasoning benchmarks and the efficiency saving of GPU hours from 22\% up to 43\% of the sampling design for the trained models, meanwhile showing up to 40\% reduction at trajectory-level and 35\% at token-level sampling compute for the existing models. While offering a free lunch of inference efficiency, TreePO reveals a practical path toward scaling RL-based post-training with fewer samples and less compute. Home page locates at https://m-a-p.ai/TreePO.
URLs: https://m-a-p.ai/TreePO.
Authors: Shaocong Ma, Ziyi Chen, Yi Zhou, Heng Huang
Abstract: The goal of robust constrained reinforcement learning (RL) is to optimize an agent's performance under the worst-case model uncertainty while satisfying safety or resource constraints. In this paper, we demonstrate that strong duality does not generally hold in robust constrained RL, indicating that traditional primal-dual methods may fail to find optimal feasible policies. To overcome this limitation, we propose a novel primal-only algorithm called Rectified Robust Policy Optimization (RRPO), which operates directly on the primal problem without relying on dual formulations. We provide theoretical convergence guarantees under mild regularity assumptions, showing convergence to an approximately optimal feasible policy with iteration complexity matching the best-known lower bound when the uncertainty set diameter is controlled in a specific level. Empirical results in a grid-world environment validate the effectiveness of our approach, demonstrating that RRPO achieves robust and safe performance under model uncertainties while the non-robust method can violate the worst-case safety constraints.
Authors: Dou Jiabao, Nie Jiayi, Yihang Cheng, Jinwei Liu, Yingrui Ji, Canran Xiao, Feixiang Du, Jiaping Xiao
Abstract: The Branch-and-bound (B&B) algorithm is the main solver for Mixed Integer Linear Programs (MILPs), where the selection of branching variable is essential to computational efficiency. However, traditional heuristics for branching often fail to generalize across heterogeneous problem instances, while existing learning-based methods such as imitation learning (IL) suffers from dependence on expert demonstration quality, and reinforcement learning (RL) struggles with limitations in sparse rewards and dynamic state representation challenges. To address these issues, we propose ReviBranch, a novel deep RL framework that constructs revived trajectories by reviving explicit historical correspondences between branching decisions and their corresponding graph states along search-tree paths. During training, ReviBranch enables agents to learn from complete structural evolution and temporal dependencies within the branching process. Additionally, we introduce an importance-weighted reward redistribution mechanism that transforms sparse terminal rewards into dense stepwise feedback, addressing the sparse reward challenge. Extensive experiments on different MILP benchmarks demonstrate that ReviBranch outperforms state-of-the-art RL methods, reducing B&B nodes by 4.0% and LP iterations by 2.2% on large-scale instances. The results highlight the robustness and generalizability of ReviBranch across heterogeneous MILP problem classes.
Authors: H. Freire-Oliveira, E. R. F. Paiva, J. Gama, L. Khan, R. Cerri
Abstract: Classification in the context of multi-label data streams represents a challenge that has attracted significant attention due to its high real-world applicability. However, this task faces problems inherent to dynamic environments, such as the continuous arrival of data at high speed and volume, changes in the data distribution (concept drift), the emergence of new labels (concept evolution), and the latency in the arrival of ground truth labels. This systematic literature review presents an in-depth analysis of multi-label data stream classification proposals. We characterize the latest methods in the literature, providing a comprehensive overview, building a thorough hierarchy, and discussing how the proposals approach each problem. Furthermore, we discuss the adopted evaluation strategies and analyze the methods' asymptotic complexity and resource consumption. Finally, we identify the main gaps and offer recommendations for future research directions in the field.
Authors: Liv Gorton, Owen Lewis
Abstract: Adversarial examples -- inputs with imperceptible perturbations that fool neural networks -- remain one of deep learning's most perplexing phenomena despite nearly a decade of research. While numerous defenses and explanations have been proposed, there is no consensus on the fundamental mechanism. One underexplored hypothesis is that superposition, a concept from mechanistic interpretability, may be a major contributing factor, or even the primary cause. We present four lines of evidence in support of this hypothesis, greatly extending prior arguments by Elhage et al. (2022): (1) superposition can theoretically explain a range of adversarial phenomena, (2) in toy models, intervening on superposition controls robustness, (3) in toy models, intervening on robustness (via adversarial training) controls superposition, and (4) in ResNet18, intervening on robustness (via adversarial training) controls superposition.
Authors: Krishna Teja Chitty-Venkata, Sylvia Howland, Golara Azar, Daria Soboleva, Natalia Vassilieva, Siddhisanket Raskar, Murali Emani, Venkatram Vishwanath
Abstract: Mixture of Experts (MoE) models have enabled the scaling of Large Language Models (LLMs) and Vision Language Models (VLMs) by achieving massive parameter counts while maintaining computational efficiency. However, MoEs introduce several inference-time challenges, including load imbalance across experts and the additional routing computational overhead. To address these challenges and fully harness the benefits of MoE, a systematic evaluation of hardware acceleration techniques is essential. We present MoE-Inference-Bench, a comprehensive study to evaluate MoE performance across diverse scenarios. We analyze the impact of batch size, sequence length, and critical MoE hyperparameters such as FFN dimensions and number of experts on throughput. We evaluate several optimization techniques on Nvidia H100 GPUs, including pruning, Fused MoE operations, speculative decoding, quantization, and various parallelization strategies. Our evaluation includes MoEs from the Mixtral, DeepSeek, OLMoE and Qwen families. The results reveal performance differences across configurations and provide insights for the efficient deployment of MoEs.
Authors: Martin K\"appel, Julian Neuberger, Felix M\"ohrlein, Sven Weinzierl, Martin Matzner, Stefan Jablonski
Abstract: Predictive process monitoring enables organizations to proactively react and intervene in running instances of a business process. Given an incomplete process instance, predictions about the outcome, next activity, or remaining time are created. This is done by powerful machine learning models, which have shown impressive predictive performance. However, the data-driven nature of these models makes them susceptible to finding unfair, biased, or unethical patterns in the data. Such patterns lead to biased predictions based on so-called sensitive attributes, such as the gender or age of process participants. Previous work has identified this problem and offered solutions that mitigate biases by removing sensitive attributes entirely from the process instance. However, sensitive attributes can be used both fairly and unfairly in the same process instance. For example, during a medical process, treatment decisions could be based on gender, while the decision to accept a patient should not be based on gender. This paper proposes a novel, model-agnostic approach for identifying and rectifying biased decisions in predictive business process monitoring models, even when the same sensitive attribute is used both fairly and unfairly. The proposed approach uses a human-in-the-loop approach to differentiate between fair and unfair decisions through simple alterations on a decision tree model distilled from the original prediction model. Our results show that the proposed approach achieves a promising tradeoff between fairness and accuracy in the presence of biased data. All source code and data are publicly available at https://doi.org/10.5281/zenodo.15387576.
Authors: Yang Qiao, Yuntong Hu, Liang Zhao
Abstract: Multimodal representation learning has advanced rapidly with contrastive models such as CLIP, which align image-text pairs in a shared embedding space. However, these models face limitations: (1) they typically focus on image-text pairs, underutilizing the semantic relations across different pairs. (2) they directly match global embeddings without contextualization, overlooking the need for semantic alignment along specific subspaces or relational dimensions; and (3) they emphasize cross-modal contrast, with limited support for intra-modal consistency. To address these issues, we propose Relation-Conditioned Multimodal Learning RCML, a framework that learns multimodal representations under natural-language relation descriptions to guide both feature extraction and alignment. Our approach constructs many-to-many training pairs linked by semantic relations and introduces a relation-guided cross-attention mechanism that modulates multimodal representations under each relation context. The training objective combines inter-modal and intra-modal contrastive losses, encouraging consistency across both modalities and semantically related samples. Experiments on different datasets show that RCML consistently outperforms strong baselines on both retrieval and classification tasks, highlighting the effectiveness of leveraging semantic relations to guide multimodal representation learning.
Authors: Chang Yue, Niraj K. Jha
Abstract: Differentiable logic networks (DLNs) have shown promising results in tabular domains by combining accuracy, interpretability, and computational efficiency. In this work, we apply DLNs to the domain of TSC for the first time, focusing on univariate datasets. To enable DLN application in this context, we adopt feature-based representations relying on Catch22 and TSFresh, converting sequential time series into vectorized forms suitable for DLN classification. Unlike prior DLN studies that fix the training configuration and vary various settings in isolation via ablation, we integrate all such configurations into the hyperparameter search space, enabling the search process to select jointly optimal settings. We then analyze the distribution of selected configurations to better understand DLN training dynamics. We evaluate our approach on 51 publicly available univariate TSC benchmarks. The results confirm that classification DLNs maintain their core strengths in this new domain: they deliver competitive accuracy, retain low inference cost, and provide transparent, interpretable decision logic, thus aligning well with previous DLN findings in the realm of tabular classification and regression tasks.
Authors: Kyrylo Yemets, Mykola Lukashchuk, Ivan Izonin
Abstract: Accurate univariate forecasting remains a pressing need in real-world systems, such as energy markets, hydrology, retail demand, and IoT monitoring, where signals are often intermittent and horizons span both short- and long-term. While transformers and Mixture-of-Experts (MoE) architectures are increasingly favored for time-series forecasting, a key gap persists: MoE models typically require complicated training with both the main forecasting loss and auxiliary load-balancing losses, along with careful routing/temperature tuning, which hinders practical adoption. In this paper, we propose a model architecture that simplifies the training process for univariate time series forecasting and effectively addresses both long- and short-term horizons, including intermittent patterns. Our approach combines sparse MoE computation with a novel attention-inspired gating mechanism that replaces the traditional one-layer softmax router. Through extensive empirical evaluation, we demonstrate that our gating design naturally promotes balanced expert utilization and achieves superior predictive accuracy without requiring the auxiliary load-balancing losses typically used in classical MoE implementations. The model achieves better performance while utilizing only a fraction of the parameters required by state-of-the-art transformer models, such as PatchTST. Furthermore, experiments across diverse datasets confirm that our MoE architecture with the proposed gating mechanism is more computationally efficient than LSTM for both long- and short-term forecasting, enabling cost-effective inference. These results highlight the potential of our approach for practical time-series forecasting applications where both accuracy and computational efficiency are critical.
Authors: YongKyung Oh, Dong-Young Lim, Sungil Kim, Alex Bui
Abstract: Handling missing data in time series classification remains a significant challenge in various domains. Traditional methods often rely on imputation, which may introduce bias or fail to capture the underlying temporal dynamics. In this paper, we propose TANDEM (Temporal Attention-guided Neural Differential Equations for Missingness), an attention-guided neural differential equation framework that effectively classifies time series data with missing values. Our approach integrates raw observation, interpolated control path, and continuous latent dynamics through a novel attention mechanism, allowing the model to focus on the most informative aspects of the data. We evaluate TANDEM on 30 benchmark datasets and a real-world medical dataset, demonstrating its superiority over existing state-of-the-art methods. Our framework not only improves classification accuracy but also provides insights into the handling of missing data, making it a valuable tool in practice.
Authors: YongKyung Oh, Seungsu Kam, Dong-Young Lim, Sungil Kim
Abstract: Astronomical time series from large-scale surveys like LSST are often irregularly sampled and incomplete, posing challenges for classification and anomaly detection. We introduce a new framework based on Neural Stochastic Delay Differential Equations (Neural SDDEs) that combines stochastic modeling with neural networks to capture delayed temporal dynamics and handle irregular observations. Our approach integrates a delay-aware neural architecture, a numerical solver for SDDEs, and mechanisms to robustly learn from noisy, sparse sequences. Experiments on irregularly sampled astronomical data demonstrate strong classification accuracy and effective detection of novel astrophysical events, even with partial labels. This work highlights Neural SDDEs as a principled and practical tool for time series analysis under observational constraints.
Authors: Marcel Hoffmann, Lukas Galke, Ansgar Scherp
Abstract: Graph homophily has been considered an essential property for message-passing neural networks (MPNN) in node classification. Recent findings suggest that performance is more closely tied to the consistency of neighborhood class distributions. We demonstrate that the MPNN performance depends on the number of components of the overall neighborhood distribution within a class. By breaking down the classes into their neighborhood distribution components, we increase measures of neighborhood distribution informativeness but do not observe an improvement in MPNN performance. We propose a Gumbel-Softmax-based rewiring method that reduces deviations in neighborhood distributions. Our results show that our new method enhances neighborhood informativeness, handles long-range dependencies, mitigates oversquashing, and increases the classification performance of the MPNN. The code is available at https://github.com/Bobowner/Gumbel-Softmax-MPNN.
Authors: Andrzej Szablewski, Marek Masiak
Abstract: The residual stream mediates communication between transformer decoder layers via linear reads and writes of non-linear computations. While sparse-dictionary learning-based methods locate features in the residual stream, and activation patching methods discover circuits within the model, the mechanism by which features flow through the residual stream remains understudied. Understanding this dynamic can better inform jailbreaking protections, enable early detection of model mistakes, and their correction. In this work, we propose Activation Transport Operators (ATO), linear maps from upstream to downstream residuals $k$ layers later, evaluated in feature space using downstream SAE decoder projections. We empirically demonstrate that these operators can determine whether a feature has been linearly transported from a previous layer or synthesised from non-linear layer computation. We develop the notion of transport efficiency, for which we provide an upper bound, and use it to estimate the size of the residual stream subspace that corresponds to linear transport. We empirically demonstrate the linear transport, report transport efficiency and the size of the residual stream's subspace involved in linear transport. This compute-light (no finetuning, <50 GPU-h) method offers practical tools for safety, debugging, and a clearer picture of where computation in LLMs behaves linearly.
Authors: Jerry Yao-Chieh Hu, Hude Liu, Jennifer Yuntong Zhang, Han Liu
Abstract: We prove that a minimal Transformer architecture with frozen weights is capable of emulating a broad class of algorithms by in-context prompting. In particular, for any algorithm implementable by a fixed-weight attention head (e.g. one-step gradient descent or linear/ridge regression), there exists a prompt that drives a two-layer softmax attention module to reproduce the algorithm's output with arbitrary precision. This guarantee extends even to a single-head attention layer (using longer prompts if necessary), achieving architectural minimality. Our key idea is to construct prompts that encode an algorithm's parameters into token representations, creating sharp dot-product gaps that force the softmax attention to follow the intended computation. This construction requires no feed-forward layers and no parameter updates. All adaptation happens through the prompt alone. These findings forge a direct link between in-context learning and algorithmic emulation, and offer a simple mechanism for large Transformers to serve as prompt-programmable libraries of algorithms. They illuminate how GPT-style foundation models may swap algorithms via prompts alone, establishing a form of algorithmic universality in modern Transformer models.
Authors: Shuqi Zi, Haitz S\'aez de Oc\'ariz Borde, Emma Rocheteau, Pietro Lio'
Abstract: Predicting a patient's length of stay (LOS) in the intensive care unit (ICU) is a critical task for hospital resource management, yet remains challenging due to the heterogeneous and irregularly sampled nature of electronic health records (EHRs). In this work, we propose S$^2$G-Net, a novel neural architecture that unifies state-space sequence modeling with multi-view Graph Neural Networks (GNNs) for ICU LOS prediction. The temporal path employs Mamba state-space models (SSMs) to capture patient trajectories, while the graph path leverages an optimized GraphGPS backbone, designed to integrate heterogeneous patient similarity graphs derived from diagnostic, administrative, and semantic features. Experiments on the large-scale MIMIC-IV cohort dataset show that S$^2$G-Net consistently outperforms sequence models (BiLSTM, Mamba, Transformer), graph models (classic GNNs, GraphGPS), and hybrid approaches across all primary metrics. Extensive ablation studies and interpretability analyses highlight the complementary contributions of each component of our architecture and underscore the importance of principled graph construction. These results demonstrate that S$^2$G-Net provides an effective and scalable solution for ICU LOS prediction with multi-modal clinical data.
Authors: Daniel Frees, Aditri Bhagirath, Moritz Bolling
Abstract: While Large Language Models (LLMs) have revolutionized artificial intelligence, fine-tuning LLMs is extraordinarily computationally expensive, preventing smaller businesses and research teams with limited GPU resources from engaging with new research. Hu et al and Liu et al introduce Low-Rank Adaptation (LoRA) and Weight-Decomposed Low-Rank Adaptation (DoRA) as highly efficient and performant solutions to the computational challenges of LLM fine-tuning, demonstrating huge speedups and memory usage savings for models such as GPT-3 and RoBERTa. We seek to expand upon the original LoRA and DoRA papers by benchmarking efficiency and performance of LoRA and DoRA when applied to a much smaller scale of language model: our case study here is the compact minBERT model. Our findings reveal that optimal custom configurations of LoRA and DoRA, coupled with Automatic Mixed Precision (AMP), significantly enhance training efficiency without compromising performance. Furthermore, while the parameterization of minBERT is significantly smaller than GPT-3, our results validate the observation that gradient updates to language models are inherently low-rank even in small model space, observing that rank 1 decompositions yield negligible performance deficits. Furthermore, aided by our highly efficient minBERT implementation, we investigate numerous architectures, custom loss functions, and hyperparameters to ultimately train an optimal ensembled multitask minBERT model to simultaneously perform sentiment analysis, paraphrase detection, and similarity scoring.
Authors: Wentao Tan, Qiong Cao, Chao Xue, Yibing Zhan, Changxing Ding, Xiaodong He
Abstract: The chart-to-code generation task requires MLLMs to convert chart images into executable code. This task faces two major challenges: limited data diversity and insufficient maintenance of visual consistency between generated and original charts during training. Existing datasets mainly rely on seed data to prompt GPT models for code generation, resulting in homogeneous samples. To address this, we propose ReChartPrompt, which leverages real-world, human-designed charts from arXiv papers as prompts instead of synthetic seeds. Using the diverse styles and rich content of arXiv charts, we construct ReChartPrompt-240K, a large-scale and highly diverse dataset. Another challenge is that although SFT effectively improve code understanding, it often fails to ensure that generated charts are visually consistent with the originals. To address this, we propose ChartSimRL, a GRPO-based reinforcement learning algorithm guided by a novel chart similarity reward. This reward consists of attribute similarity, which measures the overlap of chart attributes such as layout and color between the generated and original charts, and visual similarity, which assesses similarity in texture and other overall visual features using convolutional neural networks. Unlike traditional text-based rewards such as accuracy or format rewards, our reward considers the multimodal nature of the chart-to-code task and effectively enhances the model's ability to accurately reproduce charts. By integrating ReChartPrompt and ChartSimRL, we develop the ChartMaster model, which achieves state-of-the-art results among 7B-parameter models and even rivals GPT-4o on various chart-to-code generation benchmarks. All resources are available at https://github.com/WentaoTan/ChartMaster.
Authors: Jinli Li, Shiyu Long, Minglian Han
Abstract: In industrial big data scenarios, high-dimensional sparse matrices (HDI) are widely used to characterize high-order interaction relationships among massive nodes. The stochastic gradient descent-based latent factor analysis (SGD-LFA) method can effectively extract deep feature information embedded in HDI matrices. However, existing SGD-LFA methods exhibit significant limitations: their parameter update process relies solely on the instantaneous gradient information of current samples, failing to incorporate accumulated experiential knowledge from historical iterations or account for intrinsic correlations between samples, resulting in slow convergence speed and suboptimal generalization performance. Thus, this paper proposes a PILF model by developing a PI-accelerated SGD algorithm by integrating correlated instances and refining learning errors through proportional-integral (PI) control mechanism that current and historical information; Comparative experiments demonstrate the superior representation capability of the PILF model on HDI matrices
Authors: An Ning, Tai Yue Li, Nan Yow Chen
Abstract: We propose the Quantum Graph Attention Network (QGAT), a hybrid graph neural network that integrates variational quantum circuits into the attention mechanism. At its core, QGAT employs strongly entangling quantum circuits with amplitude-encoded node features to enable expressive nonlinear interactions. Distinct from classical multi-head attention that separately computes each head, QGAT leverages a single quantum circuit to simultaneously generate multiple attention coefficients. This quantum parallelism facilitates parameter sharing across heads, substantially reducing computational overhead and model complexity. Classical projection weights and quantum circuit parameters are optimized jointly in an end-to-end manner, ensuring flexible adaptation to learning tasks. Empirical results demonstrate QGAT's effectiveness in capturing complex structural dependencies and improved generalization in inductive scenarios, highlighting its potential for scalable quantum-enhanced learning across domains such as chemistry, biology, and network analysis. Furthermore, experiments confirm that quantum embedding enhances robustness against feature and structural noise, suggesting advantages in handling real-world noisy data. The modularity of QGAT also ensures straightforward integration into existing architectures, allowing it to easily augment classical attention-based models.
Authors: Nima Kondori, Hanwen Liang, Hooman Vaseli, Bingyu Xie, Christina Luong, Purang Abolmaesumi, Teresa Tsang, Renjie Liao
Abstract: Synthetic data generation represents a significant advancement in boosting the performance of machine learning (ML) models, particularly in fields where data acquisition is challenging, such as echocardiography. The acquisition and labeling of echocardiograms (echo) for heart assessment, crucial in point-of-care ultrasound (POCUS) settings, often encounter limitations due to the restricted number of echo views available, typically captured by operators with varying levels of experience. This study proposes a novel approach for enhancing clinical diagnosis accuracy by synthetically generating echo views. These views are conditioned on existing, real views of the heart, focusing specifically on the estimation of ejection fraction (EF), a critical parameter traditionally measured from biplane apical views. By integrating a conditional generative model, we demonstrate an improvement in EF estimation accuracy, providing a comparative analysis with traditional methods. Preliminary results indicate that our synthetic echoes, when used to augment existing datasets, not only enhance EF estimation but also show potential in advancing the development of more robust, accurate, and clinically relevant ML models. This approach is anticipated to catalyze further research in synthetic data applications, paving the way for innovative solutions in medical imaging diagnostics.
Authors: Yilang Ding, Jiawen Ren, Jiaying Lu, Gloria Hyunjung Kwak, Armin Iraji, Alex Fedorov
Abstract: Alzheimer's disease is a progressive neurodegenerative disorder that remains challenging to predict due to its multifactorial etiology and the complexity of multimodal clinical data. Accurate forecasting of clinically relevant biomarkers, including diagnostic and quantitative measures, is essential for effective monitoring of disease progression. This work introduces L2C-TabPFN, a method that integrates a longitudinal-to-cross-sectional (L2C) transformation with a pre-trained Tabular Foundation Model (TabPFN) to predict Alzheimer's disease outcomes using the TADPOLE dataset. L2C-TabPFN converts sequential patient records into fixed-length feature vectors, enabling robust prediction of diagnosis, cognitive scores, and ventricular volume. Experimental results demonstrate that, while L2C-TabPFN achieves competitive performance on diagnostic and cognitive outcomes, it provides state-of-the-art results in ventricular volume prediction. This key imaging biomarker reflects neurodegeneration and progression in Alzheimer's disease. These findings highlight the potential of tabular foundational models for advancing longitudinal prediction of clinically relevant imaging markers in Alzheimer's disease.
Authors: Takuro Ishida, Tetsuo Furukawa
Abstract: This paper proposes an embedding method for co-occurrence data aimed at visual information exploration. We consider cases where co-occurrence probabilities are measured between pairs of elements from heterogeneous domains. The proposed method maps these heterogeneous elements into corresponding two-dimensional latent spaces, enabling visualization of asymmetric relationships between the domains. The key idea is to embed the elements in a way that maximizes their mutual information, thereby preserving the original dependency structure as much as possible. This approach can be naturally extended to cases involving three or more domains, using a generalization of mutual information known as total correlation. For inter-domain analysis, we also propose a visualization method that assigns colors to the latent spaces based on conditional probabilities, allowing users to explore asymmetric relationships interactively. We demonstrate the utility of the method through applications to an adjective-noun dataset, the NeurIPS dataset, and a subject-verb-object dataset, showcasing both intra- and inter-domain analysis.
Authors: Victoria Yan, Honor Chotkowski, Fengran Wang, Alex Fedorov
Abstract: Cognitive assessments require normative data as essential benchmarks for evaluating individual performance. Hence, developing new cognitive tests based on novel image stimuli is challenging due to the lack of readily available normative data. Traditional data collection methods are costly, time-consuming, and infrequently updated, limiting their practical utility. Recent advancements in generative multimodal large language models (MLLMs) offer a new approach to generate synthetic normative data from existing cognitive test images. We investigated the feasibility of using MLLMs, specifically GPT-4o and GPT-4o-mini, to synthesize normative textual responses for established image-based cognitive assessments, such as the "Cookie Theft" picture description task. Two distinct prompting strategies-naive prompts with basic instructions and advanced prompts enriched with contextual guidance-were evaluated. Responses were analyzed using embeddings to assess their capacity to distinguish diagnostic groups and demographic variations. Performance metrics included BLEU, ROUGE, BERTScore, and an LLM-as-a-judge evaluation. Advanced prompting strategies produced synthetic responses that more effectively distinguished between diagnostic groups and captured demographic diversity compared to naive prompts. Superior models generated responses exhibiting higher realism and diversity. BERTScore emerged as the most reliable metric for contextual similarity assessment, while BLEU was less effective for evaluating creative outputs. The LLM-as-a-judge approach provided promising preliminary validation results. Our study demonstrates that generative multimodal LLMs, guided by refined prompting methods, can feasibly generate robust synthetic normative data for existing cognitive tests, thereby laying the groundwork for developing novel image-based cognitive assessments without the traditional limitations.
Authors: Yifan Wang, Binbin Liu, Fengze Liu, Yuanfan Guo, Jiyao Deng, Xuecheng Wu, Weidong Zhou, Xiaohuan Zhou, Taifeng Wang
Abstract: The data mixture used in the pre-training of a language model is a cornerstone of its final performance. However, a static mixing strategy is suboptimal, as the model's learning preferences for various data domains shift dynamically throughout training. Crucially, observing these evolving preferences in a computationally efficient manner remains a significant challenge. To address this, we propose TiKMiX, a method that dynamically adjusts the data mixture according to the model's evolving preferences. TiKMiX introduces Group Influence, an efficient metric for evaluating the impact of data domains on the model. This metric enables the formulation of the data mixing problem as a search for an optimal, influence-maximizing distribution. We solve this via two approaches: TiKMiX-D for direct optimization, and TiKMiX-M, which uses a regression model to predict a superior mixture. We trained models with different numbers of parameters, on up to 1 trillion tokens. TiKMiX-D exceeds the performance of state-of-the-art methods like REGMIX while using just 20% of the computational resources. TiKMiX-M leads to an average performance gain of 2% across 9 downstream benchmarks. Our experiments reveal that a model's data preferences evolve with training progress and scale, and we demonstrate that dynamically adjusting the data mixture based on Group Influence, a direct measure of these preferences, significantly improves performance by mitigating the underdigestion of data seen with static ratios.
Authors: Trinayan Baruah, Kaustubh Shivdikar, Sara Prescott, David Kaeli
Abstract: Mamba-based State Space Models (SSM) have emerged as a promising alternative to the ubiquitous transformers. Despite the expressive power of transformers, the quadratic complexity of computing attention is a major impediment to scaling performance as we increase the sequence length. SSMs provide an alternative path that addresses this problem, reducing the computational complexity requirements of self-attention with novel model architectures for different domains and fields such as video, text generation and graphs. Thus, it is important to characterize the behavior of these emerging workloads on GPUs and understand their requirements during GPU microarchitectural design. In this work we evaluate Mamba-based SSMs and characterize their behavior during training on GPUs. We construct a workload suite that offers representative models that span different model architectures. We then use this suite to analyze the architectural implications of running Mamba-based SSMs on GPUs. Our work sheds new light on potential optimizations to continue scaling the performance for such models.
Authors: Quanwei Wu, Jun Guo, Wei Wang, Yi Wang
Abstract: Adversarial training (AT) with projected gradient descent is the most popular method to improve model robustness under adversarial attacks. However, computational overheads become prohibitively large when AT is applied to large backbone models. AT is also known to have the issue of robust overfitting. This paper contributes to solving both problems simultaneously towards building more trustworthy foundation models. In particular, we propose a new adapter-based approach for efficient AT directly in the feature space. We show that the proposed adapter-based approach can improve the inner-loop convergence quality by eliminating robust overfitting. As a result, it significantly increases computational efficiency and improves model accuracy by generalizing adversarial robustness to unseen attacks. We demonstrate the effectiveness of the new adapter-based approach in different backbone architectures and in AT at scale.
Authors: Robert Yang
Abstract: Bold claims about AI's role in science-from "AGI will cure all diseases" to promises of radically accelerated discovery-raise a central epistemic question: do large language models (LLMs) truly generate new knowledge, or do they merely remix memorized fragments? We propose unlearning-as-ablation as a falsifiable test of constructive scientific discovery. The method systematically removes a target result and its entire forget-closure (lemmas, paraphrases, and multi-hop entailments) and then evaluates whether the model can re-derive the result from only permitted axioms and tools. Success provides evidence for genuine generative capability; failure exposes current limits. Unlike prevailing motivations for unlearning-privacy, copyright, or safety-our framing repositions it as an epistemic probe for AI-for-Science. We argue that such tests could serve as the next generation of benchmarks, much as ImageNet catalyzed progress in vision: distinguishing models that can merely recall from those that can constructively generate new scientific knowledge. We outline a minimal pilot in mathematics and algorithms, and discuss extensions to physics, chemistry, and biology. Whether models succeed or fail, unlearning-as-ablation provides a principled framework to map the true reach and limits of AI scientific discovery. This is a position paper: we advance a conceptual and methodological argument rather than new empirical results.
Authors: Sam Buchanan, Druv Pai, Yi Ma, Valentin De Bortoli
Abstract: When do diffusion models reproduce their training data, and when are they able to generate samples beyond it? A practically relevant theoretical understanding of this interplay between memorization and generalization may significantly impact real-world deployments of diffusion models with respect to issues such as copyright infringement and data privacy. In this work, to disentangle the different factors that influence memorization and generalization in practical diffusion models, we introduce a scientific and mathematical "laboratory" for investigating these phenomena in diffusion models trained on fully synthetic or natural image-like structured data. Within this setting, we hypothesize that the memorization or generalization behavior of an underparameterized trained model is determined by the difference in training loss between an associated memorizing model and a generalizing model. To probe this hypothesis, we theoretically characterize a crossover point wherein the weighted training loss of a fully generalizing model becomes greater than that of an underparameterized memorizing model at a critical value of model (under)parameterization. We then demonstrate via carefully-designed experiments that the location of this crossover predicts a phase transition in diffusion models trained via gradient descent, validating our hypothesis. Ultimately, our theory enables us to analytically predict the model size at which memorization becomes predominant. Our work provides an analytically tractable and practically meaningful setting for future theoretical and empirical investigations. Code for our experiments is available at https://github.com/DruvPai/diffusion_mem_gen.
Authors: Jiaqi Zhu, Bikramjit Das, Yong Xie, Nikolaos Pappas, Howard H. Yang
Abstract: Federated learning facilitates collaborative model training across multiple clients while preserving data privacy. However, its performance is often constrained by limited communication resources, particularly in systems supporting a large number of clients. To address this challenge, integrating over-the-air computations into the training process has emerged as a promising solution to alleviate communication bottlenecks. The system significantly increases the number of clients it can support in each communication round by transmitting intermediate parameters via analog signals rather than digital ones. This improvement, however, comes at the cost of channel-induced distortions, such as fading and noise, which affect the aggregated global parameters. To elucidate these effects, this paper develops a theoretical framework to analyze the performance of over-the-air federated learning in large-scale client scenarios. Our analysis reveals three key advantages of scaling up the number of participating clients: (1) Enhanced Privacy: The mutual information between a client's local gradient and the server's aggregated gradient diminishes, effectively reducing privacy leakage. (2) Mitigation of Channel Fading: The channel hardening effect eliminates the impact of small-scale fading in the noisy global gradient. (3) Improved Convergence: Reduced thermal noise and gradient estimation errors benefit the convergence rate. These findings solidify over-the-air model training as a viable approach for federated learning in networks with a large number of clients. The theoretical insights are further substantiated through extensive experimental evaluations.
Authors: Junying Yang, Gang Lu, Xiaoqing Yan, Peng Xia, Di Wu
Abstract: Machine learning (ML) is capable of accurate Load Forecasting from complete data. However, there are many uncertainties that affect data collection, leading to sparsity. This article proposed a model called Adaptive Ensemble Learning with Gaussian Copula to deal with sparsity, which contains three modules: data complementation, ML construction, and adaptive ensemble. First, it applies Gaussian Copula to eliminate sparsity. Then, we utilise five ML models to make predictions individually. Finally, it employs adaptive ensemble to get final weighted-sum result. Experiments have demonstrated that our model are robust.
Authors: Runwen Hu, Peilin Chen, Keyan Ding, Shiqi Wang
Abstract: Artificial intelligence (AI) revolutionizes molecule generation in bioengineering and biological research, significantly accelerating discovery processes. However, this advancement introduces critical concerns regarding intellectual property protection. To address these challenges, we propose the first robust watermarking method designed for molecules, which utilizes atom-level features to preserve molecular integrity and invariant features to ensure robustness against affine transformations. Comprehensive experiments validate the effectiveness of our method using the datasets QM9 and GEOM-DRUG, and generative models GeoBFN and GeoLDM. We demonstrate the feasibility of embedding watermarks, maintaining basic properties higher than 90.00\% while achieving watermark accuracy greater than 95.00\%. Furthermore, downstream docking simulations reveal comparable performance between original and watermarked molecules, with binding affinities reaching -6.00 kcal/mol and root mean square deviations below 1.602 \AA. These results confirm that our watermarking technique effectively safeguards molecular intellectual property without compromising scientific utility, enabling secure and responsible AI integration in molecular discovery and research applications.
Authors: Xuekang Wang, Shengyu Zhu, Xueqi Cheng
Abstract: Despite extensive efforts to align Large Language Models (LLMs) with human values and safety rules, jailbreak attacks that exploit certain vulnerabilities continuously emerge, highlighting the need to strengthen existing LLMs with additional safety properties to defend against these attacks. However, tuning large models has become increasingly resource-intensive and may have difficulty ensuring consistent performance. We introduce Speculative Safety-Aware Decoding (SSD), a lightweight decoding-time approach that equips LLMs with the desired safety property while accelerating inference. We assume that there exists a small language model that possesses this desired property. SSD integrates speculative sampling during decoding and leverages the match ratio between the small and composite models to quantify jailbreak risks. This enables SSD to dynamically switch between decoding schemes to prioritize utility or safety, to handle the challenge of different model capacities. The output token is then sampled from a new distribution that combines the distributions of the original and the small models. Experimental results show that SSD successfully equips the large model with the desired safety property, and also allows the model to remain helpful to benign queries. Furthermore, SSD accelerates the inference time, thanks to the speculative sampling design.
Authors: Sotaro Takeshita, Yurina Takeshita, Daniel Ruffinelli, Simone Paolo Ponzetto
Abstract: In this paper, we study the surprising impact that truncating text embeddings has on downstream performance. We consistently observe across 6 state-of-the-art text encoders and 26 downstream tasks, that randomly removing up to 50% of embedding dimensions results in only a minor drop in performance, less than 10%, in retrieval and classification tasks. Given the benefits of using smaller-sized embeddings, as well as the potential insights about text encoding, we study this phenomenon and find that, contrary to what is suggested in prior work, this is not the result of an ineffective use of representation space. Instead, we find that a large number of uniformly distributed dimensions actually cause an increase in performance when removed. This would explain why, on average, removing a large number of embedding dimensions results in a marginal drop in performance. We make similar observations when truncating the embeddings used by large language models to make next-token predictions on generative tasks, suggesting that this phenomenon is not isolated to classification or retrieval tasks.
Authors: Alessio Arcudi, Davide Sartor, Alberto Sinigaglia, Vincent Fran\c{c}ois-Lavet, Gian Antonio Susto
Abstract: This paper introduces MANGO (Multilayer Abstraction for Nested Generation of Options), a novel hierarchical reinforcement learning framework designed to address the challenges of long-term sparse reward environments. MANGO decomposes complex tasks into multiple layers of abstraction, where each layer defines an abstract state space and employs options to modularize trajectories into macro-actions. These options are nested across layers, allowing for efficient reuse of learned movements and improved sample efficiency. The framework introduces intra-layer policies that guide the agent's transitions within the abstract state space, and task actions that integrate task-specific components such as reward functions. Experiments conducted in procedurally-generated grid environments demonstrate substantial improvements in both sample efficiency and generalization capabilities compared to standard RL methods. MANGO also enhances interpretability by making the agent's decision-making process transparent across layers, which is particularly valuable in safety-critical and industrial applications. Future work will explore automated discovery of abstractions and abstract actions, adaptation to continuous or fuzzy environments, and more robust multi-layer training strategies.
Authors: Fanjiang Ye, Zepeng Zhao, Yi Mu, Jucheng Shen, Renjie Li, Kaijian Wang, Desen Sun, Saurabh Agarwal, Myungjin Lee, Triston Cao, Aditya Akella, Arvind Krishnamurthy, T. S. Eugene Ng, Zhengzhong Tu, Yuke Wang
Abstract: Diffusion models have recently achieved remarkable success in generative tasks (e.g., image and video generation), and the demand for high-quality content (e.g., 2K/4K videos) is rapidly increasing across various domains. However, generating ultra-high-resolution videos on existing standard-resolution (e.g., 720p) platforms remains challenging due to the excessive re-training requirements and prohibitively high computational and memory costs. To this end, we introduce SuperGen, an efficient tile-based framework for ultra-high-resolution video generation. SuperGen features a novel training-free algorithmic innovation with tiling to successfully support a wide range of resolutions without additional training efforts while significantly reducing both memory footprint and computational complexity. Moreover, SuperGen incorporates a tile-tailored, adaptive, region-aware caching strategy that accelerates video generation by exploiting redundancy across denoising steps and spatial regions. SuperGen also integrates cache-guided, communication-minimized tile parallelism for enhanced throughput and minimized latency. Evaluations demonstrate that SuperGen harvests the maximum performance gains while achieving high output quality across various benchmarks.
Authors: Jelke Wibbeke, Nico Sch\"onfisch, Sebastian Rohjans, Andreas Rauh
Abstract: In safety-critical applications data-driven models must not only be accurate but also provide reliable uncertainty estimates. This property, commonly referred to as calibration, is essential for risk-aware decision-making. In regression a wide variety of calibration metrics and recalibration methods have emerged. However, these metrics differ significantly in their definitions, assumptions and scales, making it difficult to interpret and compare results across studies. Moreover, most recalibration methods have been evaluated using only a small subset of metrics, leaving it unclear whether improvements generalize across different notions of calibration. In this work, we systematically extract and categorize regression calibration metrics from the literature and benchmark these metrics independently of specific modelling methods or recalibration approaches. Through controlled experiments with real-world, synthetic and artificially miscalibrated data, we demonstrate that calibration metrics frequently produce conflicting results. Our analysis reveals substantial inconsistencies: many metrics disagree in their evaluation of the same recalibration result, and some even indicate contradictory conclusions. This inconsistency is particularly concerning as it potentially allows cherry-picking of metrics to create misleading impressions of success. We identify the Expected Normalized Calibration Error (ENCE) and the Coverage Width-based Criterion (CWC) as the most dependable metrics in our tests. Our findings highlight the critical role of metric selection in calibration research.
Authors: Duseok Kang, Yunseong Lee, Junghoon Kim
Abstract: As deep learning models are increasingly deployed on mobile devices, modern mobile devices incorporate deep learning-specific accelerators to handle the growing computational demands, thus increasing their hardware heterogeneity. However, existing works on scheduling deep learning workloads across these processors have significant limitations: most studies focus on single-model scenarios rather than realistic multi-model scenarios, overlook performance variations from different hardware/software configurations, and struggle with accurate execution time estimation. To address these challenges, we propose a novel genetic algorithm-based methodology for scheduling multiple deep learning networks on heterogeneous processors by partitioning the networks into multiple subgraphs. Our approach incorporates three different types of chromosomes for partition/mapping/priority exploration, and leverages device-in-the-loop profiling and evaluation for accurate execution time estimation. Based on this methodology, our system, Puzzle, demonstrates superior performance in extensive evaluations with randomly generated scenarios involving nine state-of-the-art networks. The results demonstrate Puzzle can support 3.7 and 2.2 times higher request frequency on average compared to the two heuristic baselines, NPU Only and Best Mapping, respectively, while satisfying the equivalent level of real-time requirements.
Authors: Wenhong Zhu, Ruobing Xie, Rui Wang, Xingwu Sun, Di Wang, Pengfei Liu
Abstract: Supervised fine-tuning (SFT) of foundation models often leads to poor generalization, where prior capabilities deteriorate after tuning on new tasks or domains. Inspired by trust-region policy optimization (TRPO) and proximal policy optimization (PPO) in reinforcement learning (RL), we propose Proximal SFT (PSFT). This fine-tuning objective incorporates the benefits of trust-region, effectively constraining policy drift during SFT while maintaining competitive tuning. By viewing SFT as a special case of policy gradient methods with constant positive advantages, we derive PSFT that stabilizes optimization and leads to generalization, while leaving room for further optimization in subsequent post-training stages. Experiments across mathematical and human-value domains show that PSFT matches SFT in-domain, outperforms it in out-of-domain generalization, remains stable under prolonged training without causing entropy collapse, and provides a stronger foundation for the subsequent optimization.
Authors: Arne Schneuing, Ilia Igashov, Adrian W. Dobbelstein, Thomas Castiglione, Michael Bronstein, Bruno Correia
Abstract: We introduce DrugFlow, a generative model for structure-based drug design that integrates continuous flow matching with discrete Markov bridges, demonstrating state-of-the-art performance in learning chemical, geometric, and physical aspects of three-dimensional protein-ligand data. We endow DrugFlow with an uncertainty estimate that is able to detect out-of-distribution samples. To further enhance the sampling process towards distribution regions with desirable metric values, we propose a joint preference alignment scheme applicable to both flow matching and Markov bridge frameworks. Furthermore, we extend our model to also explore the conformational landscape of the protein by jointly sampling side chain angles and molecules.
Authors: Timur Mudarisov, Mikhail Burtsev, Tatiana Petrova, Radu State
Abstract: This paper investigates the limitations of the normalization in attention mechanisms. We begin with a theoretical framework that enables the identification of the model's selective ability and the geometric separation involved in token selection. Our analysis includes explicit bounds on distances and separation criteria for token vectors under softmax scaling. Through experiments with pre-trained GPT-2 model, we empirically validate our theoretical results and analyze key behaviors of the attention mechanism. Notably, we demonstrate that as the number of selected tokens increases, the model's ability to distinguish informative tokens declines, often converging toward a uniform selection pattern. We also show that gradient sensitivity under softmax normalization presents challenges during training, especially at low temperature settings. These findings advance current understanding of softmax-based attention mechanism and motivate the need for more robust normalization and selection strategies in future attention architectures.
Authors: Jonathan Rubin, Sahil Loomba, Nick S. Jones
Abstract: Message passing neural networks (MPNNs) are powerful models for node classification but suffer from performance limitations under heterophily (low same-class connectivity) and structural bottlenecks in the graph. We provide a unifying statistical framework exposing the relationship between heterophily and bottlenecks through the signal-to-noise ratio (SNR) of MPNN representations. The SNR decomposes model performance into feature-dependent parameters and feature-independent sensitivities. We prove that the sensitivity to class-wise signals is bounded by higher-order homophily -- a generalisation of classical homophily to multi-hop neighbourhoods -- and show that low higher-order homophily manifests locally as the interaction between structural bottlenecks and class labels (class-bottlenecks). Through analysis of graph ensembles, we provide a further quantitative decomposition of bottlenecking into underreaching (lack of depth implying signals cannot arrive) and oversquashing (lack of breadth implying signals arriving on fewer paths) with closed-form expressions. We prove that optimal graph structures for maximising higher-order homophily are disjoint unions of single-class and two-class-bipartite clusters. This yields BRIDGE, a graph ensemble-based rewiring algorithm that achieves near-perfect classification accuracy across all homophily regimes on synthetic benchmarks and significant improvements on real-world benchmarks, by eliminating the ``mid-homophily pitfall'' where MPNNs typically struggle, surpassing current standard rewiring techniques from the literature. Our framework, whose code we make available for public use, provides both diagnostic tools for assessing MPNN performance, and simple yet effective methods for enhancing performance through principled graph modification.
Authors: Han Zhang, Ruibin Zheng, Zexuan Yi, Hanyang Peng, Hui Wang, Yue Yu
Abstract: As single-center computing approaches power constraints, decentralized training is becoming essential. Reinforcement Learning (RL) post-training enhances Large Language Models (LLMs) but faces challenges in heterogeneous distributed environments due to its tightly-coupled sampling-learning alternation. We propose HeteroRL, an asynchronous RL architecture that decouples rollout sampling from parameter learning, enabling robust deployment across geographically distributed nodes under network delays. We identify that latency-induced KL divergence causes importance sampling failure due to high variance. To address this, we propose Group Expectation Policy Optimization (GEPO), which reduces importance weight variance through a refined sampling mechanism. Theoretically, GEPO achieves exponential variance reduction. Experiments show it maintains superior stability over methods like GRPO, with less than 3% performance degradation under 1800-second delays, demonstrating strong potential for decentralized RL in heterogeneous networks.
Authors: Dan Wang, Feng Jiang, Zhanquan Wang
Abstract: Accurate air quality prediction is becoming increasingly important in the environmental field. To address issues such as low prediction accuracy and slow real-time updates in existing models, which lead to lagging prediction results, we propose a Transformer-based spatiotemporal data prediction method (Ada-TransGNN) that integrates global spatial semantics and temporal behavior. The model constructs an efficient and collaborative spatiotemporal block set comprising a multi-head attention mechanism and a graph convolutional network to extract dynamically changing spatiotemporal dependency features from complex air quality monitoring data. Considering the interaction relationships between different monitoring points, we propose an adaptive graph structure learning module, which combines spatiotemporal dependency features in a data-driven manner to learn the optimal graph structure, thereby more accurately capturing the spatial relationships between monitoring points. Additionally, we design an auxiliary task learning module that enhances the decoding capability of temporal relationships by integrating spatial context information into the optimal graph structure representation, effectively improving the accuracy of prediction results. We conducted comprehensive evaluations on a benchmark dataset and a novel dataset (Mete-air). The results demonstrate that our model outperforms existing state-of-the-art prediction models in short-term and long-term predictions.
Authors: Yanghao Qin, Bo Zhou, Guangliang Pan, Qihui Wu, Meixia Tao
Abstract: Accurate spectrum prediction is crucial for dynamic spectrum access (DSA) and resource allocation. However, due to the unique characteristics of spectrum data, existing methods based on the time or frequency domain often struggle to separate predictable patterns from noise. To address this, we propose the Spectral Fractional Filtering and Prediction (SFFP) framework. SFFP first employs an adaptive fractional Fourier transform (FrFT) module to transform spectrum data into a suitable fractional Fourier domain, enhancing the separability of predictable trends from noise. Subsequently, an adaptive Filter module selectively suppresses noise while preserving critical predictive features within this domain. Finally, a prediction module, leveraging a complex-valued neural network, learns and forecasts these filtered trend components. Experiments on real-world spectrum data show that the SFFP outperforms leading spectrum and general forecasting methods.
Authors: Juneyoung Park, Minjae Kang, Seongbae Lee, Haegang Lee, Seongwan Kim, Jaeho Lee
Abstract: While powerful, large language models (LLMs) present significant fine-tuning challenges due to their size. Parameter-efficient fine-tuning (PEFT) methods like LoRA provide solutions, yet suffer from critical optimizer inefficiencies; notably basis redundancy in LoRA's $B$ matrix when using AdamW, which fundamentally limits performance. We address this by optimizing the $B$ matrix on the Stiefel manifold, imposing explicit orthogonality constraints that achieve near-perfect orthogonality and full effective rank. This geometric approach dramatically enhances parameter efficiency and representational capacity. Our Stiefel optimizer consistently outperforms AdamW across benchmarks with both LoRA and DoRA, demonstrating that geometric constraints are the key to unlocking LoRA's full potential for effective LLM fine-tuning.
Authors: Sarina Penquitt, Tobias Riedlinger, Timo Heller, Markus Reischl, Matthias Rottmann
Abstract: Recently, detection of label errors and improvement of label quality in datasets for supervised learning tasks has become an increasingly important goal in both research and industry. The consequences of incorrectly annotated data include reduced model performance, biased benchmark results, and lower overall accuracy. Current state-of-the-art label error detection methods often focus on a single computer vision task and, consequently, a specific type of dataset, containing, for example, either bounding boxes or pixel-wise annotations. Furthermore, previous methods are not learning-based. In this work, we overcome this research gap. We present a unified method for detecting label errors in object detection, semantic segmentation, and instance segmentation datasets. In a nutshell, our approach - learning to detect label errors by making them - works as follows: we inject different kinds of label errors into the ground truth. Then, the detection of label errors, across all mentioned primary tasks, is framed as an instance segmentation problem based on a composite input. In our experiments, we compare the label error detection performance of our method with various baselines and state-of-the-art approaches of each task's domain on simulated label errors across multiple tasks, datasets, and base models. This is complemented by a generalization study on real-world label errors. Additionally, we release 459 real label errors identified in the Cityscapes dataset and provide a benchmark for real label error detection in Cityscapes.
Authors: Ming Yang, Dongrun Li, Xin Wang, Xiaoyang Yu, Xiaoming Wu, Shibo He
Abstract: Cross-client data heterogeneity in federated learning induces biases that impede unbiased consensus condensation and the complementary fusion of generalization- and personalization-oriented knowledge. While existing approaches mitigate heterogeneity through model decoupling and representation center loss, they often rely on static and restricted metrics to evaluate local knowledge and adopt global alignment too rigidly, leading to consensus distortion and diminished model adaptability. To address these limitations, we propose FedMate, a method that implements bilateral optimization: On the server side, we construct a dynamic global prototype, with aggregation weights calibrated by holistic integration of sample size, current parameters, and future prediction; a category-wise classifier is then fine-tuned using this prototype to preserve global consistency. On the client side, we introduce complementary classification fusion to enable merit-based discrimination training and incorporate cost-aware feature transmission to balance model performance and communication efficiency. Experiments on five datasets of varying complexity demonstrate that FedMate outperforms state-of-the-art methods in harmonizing generalization and adaptation. Additionally, semantic segmentation experiments on autonomous driving datasets validate the method's real-world scalability.
Authors: Jianhao Huang, Qunsong Zeng, Hongyang Du, Kaibin Huang
Abstract: Semantic communication (SemCom) has emerged as a promising paradigm for achieving unprecedented communication efficiency in sixth-generation (6G) networks by leveraging artificial intelligence (AI) to extract and transmit the underlying meanings of source data. However, deploying SemCom over digital systems presents new challenges, particularly in ensuring robustness against transmission errors that may distort semantically critical content. To address this issue, this paper proposes a novel framework, termed generative feature imputing, which comprises three key techniques. First, we introduce a spatial error concentration packetization strategy that spatially concentrates feature distortions by encoding feature elements based on their channel mappings, a property crucial for both the effectiveness and reduced complexity of the subsequent techniques. Second, building on this strategy, we propose a generative feature imputing method that utilizes a diffusion model to efficiently reconstruct missing features caused by packet losses. Finally, we develop a semantic-aware power allocation scheme that enables unequal error protection by allocating transmission power according to the semantic importance of each packet. Experimental results demonstrate that the proposed framework outperforms conventional approaches, such as Deep Joint Source-Channel Coding (DJSCC) and JPEG2000, under block fading conditions, achieving higher semantic accuracy and lower Learned Perceptual Image Patch Similarity (LPIPS) scores.
Authors: Mohamed Kissi, Keanu Sisouk, Joshua A. Levine, Julien Tierny
Abstract: This paper presents a neural scheme for the topology-aware interpolation of time-varying scalar fields. Given a time-varying sequence of persistence diagrams, along with a sparse temporal sampling of the corresponding scalar fields, denoted as keyframes, our interpolation approach aims at "inverting" the non-keyframe diagrams to produce plausible estimations of the corresponding, missing data. For this, we rely on a neural architecture which learns the relation from a time value to the corresponding scalar field, based on the keyframe examples, and reliably extends this relation to the non-keyframe time steps. We show how augmenting this architecture with specific topological losses exploiting the input diagrams both improves the geometrical and topological reconstruction of the non-keyframe time steps. At query time, given an input time value for which an interpolation is desired, our approach instantaneously produces an output, via a single propagation of the time input through the network. Experiments interpolating 2D and 3D time-varying datasets show our approach superiority, both in terms of data and topological fitting, with regard to reference interpolation schemes.
Authors: Sebastian G. Gruber
Abstract: In this PhD thesis, we propose a novel framework for uncertainty quantification in machine learning, which is based on proper scores. Uncertainty quantification is an important cornerstone for trustworthy and reliable machine learning applications in practice. Usually, approaches to uncertainty quantification are problem-specific, and solutions and insights cannot be readily transferred from one task to another. Proper scores are loss functions minimized by predicting the target distribution. Due to their very general definition, proper scores apply to regression, classification, or even generative modeling tasks. We contribute several theoretical results, that connect epistemic uncertainty, aleatoric uncertainty, and model calibration with proper scores, resulting in a general and widely applicable framework. We achieve this by introducing a general bias-variance decomposition for strictly proper scores via functional Bregman divergences. Specifically, we use the kernel score, a kernel-based proper score, for evaluating sample-based generative models in various domains, like image, audio, and natural language generation. This includes a novel approach for uncertainty estimation of large language models, which outperforms state-of-the-art baselines. Further, we generalize the calibration-sharpness decomposition beyond classification, which motivates the definition of proper calibration errors. We then introduce a novel estimator for proper calibration errors in classification, and a novel risk-based approach to compare different estimators for squared calibration errors. Last, we offer a decomposition of the kernel spherical score, another kernel-based proper score, allowing a more fine-grained and interpretable evaluation of generative image models.
Authors: William Brooks, Marelie H. Davel, Coenraad Mouton
Abstract: Deep Neural Networks (DNNs) have shown substantial success in various applications but remain vulnerable to adversarial attacks. This study aims to identify and isolate the components of two different adversarial training techniques that contribute most to increased adversarial robustness, particularly through the lens of margins in the input space -- the minimal distance between data points and decision boundaries. Specifically, we compare two methods that maximize margins: a simple approach which modifies the loss function to increase an approximation of the margin, and a more complex state-of-the-art method (Dynamics-Aware Robust Training) which builds upon this approach. Using a VGG-16 model as our base, we systematically isolate and evaluate individual components from these methods to determine their relative impact on adversarial robustness. We assess the effect of each component on the model's performance under various adversarial attacks, including AutoAttack and Projected Gradient Descent (PGD). Our analysis on the CIFAR-10 dataset reveals which elements most effectively enhance adversarial robustness, providing insights for designing more robust DNNs.
Authors: Aditri Paul, Archan Paul
Abstract: Autonomous planetary exploration missions are critically dependent on real-time, accurate environmental perception for navigation and hazard avoidance. However, deploying deep learning models on the resource-constrained computational hardware of planetary exploration platforms remains a significant challenge. This paper introduces the Adaptive Quantized Planetary Crater Detection System (AQ-PCDSys), a novel framework specifically engineered for real-time, onboard deployment in the computationally constrained environments of space exploration missions. AQ-PCDSys synergistically integrates a Quantized Neural Network (QNN) architecture, trained using Quantization-Aware Training (QAT), with an Adaptive Multi-Sensor Fusion (AMF) module. The QNN architecture significantly optimizes model size and inference latency suitable for real-time onboard deployment in space exploration missions, while preserving high accuracy. The AMF module intelligently fuses data from Optical Imagery (OI) and Digital Elevation Models (DEMs) at the feature level, utilizing an Adaptive Weighting Mechanism (AWM) to dynamically prioritize the most relevant and reliable sensor modality based on planetary ambient conditions. This approach enhances detection robustness across diverse planetary landscapes. Paired with Multi-Scale Detection Heads specifically designed for robust and efficient detection of craters across a wide range of sizes, AQ-PCDSys provides a computationally efficient, reliable and accurate solution for planetary crater detection, a critical capability for enabling the next generation of autonomous planetary landing, navigation, and scientific exploration.
Authors: Zilong Cao (The School of Mathematics, Northwest University), Hai Zhang (The School of Mathematics, Northwest University)
Abstract: Leveraging information from public data has become increasingly crucial in enhancing the utility of differentially private (DP) methods. Traditional DP approaches often require adding noise based solely on private data, which can significantly degrade utility. In this paper, we address this limitation in the context of the ordinary least squares estimator (OLSE) of linear regression based on sufficient statistics perturbation (SSP) under the unbounded data assumption. We propose a novel method that involves transforming private data using the public second-moment matrix to compute a transformed SSP-OLSE, whose second-moment matrix yields a better condition number and improves the OLSE accuracy and robustness. We derive theoretical error bounds about our method and the standard SSP-OLSE to the non-DP OLSE, which reveal the improved robustness and accuracy achieved by our approach. Experiments on synthetic and real-world datasets demonstrate the utility and effectiveness of our method.
Authors: Xiuheng Wang, Ricardo Borsoi, Arnaud Breloy, C\'edric Richard
Abstract: Non-parametric change-point detection in streaming time series data is a long-standing challenge in signal processing. Recent advancements in statistics and machine learning have increasingly addressed this problem for data residing on Riemannian manifolds. One prominent strategy involves monitoring abrupt changes in the center of mass of the time series. Implemented in a streaming fashion, this strategy, however, requires careful step size tuning when computing the updates of the center of mass. In this paper, we propose to leverage robust centroid on manifolds from M-estimation theory to address this issue. Our proposal consists of comparing two centroid estimates: the classical Karcher mean (sensitive to change) versus one defined from Huber's function (robust to change). This comparison leads to the definition of a test statistic whose performance is less sensitive to the underlying estimation method. We propose a stochastic Riemannian optimization algorithm to estimate both robust centroids efficiently. Experiments conducted on both simulated and real-world data across two representative manifolds demonstrate the superior performance of our proposed method.
Authors: Paul Garnier, Vincent Lannelongue, Jonathan Viquerat, Elie Hachem
Abstract: Simulating physics using Graph Neural Networks (GNNs) is predominantly driven by message-passing architectures, which face challenges in scaling and efficiency, particularly in handling large, complex meshes. These architectures have inspired numerous enhancements, including multigrid approaches and $K$-hop aggregation (using neighbours of distance $K$), yet they often introduce significant complexity and suffer from limited in-depth investigations. In response to these challenges, we propose a novel Graph Transformer architecture that leverages the adjacency matrix as an attention mask. The proposed approach incorporates innovative augmentations, including Dilated Sliding Windows and Global Attention, to extend receptive fields without sacrificing computational efficiency. Through extensive experimentation, we evaluate model size, adjacency matrix augmentations, positional encoding and $K$-hop configurations using challenging 3D computational fluid dynamics (CFD) datasets. We also train over 60 models to find a scaling law between training FLOPs and parameters. The introduced models demonstrate remarkable scalability, performing on meshes with up to 300k nodes and 3 million edges. Notably, the smallest model achieves parity with MeshGraphNet while being $7\times$ faster and $6\times$ smaller. The largest model surpasses the previous state-of-the-art by $38.8$\% on average and outperforms MeshGraphNet by $52$\% on the all-rollout RMSE, while having a similar training speed. Code and datasets are available at https://github.com/DonsetPG/graph-physics.
Authors: Silvia Beddar-Wiesing, Alice Moallemy-Oureh
Abstract: Graph Neural Networks (GNNs) are known to match the distinguishing power of the 1-Weisfeiler-Lehman (1-WL) test, and the resulting partitions coincide with the unfolding tree equivalence classes of graphs. Preserving this equivalence, GNNs can universally approximate any target function on graphs in probability up to any precision. However, these results are limited to attributed discrete-dynamic graphs represented as sequences of connected graph snapshots. Real-world systems, such as communication networks, financial transaction networks, and molecular interactions, evolve asynchronously and may split into disconnected components. In this paper, we extend the theory of attributed discrete-dynamic graphs to attributed continuous-time dynamic graphs with arbitrary connectivity. To this end, we introduce a continuous-time dynamic 1-WL test, prove its equivalence to continuous-time dynamic unfolding trees, and identify a class of continuous-time dynamic GNNs (CGNNs) based on discrete-dynamic GNN architectures that retain both distinguishing power and universal approximation guarantees. Our constructive proofs further yield practical design guidelines, emphasizing a compact and expressive CGNN architecture with piece-wise continuously differentiable temporal functions to process asynchronous, disconnected graphs.
Authors: Emmanouil Kritharakis, Antonios Makris, Dusan Jakovetic, Konstantinos Tserpes
Abstract: Federated Learning (FL) enables collaborative model training across multiple clients while preserving data privacy by keeping local datasets on-device. In this work, we address FL settings where clients may behave adversarially, exhibiting Byzantine attacks, while the central server is trusted and equipped with a reference dataset. We propose FedGreed, a resilient aggregation strategy for federated learning that does not require any assumptions about the fraction of adversarial participants. FedGreed orders clients' local model updates based on their loss metrics evaluated against a trusted dataset on the server and greedily selects a subset of clients whose models exhibit the minimal evaluation loss. Unlike many existing approaches, our method is designed to operate reliably under heterogeneous (non-IID) data distributions, which are prevalent in real-world deployments. FedGreed exhibits convergence guarantees and bounded optimality gaps under strong adversarial behavior. Experimental evaluations on MNIST, FMNIST, and CIFAR-10 demonstrate that our method significantly outperforms standard and robust federated learning baselines, such as Mean, Trimmed Mean, Median, Krum, and Multi-Krum, in the majority of adversarial scenarios considered, including label flipping and Gaussian noise injection attacks. All experiments were conducted using the Flower federated learning framework.
Authors: Abyad Enan, Mashrur Chowdhury, Sagar Dasgupta, Mizanur Rahman
Abstract: Global Navigation Satellite Systems (GNSS) are critical for Positioning, Navigation, and Timing (PNT) applications. However, GNSS are highly vulnerable to spoofing attacks, where adversaries transmit counterfeit signals to mislead receivers. Such attacks can lead to severe consequences, including misdirected navigation, compromised data integrity, and operational disruptions. Most existing spoofing detection methods depend on supervised learning techniques and struggle to detect novel, evolved, and unseen attacks. To overcome this limitation, we develop a zero-day spoofing detection method using a Hybrid Quantum-Classical Autoencoder (HQC-AE), trained solely on authentic GNSS signals without exposure to spoofed data. By leveraging features extracted during the tracking stage, our method enables proactive detection before PNT solutions are computed. We focus on spoofing detection in static GNSS receivers, which are particularly susceptible to time-push spoofing attacks, where attackers manipulate timing information to induce incorrect time computations at the receiver. We evaluate our model against different unseen time-push spoofing attack scenarios: simplistic, intermediate, and sophisticated. Our analysis demonstrates that the HQC-AE consistently outperforms its classical counterpart, traditional supervised learning-based models, and existing unsupervised learning-based methods in detecting zero-day, unseen GNSS time-push spoofing attacks, achieving an average detection accuracy of 97.71% with an average false negative rate of 0.62% (when an attack occurs but is not detected). For sophisticated spoofing attacks, the HQC-AE attains an accuracy of 98.23% with a false negative rate of 1.85%. These findings highlight the effectiveness of our method in proactively detecting zero-day GNSS time-push spoofing attacks across various stationary GNSS receiver platforms.
Authors: Paul Hagemann, Robert Gruhlke, Bernhard Stankewitz, Claudia Schillings, Gabriele Steidl
Abstract: We study Bayesian inverse problems with mixed noise, modeled as a combination of additive and multiplicative Gaussian components. While traditional inference methods often assume fixed or known noise characteristics, real-world applications, particularly in physics and chemistry, frequently involve noise with unknown and heterogeneous structure. Motivated by recent advances in flow-based generative modeling, we propose a novel inference framework based on conditional flow matching embedded within an Expectation-Maximization (EM) algorithm to jointly estimate posterior samplers and noise parameters. To enable high-dimensional inference and improve scalability, we use simulation-free ODE-based flow matching as the generative model in the E-step of the EM algorithm. We prove that, under suitable assumptions, the EM updates converge to the true noise parameters in the population limit of infinite observations. Our numerical results illustrate the effectiveness of combining EM inference with flow matching for mixed-noise Bayesian inverse problems.
Authors: Weida Wang, Dongchen Huang, Jiatong Li, Tengchao Yang, Ziyang Zheng, Di Zhang, Dong Han, Benteng Chen, Binzhao Luo, Zhiyu Liu, Kunling Liu, Zhiyuan Gao, Shiqi Geng, Wei Ma, Jiaming Su, Xin Li, Shuchen Pu, Yuhan Shui, Qianjia Cheng, Zhihao Dou, Dongfei Cui, Changyong He, Jin Zeng, Zeke Xie, Mao Su, Dongzhan Zhou, Yuqiang Li, Wanli Ouyang, Lei Bai, Yunqi Cai, Xi Dai, Shufei Zhang, Jinguang Cheng, Zhong Fang, Hongming Weng
Abstract: We introduce CMPhysBench, designed to assess the proficiency of Large Language Models (LLMs) in Condensed Matter Physics, as a novel Benchmark. CMPhysBench is composed of more than 520 graduate-level meticulously curated questions covering both representative subfields and foundational theoretical frameworks of condensed matter physics, such as magnetism, superconductivity, strongly correlated systems, etc. To ensure a deep understanding of the problem-solving process,we focus exclusively on calculation problems, requiring LLMs to independently generate comprehensive solutions. Meanwhile, leveraging tree-based representations of expressions, we introduce the Scalable Expression Edit Distance (SEED) score, which provides fine-grained (non-binary) partial credit and yields a more accurate assessment of similarity between prediction and ground-truth. Our results show that even the best models, Grok-4, reach only 36 average SEED score and 28% accuracy on CMPhysBench, underscoring a significant capability gap, especially for this practical and frontier domain relative to traditional physics. The code anddataset are publicly available at https://github.com/CMPhysBench/CMPhysBench.
Authors: Pradeep Singh, Mehak Sharma, Anupriya Dey, Balasubramanian Raman
Abstract: Transformers are the de-facto choice for sequence modelling, yet their quadratic self-attention and weak temporal bias can make long-range forecasting both expensive and brittle. We introduce FreezeTST, a lightweight hybrid that interleaves frozen random-feature (reservoir) blocks with standard trainable Transformer layers. The frozen blocks endow the network with rich nonlinear memory at no optimisation cost; the trainable layers learn to query this memory through self-attention. The design cuts trainable parameters and also lowers wall-clock training time, while leaving inference complexity unchanged. On seven standard long-term forecasting benchmarks, FreezeTST consistently matches or surpasses specialised variants such as Informer, Autoformer, and PatchTST; with substantially lower compute. Our results show that embedding reservoir principles within Transformers offers a simple, principled route to efficient long-term time-series prediction.
Authors: Riccardo Cappi, Paolo Frazzetto, Nicol\`o Navarin, Alessandro Sperduti
Abstract: The ``black-box'' nature of deep learning models presents a significant barrier to their adoption for scientific discovery, where interpretability is paramount. This challenge is especially pronounced in discovering the governing equations of dynamical processes on networks or graphs, since even their topological structure further affects the processes' behavior. This paper provides a rigorous, comparative assessment of state-of-the-art symbolic regression techniques for this task. We evaluate established methods, including sparse regression and MLP-based architectures, and introduce a novel adaptation of Kolmogorov-Arnold Networks (KANs) for graphs, designed to exploit their inherent interpretability. Across a suite of synthetic and real-world dynamical systems, our results demonstrate that both MLP and KAN-based architectures can successfully identify the underlying symbolic equations, significantly surpassing existing baselines. Critically, we show that KANs achieve this performance with greater parsimony and transparency, as their learnable activation functions provide a clearer mapping to the true physical dynamics. This study offers a practical guide for researchers, clarifying the trade-offs between model expressivity and interpretability, and establishes the viability of neural-based architectures for robust scientific discovery on complex systems.
Authors: Charlie B. Tan, Majdi Hassan, Leon Klein, Saifuddin Syed, Dominique Beaini, Michael M. Bronstein, Alexander Tong, Kirill Neklyudov
Abstract: Efficient equilibrium sampling of molecular conformations remains a core challenge in computational chemistry and statistical inference. Classical approaches such as molecular dynamics or Markov chain Monte Carlo inherently lack amortization; the computational cost of sampling must be paid in-full for each system of interest. The widespread success of generative models has inspired interest into overcoming this limitation through learning sampling algorithms. Despite performing on par with conventional methods when trained on a single system, learned samplers have so far demonstrated limited ability to transfer across systems. We prove that deep learning enables the design of scalable and transferable samplers by introducing Prose, a 280 million parameter all-atom transferable normalizing flow trained on a corpus of peptide molecular dynamics trajectories up to 8 residues in length. Prose draws zero-shot uncorrelated proposal samples for arbitrary peptide systems, achieving the previously intractable transferability across sequence length, whilst retaining the efficient likelihood evaluation of normalizing flows. Through extensive empirical evaluation we demonstrate the efficacy of Prose as a proposal for a variety of sampling algorithms, finding a simple importance sampling-based finetuning procedure to achieve superior performance to established methods such as sequential Monte Carlo on unseen tetrapeptides. We open-source the Prose codebase, model weights, and training dataset, to further stimulate research into amortized sampling methods and finetuning objectives.
Authors: Nikolay Kutuzov, Makar Baderko, Stepan Kulibaba, Artem Dzhalilov, Daniel Bobrov, Maxim Mashtaler, Alexander Gasnikov
Abstract: Scaling distributed training of Large Language Models (LLMs) requires not only algorithmic advances but also efficient utilization of heterogeneous hardware resources. While existing methods such as DiLoCo have demonstrated promising results, they often fail to fully exploit computational clusters under dynamic workloads. To address this limitation, we propose a three-stage method that combines Multi-Instance Training (MIT), Adaptive Batched DiLoCo, and switch mode mechanism. MIT allows individual nodes to run multiple lightweight training streams with different model instances in parallel and merge them to combine knowledge, increasing throughput and reducing idle time. Adaptive Batched DiLoCo dynamically adjusts local batch sizes to balance computation and communication, substantially lowering synchronization delays. Switch mode further stabilizes training by seamlessly introducing gradient accumulation once adaptive batch sizes grow beyond hardware-friendly limits. Together, these innovations improve both convergence speed and system efficiency. We also provide a theoretical estimate of the number of communications required for the full convergence of a model trained using our method.
Authors: Pradeep Singh, Sutirtha Ghosh, Ashutosh Kumar, Hrishit B P, Balasubramanian Raman
Abstract: Forecasting chaotic dynamics beyond a few Lyapunov times is difficult because infinitesimal errors grow exponentially. Existing Echo State Networks (ESNs) mitigate this growth but employ reservoirs whose Euclidean geometry is mismatched to the stretch-and-fold structure of chaos. We introduce the Hyperbolic Embedding Reservoir (HypER), an ESN whose neurons are sampled in the Poincare ball and whose connections decay exponentially with hyperbolic distance. This negative-curvature construction embeds an exponential metric directly into the latent space, aligning the reservoir's local expansion-contraction spectrum with the system's Lyapunov directions while preserving standard ESN features such as sparsity, leaky integration, and spectral-radius control. Training is limited to a Tikhonov-regularized readout. On the chaotic Lorenz-63 and Roessler systems, and the hyperchaotic Chen-Ueta attractor, HypER consistently lengthens the mean valid-prediction horizon beyond Euclidean and graph-structured ESN baselines, with statistically significant gains confirmed over 30 independent runs; parallel results on real-world benchmarks, including heart-rate variability from the Santa Fe and MIT-BIH datasets and international sunspot numbers, corroborate its advantage. We further establish a lower bound on the rate of state divergence for HypER, mirroring Lyapunov growth.
Authors: Sunwoo Kim
Abstract: In this paper, we propose a deep learning and matrix completion aided approach for recovering an outlier contaminated Euclidean distance matrix D in IoT network localization. Unlike conventional localization techniques that search the solution over a whole set of matrices, the proposed technique restricts the search to the set of Euclidean distance matrices. Specifically, we express D as a function of the sensor coordinate matrix X that inherently satisfies the unique properties of D, and then jointly recover D and X using a deep neural network. To handle outliers effectively, we model them as a sparse matrix L and add a regularization term of L into the optimization problem. We then solve the problem by alternately updating X, D, and L. Numerical experiments demonstrate that the proposed technique can recover the location information of sensors accurately even in the presence of outliers.
Authors: Chu-Cheng Lin, Daiyi Peng, Yifeng Lu, Ming Zhang, Eugene Ie
Abstract: Reliably composing Large Language Models (LLMs) for complex, multi-step workflows remains a significant challenge. The dominant paradigm-optimizing discrete prompts in a pipeline-is notoriously brittle and struggles to enforce the formal compliance required for structured tasks. We introduce Type-Compliant Adaptation Cascades (TACs), a framework that recasts workflow adaptation as learning typed probabilistic programs. TACs treats the entire workflow, which is composed of parameter-efficiently adapted LLMs and deterministic logic, as an unnormalized joint distribution. This enables principled, gradient-based training even with latent intermediate structures. We provide theoretical justification for our tractable optimization objective, proving that the optimization bias vanishes as the model learns type compliance. Empirically, TACs significantly outperforms state-of-the-art prompt-optimization baselines. Gains are particularly pronounced on structured tasks, improving MGSM-SymPy from $57.1\%$ to $75.9\%$ for a 27B model, MGSM from $1.6\%$ to $27.3\%$ for a 7B model. TACs offers a robust and theoretically grounded paradigm for developing reliable, task-compliant LLM systems.
Authors: Novin Shahroudi, Viacheslav Komisarenko, Meelis Kull
Abstract: Every prediction is ultimately used in a downstream task. Consequently, evaluating prediction quality is more meaningful when considered in the context of its downstream use. Metrics based solely on predictive performance often diverge from measures of real-world downstream impact. Existing approaches incorporate the downstream view by relying on multiple task-specific metrics, which can be burdensome to analyze, or by formulating cost-sensitive evaluations that require an explicit cost structure, typically assumed to be known a priori. We frame this mismatch as an evaluation alignment problem and propose a data-driven method to learn a proxy evaluation function aligned with the downstream evaluation. Building on the theory of proper scoring rules, we explore transformations of scoring rules that ensure the preservation of propriety. Our approach leverages weighted scoring rules parametrized by a neural network, where weighting is learned to align with the performance in the downstream task. This enables fast and scalable evaluation cycles across tasks where the weighting is complex or unknown a priori. We showcase our framework through synthetic and real-data experiments for regression tasks, demonstrating its potential to bridge the gap between predictive evaluation and downstream utility in modular prediction systems.
Authors: Adrien Kegreisz
Abstract: Stochastic optimizers are central to deep learning, yet widely used methods such as Adam and Adan can degrade in non-stationary or noisy environments, partly due to their reliance on momentum-based magnitude estimates. We introduce Ano, a novel optimizer that decouples direction and magnitude: momentum is used for directional smoothing, while instantaneous gradient magnitudes determine step size. This design improves robustness to gradient noise while retaining the simplicity and efficiency of first-order methods. We further propose Anolog, which removes sensitivity to the momentum coefficient by expanding its window over time via a logarithmic schedule. We establish non-convex convergence guarantees with a convergence rate similar to other sign-based methods, and empirically show that Ano provides substantial gains in noisy and non-stationary regimes such as reinforcement learning, while remaining competitive on low-noise tasks such as standard computer vision benchmarks.
Authors: Jaydip Sen, Subhasis Dasgupta, Hetvi Waghela
Abstract: Speculative decoding has emerged as an effective approach for accelerating autoregressive inference by parallelizing token generation through a draft-then-verify paradigm. However, existing methods rely on static drafting lengths and rigid verification criteria, limiting their adaptability across varying model uncertainties and input complexities. This paper proposes an information-theoretic framework for speculative decoding based on confidence-modulated drafting. By leveraging entropy and margin-based uncertainty measures over the drafter's output distribution, the proposed method dynamically adjusts the number of speculatively generated tokens at each iteration. This adaptive mechanism reduces rollback frequency, improves resource utilization, and maintains output fidelity. Additionally, the verification process is modulated using the same confidence signals, enabling more flexible acceptance of drafted tokens without sacrificing generation quality. Experiments on machine translation and summarization tasks demonstrate significant speedups over standard speculative decoding while preserving or improving BLEU and ROUGE scores. The proposed approach offers a principled, plug-in method for efficient and robust decoding in large language models under varying conditions of uncertainty.
Authors: Micha{\l} Patryk Miazga, Patrick Ebel
Abstract: Biomechanical forward simulation holds great potential for HCI, enabling the generation of human-like movements in interactive tasks. However, training biomechanical models with reinforcement learning is challenging, particularly for precise and dexterous movements like those required for touchscreen interactions on mobile devices. Current approaches are limited in their interaction fidelity, require restricting the underlying biomechanical model to reduce complexity, and do not generalize well. In this work, we propose practical improvements to training routines that reduce training time, increase interaction fidelity beyond existing methods, and enable the use of more complex biomechanical models. Using a touchscreen pointing task, we demonstrate that curriculum learning, action masking, more complex network configurations, and simple adjustments to the simulation environment can significantly improve the agent's ability to learn accurate touch behavior. Our work provides HCI researchers with practical tips and training routines for developing better biomechanical models of human-like interaction fidelity.
Authors: Linghao Zeng
Abstract: Predicting user intentions in virtual reality (VR) is crucial for creating immersive experiences, particularly in tasks involving complex grasping motions where accurate haptic feedback is essential. In this work, we leverage time-series data from hand movements to evaluate both classification and regression approaches across 810 trials with varied object types, sizes, and manipulations. Our findings reveal that classification models struggle to generalize across users, leading to inconsistent performance. In contrast, regression-based approaches, particularly those using Long Short Term Memory (LSTM) networks, demonstrate more robust performance, with timing errors within 0.25 seconds and distance errors around 5-20 cm in the critical two-second window before a grasp. Despite these improvements, predicting precise hand postures remains challenging. Through a comprehensive analysis of user variability and model interpretability, we explore why certain models fail and how regression models better accommodate the dynamic and complex nature of user behavior in VR. Our results underscore the potential of machine learning models to enhance VR interactions, particularly through adaptive haptic feedback, and lay the groundwork for future advancements in real-time prediction of user actions in VR.
Authors: Rakesh Thakur, Riya Gupta
Abstract: Comprehending the long-timescale dynamics of protein-ligand complexes is very important for drug discovery and structural biology, but it continues to be computationally challenging for large biomolecular systems. We introduce HemePLM-Diffuse, an innovative generative transformer model that is designed for accurate simulation of protein-ligand trajectories, inpaints the missing ligand fragments, and sample transition paths in systems with more than 10,000 atoms. HemePLM-Diffuse has features of SE(3)-Invariant tokenization approach for proteins and ligands, that utilizes time-aware cross-attentional diffusion to effectively capture atomic motion. We also demonstrate its capabilities using the 3CQV HEME system, showing enhanced accuracy and scalability compared to leading models such as TorchMD-Net, MDGEN, and Uni-Mol.
Authors: Yanwen Wang, Xinglin Zhao, Yijin Song, Xiaobo Liu, Yanrong Hao, Rui Cao, Xin Wen
Abstract: Functional connectivity (FC) derived from resting-state fMRI plays a critical role in personalized predictions such as age and cognitive performance. However, applying foundation models(FM) to fMRI data remains challenging due to its high dimensionality, computational complexity, and the difficulty in capturing complex spatiotemporal dynamics and indirect region-of-interest (ROI) interactions. To address these limitations, we propose a modular neuroimaging framework that integrates principles from FM with efficient, domain-specific architectures. Our approach begins with a Local Masked Autoencoder (LMAE) for pretraining, which reduces the influence of hemodynamic response function (HRF) dynamics and suppresses noise. This is followed by a Random Walk Mixture of Experts (RWMOE) module that clusters features across spatial and temporal dimensions, effectively capturing intricate brain interactions. Finally, a state-space model (SSM)-based predictor performs downstream task inference. Evaluated on the Cambridge Centre for Ageing and Neuroscience (Cam-CAN) dataset, our framework achieved mean absolute errors (MAEs) of 5.343 for age prediction and 2.940 for fluid intelligence, with Pearson correlation coefficients (PCCs) of 0.928 and 0.887, respectively-outperforming existing state-of-the-art methods. Visualization of expert distribution weights further enhances interpretability by identifying key brain regions. This work provides a robust, interpretable alternative to LLM-based approaches for fMRI analysis, offering novel insights into brain aging and cognitive function.
Authors: Zheng Dong, Luming Shang, Gabriela Olinto
Abstract: High-quality prompts are crucial for Large Language Models (LLMs) to achieve exceptional performance. However, manually crafting effective prompts is labor-intensive and demands significant domain expertise, limiting its scalability. Existing automatic prompt optimization methods either extensively explore new prompt candidates, incurring high computational costs due to inefficient searches within a large solution space, or overly exploit feedback on existing prompts, risking suboptimal optimization because of the complex prompt landscape. To address these challenges, we introduce GreenTEA, an agentic LLM workflow for automatic prompt optimization that balances candidate exploration and knowledge exploitation. It leverages a collaborative team of agents to iteratively refine prompts based on feedback from error samples. An analyzing agent identifies common error patterns resulting from the current prompt via topic modeling, and a generation agent revises the prompt to directly address these key deficiencies. This refinement process is guided by a genetic algorithm framework, which simulates natural selection by evolving candidate prompts through operations such as crossover and mutation to progressively optimize model performance. Extensive numerical experiments conducted on public benchmark datasets suggest the superior performance of GreenTEA against human-engineered prompts and existing state-of-the-arts for automatic prompt optimization, covering logical and quantitative reasoning, commonsense, and ethical decision-making.
Authors: Maximilian Burzer, Tobias King, Till Riedel, Michael Beigl, Tobias R\"oddiger
Abstract: The lack of standardization across Wearable Human Activity Recognition (WHAR) datasets limits reproducibility, comparability, and research efficiency. We introduce WHAR datasets, an open-source library designed to simplify WHAR data handling through a standardized data format and a configuration-driven design, enabling reproducible and computationally efficient workflows with minimal manual intervention. The library currently supports 9 widely-used datasets, integrates with PyTorch and TensorFlow, and is easily extensible to new datasets. To demonstrate its utility, we trained two state-of-the-art models, TinyHar and MLP-HAR, on the included datasets, approximately reproducing published results and validating the library's effectiveness for experimentation and benchmarking. Additionally, we evaluated preprocessing performance and observed speedups of up to 3.8x using multiprocessing. We hope this library contributes to more efficient, reproducible, and comparable WHAR research.
Authors: Yogesh Kumar Meena, Manish Salvi
Abstract: Over the past decade, the demand for communication devices has increased among individuals with mobility and speech impairments. Eye-gaze tracking has emerged as a promising solution for hands-free communication; however, traditional appearance-based interfaces often face challenges such as accuracy issues, involuntary eye movements, and difficulties with extensive command sets. This work presents a multimodal appearance-based gaze-controlled virtual keyboard that utilises deep learning in conjunction with standard camera hardware, incorporating both synchronous and asynchronous modes for command selection. The virtual keyboard application supports menu-based selection with nine commands, enabling users to spell and type up to 56 English characters, including uppercase and lowercase letters, punctuation, and a delete function for corrections. The proposed system was evaluated with twenty able-bodied participants who completed specially designed typing tasks using three input modalities: (i) a mouse, (ii) an eye-tracker, and (iii) an unmodified webcam. Typing performance was measured in terms of speed and information transfer rate (ITR) at both command and letter levels. Average typing speeds were 18.3+-5.31 letters/min (mouse), 12.60+-2.99letters/min (eye-tracker, synchronous), 10.94 +- 1.89 letters/min (webcam, synchronous), 11.15 +- 2.90 letters/min (eye-tracker, asynchronous), and 7.86 +- 1.69 letters/min (webcam, asynchronous). ITRs were approximately 80.29 +- 15.72 bits/min (command level) and 63.56 +- 11 bits/min (letter level) with webcam in synchronous mode. The system demonstrated good usability and low workload with webcam input, highlighting its user-centred design and promise as an accessible communication tool in low-resource settings.
Authors: Zhao Feng, Xin-Yang Liu, Meet Hemant Parikh, Junyi Guo, Pan Du, Bicheng Yan, Jian-Xun Wang
Abstract: Geological Carbon Sequestration (GCS) has emerged as a promising strategy for mitigating global warming, yet its effectiveness heavily depends on accurately characterizing subsurface flow dynamics. The inherent geological uncertainty, stemming from limited observations and reservoir heterogeneity, poses significant challenges to predictive modeling. Existing methods for inverse modeling and uncertainty quantification are computationally intensive and lack generalizability, restricting their practical utility. Here, we introduce a Conditional Neural Field Latent Diffusion (CoNFiLD-geo) model, a generative framework for efficient and uncertainty-aware forward and inverse modeling of GCS processes. CoNFiLD-geo synergistically combines conditional neural field encoding with Bayesian conditional latent-space diffusion models, enabling zero-shot conditional generation of geomodels and reservoir responses across complex geometries and grid structures. The model is pretrained unconditionally in a self-supervised manner, followed by a Bayesian posterior sampling process, allowing for data assimilation for unseen/unobserved states without task-specific retraining. Comprehensive validation across synthetic and real-world GCS scenarios demonstrates CoNFiLD-geo's superior efficiency, generalization, scalability, and robustness. By enabling effective data assimilation, uncertainty quantification, and reliable forward modeling, CoNFiLD-geo significantly advances intelligent decision-making in geo-energy systems, supporting the transition toward a sustainable, net-zero carbon future.
Authors: Naren Sengodan
Abstract: Fine-Grained Visual Classification (FGVC) is a critical and challenging area within computer vision, demanding the identification of highly subtle, localized visual cues. The importance of FGVC extends to critical applications such as biodiversity monitoring and medical diagnostics, where precision is paramount. While large-scale Vision Transformers have achieved state-of-the-art performance, their decision-making processes often lack the interpretability required for trust and verification in such domains. In this paper, we introduce The Loupe, a novel, lightweight, and plug-and-play attention module designed to be inserted into pre-trained backbones like the Swin Transformer. The Loupe is trained end-to-end with a composite loss function that implicitly guides the model to focus on the most discriminative object parts without requiring explicit part-level annotations. Our unique contribution lies in demonstrating that a simple, intrinsic attention mechanism can act as a powerful regularizer, significantly boosting performance while simultaneously providing clear visual explanations. Our experimental evaluation on the challenging CUB-200-2011 dataset shows that The Loupe improves the accuracy of a Swin-Base model from 85.40% to 88.06%, a significant gain of 2.66%. Crucially, our qualitative analysis of the learned attention maps reveals that The Loupe effectively localizes semantically meaningful features, providing a valuable tool for understanding and trusting the model's decision-making process.
Authors: Deborup Sanyal
Abstract: COVID19 took the world by storm since December 2019. A highly infectious communicable disease, COVID19 is caused by the SARSCoV2 virus. By March 2020, the World Health Organization (WHO) declared COVID19 as a global pandemic. A pandemic in the 21st century after almost 100 years was something the world was not prepared for, which resulted in the deaths of around 1.6 million people worldwide. The most common symptoms of COVID19 were associated with the respiratory system and resembled a cold, flu, or pneumonia. After extensive research, doctors and scientists concluded that the main reason for lives being lost due to COVID19 was failure of the respiratory system. Patients were dying gasping for breath. Top healthcare systems of the world were failing badly as there was an acute shortage of hospital beds, oxygen cylinders, and ventilators. Many were dying without receiving any treatment at all. The aim of this project is to help doctors decide the severity of COVID19 by reading the patient's Computed Tomography (CT) scans of the lungs. Computer models are less prone to human error, and Machine Learning or Neural Network models tend to give better accuracy as training improves over time. We have decided to use a Convolutional Neural Network model. Given that a patient tests positive, our model will analyze the severity of COVID19 infection within one month of the positive test result. The severity of the infection may be promising or unfavorable (if it leads to intubation or death), based entirely on the CT scans in the dataset.
Authors: Nicole Cho, William Watson, Alec Koppel, Sumitra Ganesh, Manuela Veloso
Abstract: Advanced reasoning capabilities in Large Language Models (LLMs) have caused higher hallucination prevalence; yet most mitigation work focuses on after-the-fact filtering rather than shaping the queries that trigger them. We introduce QueryBandits, a bandit framework that designs rewrite strategies to maximize a reward model, that encapsulates hallucination propensity based upon the sensitivities of 17 linguistic features of the input query-and therefore, proactively steer LLMs away from generating hallucinations. Across 13 diverse QA benchmarks and 1,050 lexically perturbed queries per dataset, our top contextual QueryBandit (Thompson Sampling) achieves an 87.5% win rate over a no-rewrite baseline and also outperforms zero-shot static prompting ("paraphrase" or "expand") by 42.6% and 60.3% respectively. Therefore, we empirically substantiate the effectiveness of QueryBandits in mitigating hallucination via the intervention that takes the form of a query rewrite. Interestingly, certain static prompting strategies, which constitute a considerable number of current query rewriting literature, have a higher cumulative regret than the no-rewrite baseline, signifying that static rewrites can worsen hallucination. Moreover, we discover that the converged per-arm regression feature weight vectors substantiate that there is no single rewrite strategy optimal for all queries. In this context, guided rewriting via exploiting semantic features with QueryBandits can induce significant shifts in output behavior through forward-pass mechanisms, bypassing the need for retraining or gradient-based adaptation.
Authors: Wangsong Yin, Daliang Xu, Mengwei Xu, Gang Huang, Xuanzhe Liu
Abstract: On-device running Large Language Models (LLMs) is nowadays a critical enabler towards preserving user privacy. We observe that the attention operator falls back from the special-purpose NPU to the general-purpose CPU/GPU because of quantization sensitivity in state-of-the-art frameworks. This fallback results in a degraded user experience and increased complexity in system scheduling. To this end, this paper presents shadowAttn, a system-algorithm codesigned sparse attention module with minimal reliance on CPU/GPU by only sparsely calculating the attention on a tiny portion of tokens. The key idea is to hide the overhead of estimating the important tokens with a NPU-based pilot compute. Further, shadowAttn proposes insightful techniques such as NPU compute graph bucketing, head-wise NPU-CPU/GPU pipeline and per-head fine-grained sparsity ratio to achieve high accuracy and efficiency. shadowAttn delivers the best performance with highly limited CPU/GPU resource; it requires much less CPU/GPU resource to deliver on-par performance of SoTA frameworks.
Authors: Jonghyun Song, Youngjune Lee, Gyu-Hwung Cho, Ilhyeon Song, Saehun Kim, Yohan Jo
Abstract: Vision-Language Pretrained (VLP) models have achieved impressive performance on multimodal tasks, including text-image retrieval, based on dense representations. Meanwhile, Learned Sparse Retrieval (LSR) has gained traction in text-only settings due to its interpretability and efficiency with fast term-based lookup via inverted indexes. Inspired by these advantages, recent work has extended LSR to the multimodal domain. However, these methods often rely on computationally expensive contrastive pre-training, or distillation from a frozen dense model, which limits the potential for mutual enhancement. To address these limitations, we propose a simple yet effective framework that enables bi-directional learning between dense and sparse representations through Self-Knowledge Distillation. This bi-directional learning is achieved using an integrated similarity score-a weighted sum of dense and sparse similarities-which serves as a shared teacher signal for both representations. To ensure efficiency, we fine-tune the final layer of the dense encoder and the sparse projection head, enabling easy adaptation of any existing VLP model. Experiments on MSCOCO and Flickr30k demonstrate that our sparse retriever not only outperforms existing sparse baselines, but also achieves performance comparable to-or even surpassing-its dense counterparts, while retaining the benefits of sparse models.
Authors: Tianyao Shi, Yi Ding
Abstract: Large language models (LLMs) have demonstrated remarkable capabilities across diverse domains, but their heavy resource demands make quantization-reducing precision to lower-bit formats-critical for efficient serving. While many quantization methods exist, a systematic understanding of their performance, energy, and quality tradeoffs in realistic serving conditions remains a gap. In this work, we first develop a fully automated online characterization framework qMeter, and then conduct an in-depth characterization of 11 post-training LLM quantization methods across 4 model sizes (7B-70B) and two GPU architectures (A100, H100). We evaluate quantization at the application, workload, parallelism, and hardware levels under online serving conditions. Our study reveals highly task- and method-dependent tradeoffs, strong sensitivity to workload characteristics, and complex interactions with parallelism and GPU architecture. We further present three optimization case studies illustrating deployment challenges in capacity planning, energy-efficient scheduling, and multi-objective tuning. To the best of our knowledge, this is one of the first comprehensive application-, system-, and hardware-level characterization of LLM quantization from a joint performance, energy, and quality perspective.
Authors: Prabhant Singh, Yiping Li, Yasmina Al Khalil
Abstract: Fine-tuning pre-trained models has become a cornerstone of modern machine learning, allowing practitioners to achieve high performance with limited labeled data. In surgical video analysis, where expert annotations are especially time-consuming and costly, identifying the most suitable pre-trained model for a downstream task is both critical and challenging. Source-independent transferability estimation (SITE) offers a solution by predicting how well a model will fine-tune on target data using only its embeddings or outputs, without requiring full retraining. In this work, we formalize SITE for surgical phase recognition and provide the first comprehensive benchmark of three representative metrics, LogME, H-Score, and TransRate, on two diverse datasets (RAMIE and AutoLaparo). Our results show that LogME, particularly when aggregated by the minimum per-subset score, aligns most closely with fine-tuning accuracy; H-Score yields only weak predictive power; and TransRate often inverses true model rankings. Ablation studies show that when candidate models have similar performances, transferability estimates lose discriminative power, emphasizing the importance of maintaining model diversity or using additional validation. We conclude with practical guidelines for model selection and outline future directions toward domain-specific metrics, theoretical foundations, and interactive benchmarking tools.
Authors: Abdul Rehman Akbar, Usama Sajjad, Ziyu Su, Wencheng Li, Fei Xing, Jimmy Ruiz, Wei Chen, Muhammad Khalid Khan Niazi
Abstract: Despite surgical resection, ~70% of invasive lung adenocarcinoma (ILA) patients recur within five years, and current tools fail to identify those needing adjuvant therapy. To address this unmet clinical need, we introduce CellEcoNet, a novel spatially aware deep learning framework that models whole slide images (WSIs) through natural language analogy, defining a "language of pathology," where cells act as words, cellular neighborhoods become phrases, and tissue architecture forms sentences. CellEcoNet learns these context-dependent meanings automatically, capturing how subtle variations and spatial interactions derive recurrence risk. On a dataset of 456 H&E-stained WSIs, CellEcoNet achieved superior predictive performance (AUC:77.8% HR:9.54), outperforming IASLC grading system (AUC:71.4% HR:2.36), AJCC Stage (AUC:64.0% HR:1.17) and state-of-the-art computational methods (AUCs:62.2-67.4%). CellEcoNet demonstrated fairness and consistent performance across diverse demographic and clinical subgroups. Beyond prognosis, CellEcoNet marks a paradigm shift by decoding the tumor microenvironment's cellular "language" to reveal how subtle cell variations encode recurrence risk.
Authors: Liu Liu, Rui Dai
Abstract: Understanding the factors that shape students' mathematics performance is vital for designing effective educational policies. This study applies explainable artificial intelligence (XAI) techniques to PISA 2018 data to predict math achievement and identify key predictors across ten countries (67,329 students). We tested four models: Multiple Linear Regression (MLR), Random Forest (RF), CATBoost, and Artificial Neural Networks (ANN), using student, family, and school variables. Models were trained on 70% of the data (with 5-fold cross-validation) and tested on 30%, stratified by country. Performance was assessed with R^2 and Mean Absolute Error (MAE). To ensure interpretability, we used feature importance, SHAP values, and decision tree visualizations. Non-linear models, especially RF and ANN, outperformed MLR, with RF balancing accuracy and generalizability. Key predictors included socio-economic status, study time, teacher motivation, and students' attitudes toward mathematics, though their impact varied across countries. Visual diagnostics such as scatterplots of predicted vs actual scores showed RF and CATBoost aligned closely with actual performance. Findings highlight the non-linear and context-dependent nature of achievement and the value of XAI in educational research. This study uncovers cross-national patterns, informs equity-focused reforms, and supports the development of personalized learning strategies.
Authors: Xinwen Ding, Adam R Stinchcombe
Abstract: Elliptic interface problems arise in numerous scientific and engineering applications, modeling heterogeneous materials in which physical properties change discontinuously across interfaces. In this paper, we present \textit{Walk-on-Interfaces} (WoI), a grid-free Monte Carlo estimator for a class of Neumann elliptic interface problems with nonhomogeneous flux jump conditions. Our Monte Carlo estimators maintain consistent accuracy throughout the domain and, thus, do not suffer from the well-known close-to-source evaluation issue near the interfaces. We also presented a simple modification with reduced variance. Estimation of the gradient of the solution can be performed, with almost no additional cost, by simply computing the gradient of the Green's function in WoI. Taking a scientific machine learning approach, we use our estimators to provide training data for a deep neural network that outputs a continuous representation of the solution. This regularizes our solution estimates by removing the high-frequency Monte Carlo error. All of our estimators are highly parallelizable, have a $\mathcal{O}(1 / \sqrt{\mathcal{W}})$ convergence rate in the number of samples, and generalize naturally to higher dimensions. We solve problems with many interfaces that have irregular geometry and in up to dimension six. Numerical experiments demonstrate the effectiveness of the approach and to highlight its potential in solving problems motivated by real-world applications.
Authors: Yuancheng Wang, Dekun Chen, Xueyao Zhang, Junan Zhang, Jiaqi Li, Zhizheng Wu
Abstract: Speech tokenizers serve as foundational components for speech language models, yet current designs exhibit several limitations, including: 1) dependence on multi-layer residual vector quantization structures or high frame rates, 2) reliance on auxiliary pre-trained models for semantic distillation, and 3) requirements for complex two-stage training processes. In this work, we introduce the Text-aware Diffusion Transformer Speech Codec (TaDiCodec), a novel approach designed to overcome these challenges. TaDiCodec employs end-to-end optimization for quantization and reconstruction through a diffusion autoencoder, while integrating text guidance into the diffusion decoder to enhance reconstruction quality and achieve optimal compression. TaDiCodec achieves an extremely low frame rate of 6.25 Hz and a corresponding bitrate of 0.0875 kbps with a single-layer codebook for 24 kHz speech, while maintaining superior performance on critical speech generation evaluation metrics such as Word Error Rate (WER), speaker similarity (SIM), and speech quality (UTMOS). Notably, TaDiCodec employs a single-stage, end-to-end training paradigm, and obviating the need for auxiliary pre-trained models. We also validate the compatibility of TaDiCodec in language model based zero-shot text-to-speech with both autoregressive modeling and masked generative modeling, demonstrating its effectiveness and efficiency for speech language modeling, as well as a significantly small reconstruction-generation gap. We will open source our code and model checkpoints. Audio samples are are available at https:/tadicodec.github.io/. We release code and model checkpoints at https:/github.com/HeCheng0625/Diffusion-Speech-Tokenizer.
Authors: Hongtao Lin, Haoyu Chen, Jaewon Jang, Jiajing Xu
Abstract: User-to-item retrieval has been an active research area in recommendation system, and two tower models are widely adopted due to model simplicity and serving efficiency. In this work, we focus on a variant called \textit{conditional retrieval}, where we expect retrieved items to be relevant to a condition (e.g. topic). We propose a method that uses the same training data as standard two tower models but incorporates item-side information as conditions in query. This allows us to bootstrap new conditional retrieval use cases and encourages feature interactions between user and condition. Experiments show that our method can retrieve highly relevant items and outperforms standard two tower models with filters on engagement metrics. The proposed model is deployed to power a topic-based notification feed at Pinterest and led to +0.26\% weekly active users.
Authors: Marco S. Tayar, Lucas K. de Oliveira, Juliano D. Negri, Thiago H. Segreto, Ricardo V. Godoy, Marcelo Becker
Abstract: Inspecting confined industrial infrastructure, such as ventilation shafts, is a hazardous and inefficient task for humans. Unmanned Aerial Vehicles (UAVs) offer a promising alternative, but GPS-denied environments require robust control policies to prevent collisions. Deep Reinforcement Learning (DRL) has emerged as a powerful framework for developing such policies, and this paper provides a comparative study of two leading DRL algorithms for this task: the on-policy Proximal Policy Optimization (PPO) and the off-policy Soft Actor-Critic (SAC). The training was conducted with procedurally generated duct environments in Genesis simulation environment. A reward function was designed to guide a drone through a series of waypoints while applying a significant penalty for collisions. PPO learned a stable policy that completed all evaluation episodes without collision, producing smooth trajectories. By contrast, SAC consistently converged to a suboptimal behavior that traversed only the initial segments before failure. These results suggest that, in hazard-dense navigation, the training stability of on-policy methods can outweigh the nominal sample efficiency of off-policy algorithms. More broadly, the study provides evidence that procedurally generated, high-fidelity simulations are effective testbeds for developing and benchmarking robust navigation policies.
Authors: Xavier Gonzalez, Leo Kozachkov, David M. Zoltowski, Kenneth L. Clarkson, Scott W. Linderman
Abstract: The rise of parallel computing hardware has made it increasingly important to understand which nonlinear state space models can be efficiently parallelized. Recent advances like DEER (arXiv:2309.12252) or DeepPCR (arXiv:2309.16318) have shown that evaluating a state space model can be recast as solving a parallelizable optimization problem, and sometimes this approach can yield dramatic speed-ups in evaluation time. However, the factors that govern the difficulty of these optimization problems remain unclear, limiting the larger adoption of the technique. In this work, we establish a precise relationship between the dynamics of a nonlinear system and the conditioning of its corresponding optimization formulation. We show that the predictability of a system, defined as the degree to which small perturbations in state influence future behavior, impacts the number of optimization steps required for evaluation. In predictable systems, the state trajectory can be computed in $O((\log T)^2)$ time, where $T$ is the sequence length, a major improvement over the conventional sequential approach. In contrast, chaotic or unpredictable systems exhibit poor conditioning, with the consequence that parallel evaluation converges too slowly to be useful. Importantly, our theoretical analysis demonstrates that for predictable systems, the optimization problem is always well-conditioned, whereas for unpredictable systems, the conditioning degrades exponentially as a function of the sequence length. We validate our claims through extensive experiments, providing practical guidance on when nonlinear dynamical systems can be efficiently parallelized, and highlighting predictability as a key design principle for parallelizable models.
Authors: Sam Earle, Graham Todd, Yuchen Li, Ahmed Khalifa, Muhammad Umair Nasir, Zehua Jiang, Andrzej Banburski-Fahey, Julian Togelius
Abstract: We introduce PuzzleJAX, a GPU-accelerated puzzle game engine and description language designed to support rapid benchmarking of tree search, reinforcement learning, and LLM reasoning abilities. Unlike existing GPU-accelerated learning environments that provide hard-coded implementations of fixed sets of games, PuzzleJAX allows dynamic compilation of any game expressible in its domain-specific language (DSL). This DSL follows PuzzleScript, which is a popular and accessible online game engine for designing puzzle games. In this paper, we validate in PuzzleJAX several hundred of the thousands of games designed in PuzzleScript by both professional designers and casual creators since its release in 2013, thereby demonstrating PuzzleJAX's coverage of an expansive, expressive, and human-relevant space of tasks. By analyzing the performance of search, learning, and language models on these games, we show that PuzzleJAX can naturally express tasks that are both simple and intuitive to understand, yet often deeply challenging to master, requiring a combination of control, planning, and high-level insight.
Authors: Denis Tarasov, Alexander Nikulin, Ilya Zisman, Albina Klepach, Nikita Lyubaykin, Andrei Polubarov, Alexander Derevyagin, Vladislav Kurenkov
Abstract: Recent advances in Vision-Language-Action (VLA) models have established a two-component architecture, where a pre-trained Vision-Language Model (VLM) encodes visual observations and task descriptions, and an action decoder maps these representations to continuous actions. Diffusion models have been widely adopted as action decoders due to their ability to model complex, multimodal action distributions. However, they require multiple iterative denoising steps at inference time or downstream techniques to speed up sampling, limiting their practicality in real-world settings where high-frequency control is crucial. In this work, we present NinA (Normalizing Flows in Action), a fast and expressive alter- native to diffusion-based decoders for VLAs. NinA replaces the diffusion action decoder with a Normalizing Flow (NF) that enables one-shot sampling through an invertible transformation, significantly reducing inference time. We integrate NinA into the FLOWER VLA architecture and fine-tune on the LIBERO benchmark. Our experiments show that NinA matches the performance of its diffusion-based counterpart under the same training regime, while achieving substantially faster inference. These results suggest that NinA offers a promising path toward efficient, high-frequency VLA control without compromising performance.
Authors: Md Afif Al Mamun, Gias Uddin, Lan Xia, Longyu Zhang
Abstract: Pretrained Language Models or PLMs are transformer-based architectures that can be used in bug triaging tasks. PLMs can better capture token semantics than traditional Machine Learning (ML) models that rely on statistical features (e.g., TF-IDF, bag of words). However, PLMs may still attend to less relevant tokens in a bug report, which can impact their effectiveness. In addition, the model can be sub-optimal with its recommendations when the interaction history of developers around similar bugs is not taken into account. We designed TriagerX to address these limitations. First, to assess token semantics more reliably, we leverage a dual-transformer architecture. Unlike current state-of-the-art (SOTA) baselines that employ a single transformer architecture, TriagerX collects recommendations from two transformers with each offering recommendations via its last three layers. This setup generates a robust content-based ranking of candidate developers. TriagerX then refines this ranking by employing a novel interaction-based ranking methodology, which considers developers' historical interactions with similar fixed bugs. Across five datasets, TriagerX surpasses all nine transformer-based methods, including SOTA baselines, often improving Top-1 and Top-3 developer recommendation accuracy by over 10%. We worked with our large industry partner to successfully deploy TriagerX in their development environment. The partner required both developer and component recommendations, with components acting as proxies for team assignments-particularly useful in cases of developer turnover or team changes. We trained TriagerX on the partner's dataset for both tasks, and it outperformed SOTA baselines by up to 10% for component recommendations and 54% for developer recommendations.
Authors: Masato Shibukawa, Naoya Ozaki, Maximilien Berthet
Abstract: Particle methods play an important role in computational fluid dynamics, but they are among the most difficult to implement and solve. The most common method is smoothed particle hydrodynamics, which is suitable for problem settings that involve large deformations, such as tsunamis and dam breaking. However, the calculation can become unstable depending on the distribution of particles. In contrast, the neural particle method has high computational stability for various particle distributions is a machine learning method that approximates velocity and pressure in a spatial domain using neural networks. The neural particle method has been extended to viscous flows, but until now it has been limited to incompressible flows. In this paper, we propose the compressible neural particle method, which is a new feed-forward neural network-based method that extends the original neural particle method to model compressible viscous fluid flows. The proposed method uses neural networks to calculate the velocity and pressure of fluid particles at the next time step, and the Tait equation to calculate the density to handle the compressibility. The loss function is composed of the governing equations of compressible flow and the boundary conditions, which are free surface and solid boundary conditions. We demonstrate that the proposed method can accurately solve the compressible viscous fluid flow, a problem that was difficult to solve with the smoothed particle hydrodynamics method, by applying it to a dam breaking problem.
Authors: Bin Pan, Shiyu Shen, Zongbin Wang, Zhenwei Shi, Xia Xu
Abstract: Domain generalization seeks to develop models trained on a limited set of source domains that are capable of generalizing effectively to unseen target domains. While the predominant approach leverages large-scale pre-trained vision models as initialization, recent studies have highlighted that full fine-tuning can compromise the intrinsic generalization capabilities of these models. To address this limitation, parameter-efficient adaptation strategies have emerged, wherein only a subset of model parameters is selectively fine-tuned, thereby balancing task adaptation with the preservation of generalization. Motivated by this paradigm, we introduce Joint Parameter Selection (JPS), a novel method that restricts updates to a small, sparse subset of parameters, thereby retaining and harnessing the generalization strength of pre-trained models. Theoretically, we establish a generalization error bound that explicitly accounts for the sparsity of parameter updates, thereby providing a principled justification for selective fine-tuning. Practically, we design a selection mechanism employing dual operators to identify and update parameters exhibiting consistent and significant gradients across all source domains. Extensive benchmark experiments demonstrate that JPS achieves superior performance compared to state-of-the-art domain generalization methods, substantiating both the efficiency and efficacy of the proposed approach.
Authors: Soumyasundar Pal, Liheng Ma, Amine Natik, Yingxue Zhang, Mark Coates
Abstract: Accurate modelling and quantification of predictive uncertainty is crucial in deep learning since it allows a model to make safer decisions when the data is ambiguous and facilitates the users' understanding of the model's confidence in its predictions. Along with the tremendously increasing research focus on \emph{graph neural networks} (GNNs) in recent years, there have been numerous techniques which strive to capture the uncertainty in their predictions. However, most of these approaches are specifically designed for node or link-level tasks and cannot be directly applied to graph-level learning problems. In this paper, we propose a novel variational modelling framework for the \emph{posterior predictive distribution}~(PPD) to obtain uncertainty-aware prediction in graph-level learning tasks. Based on a graph-level embedding derived from one of the existing GNNs, our framework can learn the PPD in a data-adaptive fashion. Experimental results on several benchmark datasets exhibit the effectiveness of our approach.
Authors: Jason R Brown, Lennie Wells, Edward James Young, Sergio Bacallado
Abstract: Proximal Policy Optimisation (PPO) is an established and effective policy gradient algorithm used for Language Model Reinforcement Learning from Human Feedback (LM-RLHF). PPO performs well empirically but has a heuristic motivation and handles the KL-divergence constraint used in LM-RLHF in an ad-hoc manner. In this paper, we develop a a new action-value RL method for the LM-RLHF setting, KL-regularised Q-Learning (KLQ). We then show that our method is equivalent to a version of PPO in a certain specific sense, despite its very different motivation. Finally, we benchmark KLQ on two key language generation tasks -- summarisation and single-turn dialogue. We demonstrate that KLQ performs on-par with PPO at optimising the LM-RLHF objective, and achieves a consistently higher win-rate against PPO on LLM-as-a-judge evaluations.
Authors: Yan Cathy Hua, Paul Denny, J\"org Wicker, Katerina Taskova
Abstract: Every year, most educational institutions seek and receive an enormous volume of text feedback from students on courses, teaching, and overall experience. Yet, turning this raw feedback into useful insights is far from straightforward. It has been a long-standing challenge to adopt automatic opinion mining solutions for such education review text data due to the content complexity and low-granularity reporting requirements. Aspect-based Sentiment Analysis (ABSA) offers a promising solution with its rich, sub-sentence-level opinion mining capabilities. However, existing ABSA research and resources are very heavily focused on the commercial domain. In education, they are scarce and hard to develop due to limited public datasets and strict data protection. A high-quality, annotated dataset is urgently needed to advance research in this under-resourced area. In this work, we present EduRABSA (Education Review ABSA), the first public, annotated ABSA education review dataset that covers three review subject types (course, teaching staff, university) in the English language and all main ABSA tasks, including the under-explored implicit aspect and implicit opinion extraction. We also share ASQE-DPT (Data Processing Tool), an offline, lightweight, installation-free manual data annotation tool that generates labelled datasets for comprehensive ABSA tasks from a single-task annotation. Together, these resources contribute to the ABSA community and education domain by removing the dataset barrier, supporting research transparency and reproducibility, and enabling the creation and sharing of further resources. The dataset, annotation tool, and scripts and statistics for dataset processing and sampling are available at https://github.com/yhua219/edurabsa_dataset_and_annotation_tool.
URLs: https://github.com/yhua219/edurabsa_dataset_and_annotation_tool.
Authors: Seamus Somerstep, Ya'acov Ritov, Mikhail Yurochkin, Subha Maity, Yuekai Sun
Abstract: Standard techniques for aligning large language models (LLMs) utilize human-produced data, which could limit the capability of any aligned LLM to human level. Label refinement and weak training have emerged as promising strategies to address this superalignment problem. In this work, we adopt probabilistic assumptions commonly used to study label refinement and analyze whether refinement can be outperformed by alternative approaches, including computationally intractable oracle methods. We show that both weak training and label refinement suffer from irreducible error, leaving a performance gap between label refinement and the oracle. These results motivate future research into developing alternative methods for weak to strong generalization that synthesize the practicality of label refinement or weak training and the optimality of the oracle procedure.
Authors: Melissa Kazemi Rad, Alberto Purpura, Himanshu Kumar, Emily Chen, Mohammad Shahed Sorower
Abstract: We address the problem of data scarcity in harmful text classification for guardrailing applications and introduce GRAID (Geometric and Reflective AI-Driven Data Augmentation), a novel pipeline that leverages Large Language Models (LLMs) for dataset augmentation. GRAID consists of two stages: (i) generation of geometrically controlled examples using a constrained LLM, and (ii) augmentation through a multi-agentic reflective process that promotes stylistic diversity and uncovers edge cases. This combination enables both reliable coverage of the input space and nuanced exploration of harmful content. Using two benchmark data sets, we demonstrate that augmenting a harmful text classification dataset with GRAID leads to significant improvements in downstream guardrail model performance.
Authors: Luben M. C. Cabezas, Vagner S. Santos, Thiago R. Ramos, Pedro L. C. Rodrigues, Rafael Izbicki
Abstract: Current experimental scientists have been increasingly relying on simulation-based inference (SBI) to invert complex non-linear models with intractable likelihoods. However, posterior approximations obtained with SBI are often miscalibrated, causing credible regions to undercover true parameters. We develop $\texttt{CP4SBI}$, a model-agnostic conformal calibration framework that constructs credible sets with local Bayesian coverage. Our two proposed variants, namely local calibration via regression trees and CDF-based calibration, enable finite-sample local coverage guarantees for any scoring function, including HPD, symmetric, and quantile-based regions. Experiments on widely used SBI benchmarks demonstrate that our approach improves the quality of uncertainty quantification for neural posterior estimators using both normalizing flows and score-diffusion modeling.
Authors: Yushi Lin, Peng Yang
Abstract: Financial markets are critical to global economic stability, yet trade-based manipulation (TBM) often undermines their fairness. Spoofing, a particularly deceptive TBM strategy, exhibits multilevel anomaly patterns that have not been adequately modeled. These patterns are usually concealed within the rich, hierarchical information of the Limit Order Book (LOB), which is challenging to leverage due to high dimensionality and noise. To address this, we propose a representation learning framework combining a cascaded LOB representation pipeline with supervised contrastive learning. Extensive experiments demonstrate that our framework consistently improves detection performance across diverse models, with Transformer-based architectures achieving state-of-the-art results. In addition, we conduct systematic analyses and ablation studies to investigate multilevel anomalies and the contributions of key components, offering broader insights into representation learning and anomaly detection for complex sequential data. Our code will be released later at this URL.
Authors: Yue-Jane Liu, Malinda Lu, Matthew K. Nock, Yaniv Yacoby
Abstract: Many modern probabilistic models rely on SDEs, but their adoption is hampered by instability, poor inductive bias outside bounded domains, and reliance on restrictive dynamics or training tricks. While recent work constrains SDEs to compact spaces using reflected dynamics, these approaches lack continuous dynamics and efficient high-order solvers, limiting interpretability and applicability. We propose a novel class of neural SDEs on compact polyhedral spaces with continuous dynamics, amenable to higher-order solvers, and with favorable inductive bias.
Authors: Yahya Badran, Christine Preisach
Abstract: Knowledge Tracing (KT) aims to predict a student's future performance based on their sequence of interactions with learning content. Many KT models rely on knowledge concepts (KCs), which represent the skills required for each item. However, some of these models are vulnerable to label leakage, in which input data inadvertently reveal the correct answer, particularly in datasets with multiple KCs per question. We propose a straightforward yet effective solution to prevent label leakage by masking ground-truth labels during input embedding construction in cases susceptible to leakage. To accomplish this, we introduce a dedicated MASK label, inspired by masked language modeling (e.g., BERT), to replace ground-truth labels. In addition, we introduce Recency Encoding, which encodes the step-wise distance between the current item and its most recent previous occurrence. This distance is important for modeling learning dynamics such as forgetting, which is a fundamental aspect of human learning, yet it is often overlooked in existing models. Recency Encoding demonstrates improved performance over traditional positional encodings on multiple KT benchmarks. We show that incorporating our embeddings into KT models like DKT, DKT+, AKT, and SAKT consistently improves prediction accuracy across multiple benchmarks. The approach is both efficient and widely applicable.
Authors: Shifat E. Arman, Hasan Muhammad Abdullah, Syed Nazmus Sakib, RM Saiem, Shamima Nasrin Asha, Md Mehedi Hasan, Shahrear Bin Amin, S M Mahin Abrar
Abstract: Despite progress in AI-based plant diagnostics, sugarcane farmers in low-resource regions remain vulnerable to leaf diseases due to the lack of scalable, efficient, and interpretable tools. Many deep learning models fail to generalize under real-world conditions and require substantial computational resources, limiting their use in resource-constrained regions. In this paper, we present SugarcaneLD-BD, a curated dataset for sugarcane leaf-disease classification; SugarcaneShuffleNet, an optimized lightweight model for rapid on-device diagnosis; and SugarcaneAI, a Progressive Web Application for field deployment. SugarcaneLD-BD contains 638 curated images across five classes, including four major sugarcane diseases, collected in Bangladesh under diverse field conditions and verified by expert pathologists. To enhance diversity, we combined SugarcaneLD-BD with two additional datasets, yielding a larger and more representative corpus. Our optimized model, SugarcaneShuffleNet, offers the best trade-off between speed and accuracy for real-time, on-device diagnosis. This 9.26 MB model achieved 98.02% accuracy, an F1-score of 0.98, and an average inference time of 4.14 ms per image. For comparison, we fine-tuned five other lightweight convolutional neural networks: MnasNet, EdgeNeXt, EfficientNet-Lite, MobileNet, and SqueezeNet via transfer learning and Bayesian optimization. MnasNet and EdgeNeXt achieved comparable accuracy to SugarcaneShuffleNet, but required significantly more parameters, memory, and computation, limiting their suitability for low-resource deployment. We integrate SugarcaneShuffleNet into SugarcaneAI, delivering Grad-CAM-based explanations in the field. Together, these contributions offer a diverse benchmark, efficient models for low-resource environments, and a practical tool for sugarcane disease classification. It spans varied lighting, backgrounds and devices used on-farm
Authors: Syed Nazmus Sakib, Nafiul Haque, Mohammad Zabed Hossain, Shifat E. Arman
Abstract: PlantVillageVQA is a large-scale visual question answering (VQA) dataset derived from the widely used PlantVillage image corpus. It was designed to advance the development and evaluation of vision-language models for agricultural decision-making and analysis. The PlantVillageVQA dataset comprises 193,609 high-quality question-answer (QA) pairs grounded over 55,448 images spanning 14 crop species and 38 disease conditions. Questions are organised into 3 levels of cognitive complexity and 9 distinct categories. Each question category was phrased manually following expert guidance and generated via an automated two-stage pipeline: (1) template-based QA synthesis from image metadata and (2) multi-stage linguistic re-engineering. The dataset was iteratively reviewed by domain experts for scientific accuracy and relevancy. The final dataset was evaluated using three state-of-the-art models for quality assessment. Our objective remains to provide a publicly available, standardised and expert-verified database to enhance diagnostic accuracy for plant disease identifications and advance scientific research in the agricultural domain. Our dataset will be open-sourced at https://huggingface.co/datasets/SyedNazmusSakib/PlantVillageVQA.
URLs: https://huggingface.co/datasets/SyedNazmusSakib/PlantVillageVQA.
Authors: Matej Neumann, Yunan Yang
Abstract: Full-waveform inversion (FWI) is a powerful technique for reconstructing high-resolution material parameters from seismic or ultrasound data. The conventional least-squares (\(L^{2}\)) misfit suffers from pronounced non-convexity that leads to \emph{cycle skipping}. Optimal-transport misfits, such as the Wasserstein distance, alleviate this issue; however, their use requires artificially converting the wavefields into probability measures, a preprocessing step that can modify critical amplitude and phase information of time-dependent wave data. We propose the \emph{HV metric}, a transport-based distance that acts naturally on signed signals, as an alternative metric for the \(L^{2}\) and Wasserstein objectives in time-domain FWI. After reviewing the metric's definition and its relationship to optimal transport, we derive closed-form expressions for the Fr\'echet derivative and Hessian of the map \(f \mapsto d_{\text{HV}}^2(f,g)\), enabling efficient adjoint-state implementations. A spectral analysis of the Hessian shows that, by tuning the hyperparameters \((\kappa,\lambda,\epsilon)\), the HV misfit seamlessly interpolates between \(L^{2}\), \(H^{-1}\), and \(H^{-2}\) norms, offering a tunable trade-off between the local point-wise matching and the global transport-based matching. Synthetic experiments on the Marmousi and BP benchmark models demonstrate that the HV metric-based objective function yields faster convergence and superior tolerance to poor initial models compared to both \(L^{2}\) and Wasserstein misfits. These results demonstrate the HV metric as a robust, geometry-preserving alternative for large-scale waveform inversion.
Authors: Viacheslav Yusupov, Danil Maksimov, Ameliia Alaeva, Tatiana Zaitceva, Antipina Anna, Anna Vasileva, Chenlin Liu, Rayuth Chheng, Danil Sazanakov, Andrey Chetvergov, Alina Ermilova, Egor Shvetsov
Abstract: This paper investigates token homogenization - the convergence of token representations toward uniformity across transformer layers and its relationship to positional bias in large language models. We empirically examine whether homogenization occurs and how positional bias amplifies this effect. Through layer-wise similarity analysis and controlled experiments, we demonstrate that tokens systematically lose distinctiveness during processing, particularly when biased toward extremal positions. Our findings confirm both the existence of homogenization and its dependence on positional attention mechanisms.
Authors: Carlos Soto
Abstract: Differential privacy (DP) has recently emerged as a definition of privacy to release private estimates. DP calibrates noise to be on the order of an individuals contribution. Due to the this calibration a private estimate obscures any individual while preserving the utility of the estimate. Since the original definition, many alternate definitions have been proposed. These alternates have been proposed for various reasons including improvements on composition results, relaxations, and formalizations. Nevertheless, thus far nearly all definitions of privacy have used a divergence of densities as the basis of the definition. In this paper we take an information geometry perspective towards differential privacy. Specifically, rather than define privacy via a divergence, we define privacy via the Rao distance. We show that our proposed definition of privacy shares the interpretation of previous definitions of privacy while improving on sequential composition.
Authors: Jianqing Fan, Soham Jana, Sanjeev Kulkarni, Qishuo Yin
Abstract: We investigate the problem of estimating the average treatment effect (ATE) under a very general setup where the covariates can be high-dimensional, highly correlated, and can have sparse nonlinear effects on the propensity and outcome models. We present the use of a Double Deep Learning strategy for estimation, which involves combining recently developed factor-augmented deep learning-based estimators, FAST-NN, for both the response functions and propensity scores to achieve our goal. By using FAST-NN, our method can select variables that contribute to propensity and outcome models in a completely nonparametric and algorithmic manner and adaptively learn low-dimensional function structures through neural networks. Our proposed novel estimator, FIDDLE (Factor Informed Double Deep Learning Estimator), estimates ATE based on the framework of augmented inverse propensity weighting AIPW with the FAST-NN-based response and propensity estimates. FIDDLE consistently estimates ATE even under model misspecification and is flexible to also allow for low-dimensional covariates. Our method achieves semiparametric efficiency under a very flexible family of propensity and outcome models. We present extensive numerical studies on synthetic and real datasets to support our theoretical guarantees and establish the advantages of our methods over other traditional choices, especially when the data dimension is large.
Authors: Arya Honarpisheh, Mario Sznaier
Abstract: This paper proposes a frequency-domain system identification method for learning low-order systems. The identification problem is formulated as the minimization of the l2 norm between the identified and measured frequency responses, with the nuclear norm of the Loewner matrix serving as a regularization term. This formulation results in an optimization problem that can be efficiently solved using standard convex optimization techniques. We derive an upper bound on the sampled-frequency complexity of the identification process and subsequently extend this bound to characterize the identification error over all frequencies. A detailed analysis of the sample complexity is provided, along with a thorough interpretation of its terms and dependencies. Finally, the efficacy of the proposed method is demonstrated through an example, along with numerical simulations validating the growth rate of the sample complexity bound.
Authors: Mohammed Alsobay, David G. Rand, Duncan J. Watts, Abdullah Almaatouq
Abstract: Punishment as a mechanism for promoting cooperation has been studied extensively for more than two decades, but its effectiveness remains a matter of dispute. Here, we examine how punishment's impact varies across cooperative settings through a large-scale integrative experiment. We vary 14 parameters that characterize public goods games, sampling 360 experimental conditions and collecting 147,618 decisions from 7,100 participants. Our results reveal striking heterogeneity in punishment effectiveness: while punishment consistently increases contributions, its impact on payoffs (i.e., efficiency) ranges from dramatically enhancing welfare (up to 43% improvement) to severely undermining it (up to 44% reduction) depending on the cooperative context. To characterize these patterns, we developed models that outperformed human forecasters (laypeople and domain experts) in predicting punishment outcomes in new experiments. Communication emerged as the most predictive feature, followed by contribution framing (opt-out vs. opt-in), contribution type (variable vs. all-or-nothing), game length (number of rounds), peer outcome visibility (whether participants can see others' earnings), and the availability of a reward mechanism. Interestingly, however, most of these features interact to influence punishment effectiveness rather than operating independently. For example, the extent to which longer games increase the effectiveness of punishment depends on whether groups can communicate. Together, our results refocus the debate over punishment from whether or not it "works" to the specific conditions under which it does and does not work. More broadly, our study demonstrates how integrative experiments can be combined with machine learning to uncover generalizable patterns, potentially involving interactions between multiple features, and help generate novel explanations in complex social phenomena.
Authors: Tobias Wegel, Geelon So, Junhyung Park, Fanny Yang
Abstract: In multi-objective learning (MOL), several possibly competing prediction tasks must be solved jointly by a single model. Achieving good trade-offs may require a model class $\mathcal{G}$ with larger capacity than what is necessary for solving the individual tasks. This, in turn, increases the statistical cost, as reflected in known MOL bounds that depend on the complexity of $\mathcal{G}$. We show that this cost is unavoidable for some losses, even in an idealized semi-supervised setting, where the learner has access to the Bayes-optimal solutions for the individual tasks as well as the marginal distributions over the covariates. On the other hand, for objectives defined with Bregman losses, we prove that the complexity of $\mathcal{G}$ may come into play only in terms of unlabeled data. Concretely, we establish sample complexity upper bounds, showing precisely when and how unlabeled data can significantly alleviate the need for labeled data. These rates are achieved by a simple, semi-supervised algorithm via pseudo-labeling.
Authors: Yajat Yadav, Varun Bharadwaj, Jathin Korrapati, Tanish Baranwal
Abstract: We introduce VROOM, a system for reconstructing 3D models of Formula 1 circuits using only onboard camera footage from racecars. Leveraging video data from the 2023 Monaco Grand Prix, we address video challenges such as high-speed motion and sharp cuts in camera frames. Our pipeline analyzes different methods such as DROID-SLAM, AnyCam, and Monst3r and combines preprocessing techniques such as different methods of masking, temporal chunking, and resolution scaling to account for dynamic motion and computational constraints. We show that Vroom is able to partially recover track and vehicle trajectories in complex environments. These findings indicate the feasibility of using onboard video for scalable 4D reconstruction in real-world settings. The project page can be found at https://varun-bharadwaj.github.io/vroom, and our code is available at https://github.com/yajatyadav/vroom.
URLs: https://varun-bharadwaj.github.io/vroom,, https://github.com/yajatyadav/vroom.
Authors: Nilay Pande, Sahiti Yerramilli, Jayant Sravan Tamarapalli, Rynaa Grover
Abstract: A key frontier for Multimodal Large Language Models (MLLMs) is the ability to perform deep mathematical and spatial reasoning directly from images, moving beyond their established success in semantic description. Mathematical surface plots provide a rigorous testbed for this capability, as they isolate the task of reasoning from the semantic noise common in natural images. To measure progress on this frontier, we introduce MaRVL-QA (Mathematical Reasoning over Visual Landscapes), a new benchmark designed to quantitatively evaluate these core reasoning skills. The benchmark comprises two novel tasks: Topological Counting, identifying and enumerating features like local maxima; and Transformation Recognition, recognizing applied geometric transformations. Generated from a curated library of functions with rigorous ambiguity filtering, our evaluation on MaRVL-QA reveals that even state-of-the-art MLLMs struggle significantly, often resorting to superficial heuristics instead of robust spatial reasoning. MaRVL-QA provides a challenging new tool for the research community to measure progress, expose model limitations, and guide the development of MLLMs with more profound reasoning abilities.
Authors: Md. Maruf, Md. Mahbubul Haque, Bishowjit Paul
Abstract: Acute lymphoblastic leukemia (ALL) is a prevalent hematological malignancy in both pediatric and adult populations. Early and accurate detection with precise subtyping is essential for guiding therapy. Conventional workflows are complex, time-consuming, and prone to human error. We present a deep learning framework for automated ALL diagnosis from bone marrow smear images. The method combines a robust preprocessing pipeline with convolutional neural networks (CNNs) to standardize image quality and improve inference efficiency. As a key design, we insert a multi-head self-attention (MHSA) block into a VGG19 backbone to model long-range dependencies and contextual relationships among cellular features. To mitigate class imbalance, we train with Focal Loss. Across evaluated architectures, the enhanced VGG19+MHSA trained with Focal Loss achieves 99.25% accuracy, surpassing a strong ResNet101 baseline (98.62%). These results indicate that attention-augmented CNNs, coupled with targeted loss optimization and preprocessing, yield more discriminative representations of leukemic cell morphology. Our approach offers a highly accurate and computationally efficient tool for automated ALL recognition and subtyping, with potential to accelerate diagnostic workflows and support reliable decision-making in clinical settings.
Authors: Bingyang Wu, Zili Zhang, Yinmin Zhong, Guanzhe Huang, Yibo Zhu, Xuanzhe Liu, Xin Jin
Abstract: Prefix caching is crucial to accelerate multi-turn interactions and requests with shared prefixes. At the cluster level, existing prefix caching systems are tightly coupled with request scheduling to optimize cache efficiency and computation performance together, leading to load imbalance, data redundancy, and memory fragmentation of caching systems across instances. To address these issues, memory pooling is promising to shield the scheduler from the underlying cache management so that it can focus on the computation optimization. However, because existing prefix caching systems only transfer increasingly longer prefix caches between instances, they cannot achieve low-latency memory pooling. To address these problems, we propose a unified segment-level prefix cache pool, TokenLake. It uses a declarative cache interface to expose requests' query tensors, prefix caches, and cache-aware operations to TokenLake for efficient pooling. Powered by this abstraction, TokenLake can manage prefix cache at the segment level with a heavy-hitter-aware load balancing algorithm to achieve better cache load balance, deduplication, and defragmentation. TokenLake also transparently minimizes the communication volume of query tensors and new caches. Based on TokenLake, the scheduler can schedule requests elastically by using existing techniques without considering prefix cache management. Evaluations on real-world workloads show that TokenLake can improve throughput by up to 2.6$\times$ and 2.0$\times$ and boost hit rate by 2.0$\times$ and 2.1$\times$, compared to state-of-the-art cache-aware routing and cache-centric PD-disaggregation solutions, respectively.
Authors: Sopam Dasgupta, Sadaf MD Halim, Joaqu\'in Arias, Elmer Salazar, Gopal Gupta
Abstract: Machine learning models increasingly influence decisions in high-stakes settings such as finance, law and hiring, driving the need for transparent, interpretable outcomes. However, while explainable approaches can help understand the decisions being made, they may inadvertently reveal the underlying proprietary algorithm: an undesirable outcome for many practitioners. Consequently, it is crucial to balance meaningful transparency with a form of recourse that clarifies why a decision was made and offers actionable steps following which a favorable outcome can be obtained. Counterfactual explanations offer a powerful mechanism to address this need by showing how specific input changes lead to a more favorable prediction. We propose Model-Agnostic Causally Constrained Counterfactual Generation (MC3G), a novel framework that tackles limitations in the existing counterfactual methods. First, MC3G is model-agnostic: it approximates any black-box model using an explainable rule-based surrogate model. Second, this surrogate is used to generate counterfactuals that produce a favourable outcome for the original underlying black box model. Third, MC3G refines cost computation by excluding the ``effort" associated with feature changes that occur automatically due to causal dependencies. By focusing only on user-initiated changes, MC3G provides a more realistic and fair representation of the effort needed to achieve a favourable outcome. We show that MC3G delivers more interpretable and actionable counterfactual recommendations compared to existing techniques all while having a lower cost. Our findings highlight MC3G's potential to enhance transparency, accountability, and practical utility in decision-making processes that incorporate machine-learning approaches.
Authors: Junan Zhang, Xueyao Zhang, Jing Yang, Yuancheng Wang, Fan Fan, Zhizheng Wu
Abstract: Recent generative models have significantly advanced speech restoration tasks, yet their training objectives often misalign with human perceptual preferences, resulting in suboptimal quality. While post-training alignment has proven effective in other generative domains like text and image generation, its application to generative speech restoration remains largely under-explored. This work investigates the challenges of applying preference-based post-training to this task, focusing on how to define a robust preference signal and curate high-quality data to avoid reward hacking. To address these challenges, we propose a multi-metric preference alignment strategy. We construct a new dataset, GenSR-Pref, comprising 80K preference pairs, where each chosen sample is unanimously favored by a complementary suite of metrics covering perceptual quality, signal fidelity, content consistency, and timbre preservation. This principled approach ensures a holistic preference signal. Applying Direct Preference Optimization (DPO) with our dataset, we observe consistent and significant performance gains across three diverse generative paradigms: autoregressive models (AR), masked generative models (MGM), and flow-matching models (FM) on various restoration benchmarks, in both objective and subjective evaluations. Ablation studies confirm the superiority of our multi-metric strategy over single-metric approaches in mitigating reward hacking. Furthermore, we demonstrate that our aligned models can serve as powerful ''data annotators'', generating high-quality pseudo-labels to serve as a supervision signal for traditional discriminative models in data-scarce scenarios like singing voice restoration. Demo Page:https://gensr-pref.github.io
Authors: Yunyong Ko, Da Eun Lee, Song Kyung Yu, Sang-Wook Kim
Abstract: Real-world networks have high-order relationships among objects and they evolve over time. To capture such dynamics, many works have been studied in a range of fields. Via an in-depth preliminary analysis, we observe two important characteristics of high-order dynamics in real-world networks: high-order relations tend to (O1) have a structural and temporal influence on other relations in a short term and (O2) periodically re-appear in a long term. In this paper, we propose LINCOLN, a method for Learning hIgh-order dyNamiCs Of reaL-world Networks, that employs (1) bi-interactional hyperedge encoding for short-term patterns, (2) periodic time injection and (3) intermediate node representation for long-term patterns. Via extensive experiments, we show that LINCOLN outperforms nine state-of-the-art methods in the dynamic hyperedge prediction task.
Authors: Sankalp Pandey, Xuan Bac Nguyen, Nicholas Borys, Hugh Churchill, Khoa Luu
Abstract: Identifying quantum flakes is crucial for scalable quantum hardware; however, automated layer classification from optical microscopy remains challenging due to substantial appearance shifts across different materials. In this paper, we propose a new Continual-Learning Framework for Flake Layer Classification (CLIFF). To our knowledge, this is the first systematic study of continual learning in the domain of two-dimensional (2D) materials. Our method enables the model to differentiate between materials and their physical and optical properties by freezing a backbone and base head trained on a reference material. For each new material, it learns a material-specific prompt, embedding, and a delta head. A prompt pool and a cosine-similarity gate modulate features and compute material-specific corrections. Additionally, we incorporate memory replay with knowledge distillation. CLIFF achieves competitive accuracy with significantly lower forgetting than naive fine-tuning and a prompt-based baseline.
Authors: Breenda Das, Lennart Purucker, Timur Carstensen, Frank Hutter
Abstract: Foundation models like SAM (Segment Anything Model) exhibit strong zero-shot image segmentation performance, but often fall short on domain-specific tasks. Fine-tuning these models typically requires significant manual effort and domain expertise. In this work, we introduce QTT-SEG, a meta-learning-driven approach for automating and accelerating the fine-tuning of SAM for image segmentation. Built on the Quick-Tune hyperparameter optimization framework, QTT-SEG predicts high-performing configurations using meta-learned cost and performance models, efficiently navigating a search space of over 200 million possibilities. We evaluate QTT-SEG on eight binary and five multiclass segmentation datasets under tight time constraints. Our results show that QTT-SEG consistently improves upon SAM's zero-shot performance and surpasses AutoGluon Multimodal, a strong AutoML baseline, on most binary tasks within three minutes. On multiclass datasets, QTT-SEG delivers consistent gains as well. These findings highlight the promise of meta-learning in automating model adaptation for specialized segmentation tasks. Code available at: https://github.com/ds-brx/QTT-SEG/
Authors: Omid Ghahroodi, Arshia Hemmat, Marzia Nouri, Seyed Mohammad Hadi Hosseini, Doratossadat Dastgheib, Mohammad Vali Sanian, Alireza Sahebi, Reihaneh Zohrabi, Mohammad Hossein Rohban, Ehsaneddin Asgari, Mahdieh Soleymani Baghshah
Abstract: Recent advancements in large vision-language models (VLMs) have primarily focused on English, with limited attention given to other languages. To address this gap, we introduce MEENA (also known as PersianMMMU), the first dataset designed to evaluate Persian VLMs across scientific, reasoning, and human-level understanding tasks. Our dataset comprises approximately 7,500 Persian and 3,000 English questions, covering a wide range of topics such as reasoning, mathematics, physics, diagrams, charts, and Persian art and literature. Key features of MEENA include: (1) diverse subject coverage spanning various educational levels, from primary to upper secondary school, (2) rich metadata, including difficulty levels and descriptive answers, (3) original Persian data that preserves cultural nuances, (4) a bilingual structure to assess cross-linguistic performance, and (5) a series of diverse experiments assessing various capabilities, including overall performance, the model's ability to attend to images, and its tendency to generate hallucinations. We hope this benchmark contributes to enhancing VLM capabilities beyond English.
Authors: Somraj Gautam, Abhirama Subramanyam Penamakuri, Abhishek Bhandari, Gaurav Harit
Abstract: We introduce MMCRICBENCH-3K, a benchmark for Visual Question Answering (VQA) on cricket scorecards, designed to evaluate large vision-language models (LVLMs) on complex numerical and cross-lingual reasoning over semi-structured tabular images. MMCRICBENCH-3K comprises 1,463 synthetically generated scorecard images from ODI, T20, and Test formats, accompanied by 1,500 English QA pairs. It includes two subsets: MMCRICBENCH-E-1.5K, featuring English scorecards, and MMCRICBENCH-H-1.5K, containing visually similar Hindi scorecards, with all questions and answers kept in English to enable controlled cross-script evaluation. The task demands reasoning over structured numerical data, multi-image context, and implicit domain knowledge. Empirical results show that even state-of-the-art LVLMs, such as GPT-4o and Qwen2.5VL, struggle on the English subset despite it being their primary training language and exhibit a further drop in performance on the Hindi subset. This reveals key limitations in structure-aware visual text understanding, numerical reasoning, and cross-lingual generalization. The dataset is publicly available via Hugging Face at https://huggingface.co/datasets/DIALab/MMCricBench, to promote LVLM research in this direction.
Authors: Haojie Zhang
Abstract: LoRA-based large model parameter-efficient fine-tuning (PEFT) methods use low-rank de- composition to approximate updates to model parameters. However, compared to full- parameter fine-tuning, low-rank updates often lead to a performance gap in downstream tasks. To address this, we introduce DropLoRA, a novel pruning-based approach that focuses on pruning the rank dimension. Unlike conven- tional methods that attempt to overcome the low-rank bottleneck, DropLoRA innovatively integrates a pruning module between the two low-rank matrices in LoRA to simulate dy- namic subspace learning. This dynamic low- rank subspace learning allows DropLoRA to overcome the limitations of traditional LoRA, which operates within a static subspace. By continuously adapting the learning subspace, DropLoRA significantly boosts performance without incurring additional training or infer- ence costs. Our experimental results demon- strate that DropLoRA consistently outperforms LoRA in fine-tuning the LLaMA series across a wide range of large language model gener- ation tasks, including commonsense reason- ing, mathematical reasoning, code generation, and instruction-following. Our code is avail- able at https://github.com/TayeeChang/DropLoRA.
Authors: Rajrupa Chattaraj, Sridhar Chimalakonda, Vibhu Saujanya Sharma, Vikrant Kaulgud
Abstract: The utilization of Machine Learning (ML) in contemporary software systems is extensive and continually expanding. However, its usage is energy-intensive, contributing to increased carbon emissions and demanding significant resources. While numerous studies examine the performance and accuracy of ML, only a limited few focus on its environmental aspects, particularly energy consumption. In addition, despite emerging efforts to compare energy consumption across various programming languages for specific algorithms and tasks, there remains a gap specifically in comparing these languages for ML-based tasks. This paper aims to raise awareness of the energy costs associated with employing different programming languages for ML model training and inference. Through this empirical study, we measure and compare the energy consumption along with run-time performance of five regression and five classification tasks implemented in Python and R, the two most popular programming languages in this context. Our study results reveal a statistically significant difference in costs between the two languages in 95% of the cases examined. Furthermore, our analysis demonstrates that the choice of programming language can influence energy efficiency significantly, up to 99.16% during model training and up to 99.8% during inferences, for a given ML task.
Authors: Noga Schwartz, Roy Fairstein, Avi Segal, Kobi Gal
Abstract: Early detection of struggling student programmers is crucial for providing them with personalized support. While multiple AI-based approaches have been proposed for this problem, they do not explicitly reason about students' programming skills in the model. This study addresses this gap by developing in collaboration with educators a taxonomy of proficiencies that categorizes how students solve coding tasks and is embedded in the detection model. Our model, termed the Proficiency Taxonomy Model (PTM), simultaneously learns the student's coding skills based on their coding history and predicts whether they will struggle on a new task. We extensively evaluated the effectiveness of the PTM model on two separate datasets from introductory Java and Python courses for beginner programmers. Experimental results demonstrate that PTM outperforms state-of-the-art models in predicting struggling students. The paper showcases the potential of combining structured insights from teachers for early identification of those needing assistance in learning to code.
Authors: Po-Hsien Yu, Yu-Syuan Tseng, Shao-Yi Chien
Abstract: Person re-identification (Re-ID) is a fundamental task in intelligent surveillance and public safety. Federated learning (FL) offers a privacy-preserving solution by enabling collaborative model training without centralized data collection. However, applying FL to real-world re-ID systems faces two major challenges: statistical heterogeneity across clients due to non-IID data distributions, and substantial communication overhead caused by frequent transmission of large-scale models. To address these issues, we propose FedKLPR, a lightweight and communication-efficient federated learning framework for person re-identification. FedKLPR introduces four key components. First, the KL-Divergence Regularization Loss (KLL) constrains local models by minimizing the divergence from the global feature distribution, effectively mitigating the effects of statistical heterogeneity and improving convergence stability under non-IID conditions. Secondly, KL-Divergence-Prune Weighted Aggregation (KLPWA) integrates pruning ratio and distributional similarity into the aggregation process, thereby improving the robustness of the global model while significantly reducing communication overhead. Furthermore, sparse Activation Skipping (SAS) mitigates the dilution of critical parameters during the aggregation of pruned client models by excluding zero-valued weights from the update process. Finally, Cross-Round Recovery (CRR) introduces a dynamic pruning control mechanism that halts pruning when necessary, enabling deeper compression while maintaining model accuracy. Experimental results on eight benchmark datasets demonstrate that FedKLPR achieves significant communication reduction. Compared with the state-of-the-art, FedKLPR reduces 33\%-38\% communication cost on ResNet-50 and 20\%-40\% communication cost on ResNet-34, while maintaining model accuracy within 1\% degradation.
Authors: Nikita Stroev, Natalia G. Berloff
Abstract: We unify k-local Ising optimization and optical KAN function learning on a single photonic platform, establishing a critical convergence point in optical computing that enables interleaved discrete-continuous workflows. We introduce a single spacial light modulator (SLM)-centric primitive that realizes, in one stroke, all-optical k-local Ising interactions and fully optical Kolmogorov-Arnold network (KAN) layers. The central idea is to convert structural nonlinearity of a nominally linear photonic scatterer into a per-window computational resource by adding one relay pass through the same spatial light modulator. A folded 4f relay reimages the first Fourier plane onto the SLM so that each chosen spin clique or ridge channel occupies a disjoint window with its own second-pass phase patch. Propagation remains linear in the optical field, yet the measured intensity in each window becomes a freely programmable polynomial of the clique sum or projection amplitude. This yields native, per-clique k-local couplings without nonlinear media and, in parallel, the many independent univariate nonlinearities required by KAN layers, all with in-situ physical gradients for training using two-frame (forward and adjoint) physical gradients. We outline implementation on spatial photonic Ising machines, injection-locked VCSEL arrays, and the Microsoft analog optical computers. In all cases the hardware change is one extra lens and a fold (or an on-chip 4f loop), enabling a minimal overhead, massively parallel route to high-order optical Ising optimization and trainable, all-optical KAN processing.
Authors: Suramya Jadhav, Abhay Shanbhag, Amogh Thakurdesai, Ridhima Sinare, Ananya Joshi, Raviraj Joshi
Abstract: Paraphrases are a vital tool to assist language understanding tasks such as question answering, style transfer, semantic parsing, and data augmentation tasks. Indic languages are complex in natural language processing (NLP) due to their rich morphological and syntactic variations, diverse scripts, and limited availability of annotated data. In this work, we present the L3Cube-MahaParaphrase Dataset, a high-quality paraphrase corpus for Marathi, a low resource Indic language, consisting of 8,000 sentence pairs, each annotated by human experts as either Paraphrase (P) or Non-paraphrase (NP). We also present the results of standard transformer-based BERT models on these datasets. The dataset and model are publicly shared at https://github.com/l3cube-pune/MarathiNLP
Authors: Dilermando Almeida, Guilherme Lazzarini, Juliano Negri, Thiago H. Segreto, Ricardo V. Godoy, Marcelo Becker
Abstract: Quadruped robots have emerged as highly efficient and versatile platforms, excelling in navigating complex and unstructured terrains where traditional wheeled robots might fail. Equipping these robots with manipulator arms unlocks the advanced capability of loco-manipulation to perform complex physical interaction tasks in areas ranging from industrial automation to search-and-rescue missions. However, achieving precise and adaptable grasping in such dynamic scenarios remains a significant challenge, often hindered by the need for extensive real-world calibration and pre-programmed grasp configurations. This paper introduces a deep learning framework designed to enhance the grasping capabilities of quadrupeds equipped with arms, focusing on improved precision and adaptability. Our approach centers on a sim-to-real methodology that minimizes reliance on physical data collection. We developed a pipeline within the Genesis simulation environment to generate a synthetic dataset of grasp attempts on common objects. By simulating thousands of interactions from various perspectives, we created pixel-wise annotated grasp-quality maps to serve as the ground truth for our model. This dataset was used to train a custom CNN with a U-Net-like architecture that processes multi-modal input from an onboard RGB and depth cameras, including RGB images, depth maps, segmentation masks, and surface normal maps. The trained model outputs a grasp-quality heatmap to identify the optimal grasp point. We validated the complete framework on a four-legged robot. The system successfully executed a full loco-manipulation task: autonomously navigating to a target object, perceiving it with its sensors, predicting the optimal grasp pose using our model, and performing a precise grasp. This work proves that leveraging simulated training with advanced sensing offers a scalable and effective solution for object handling.
Authors: Pedro Antonio Rabelo Saraiva, Enzo Ferreira de Souza, Joao Manoel Herrera Pinheiro, Thiago H. Segreto, Ricardo V. Godoy, Marcelo Becker
Abstract: This work addresses the challenges of data scarcity and high acquisition costs for training robust object detection models in complex industrial environments, such as offshore oil platforms. The practical and economic barriers to collecting real-world data in these hazardous settings often hamper the development of autonomous inspection systems. To overcome this, in this work we propose and validate a hybrid data synthesis pipeline that combines procedural rendering with AI-driven video generation. Our methodology leverages BlenderProc to create photorealistic images with precise annotations and controlled domain randomization, and integrates NVIDIA's Cosmos-Predict2 world-foundation model to synthesize physically plausible video sequences with temporal diversity, capturing rare viewpoints and adverse conditions. We demonstrate that a YOLO-based detection network trained on a composite dataset, blending real images with our synthetic data, achieves superior performance compared to models trained exclusively on real-world data. Notably, a 1:1 mixture of real and synthetic data yielded the highest accuracy, surpassing the real-only baseline. These findings highlight the viability of a synthetic-first approach as an efficient, cost-effective, and safe alternative for developing reliable perception systems in safety-critical and resource-constrained industrial applications.
Authors: Prathamesh Kokate, Mitali Sarnaik, Manavi Khopade, Mukta Takalikar, Raviraj Joshi
Abstract: Transformer-based models like BERT excel at short text classification but struggle with long document classification (LDC) due to input length limitations and computational inefficiencies. In this work, we propose an efficient, zero-shot approach to LDC that leverages sentence ranking to reduce input context without altering the model architecture. Our method enables the adaptation of models trained on short texts, such as headlines, to long-form documents by selecting the most informative sentences using a TF-IDF-based ranking strategy. Using the MahaNews dataset of long Marathi news articles, we evaluate three context reduction strategies that prioritize essential content while preserving classification accuracy. Our results show that retaining only the top 50\% ranked sentences maintains performance comparable to full-document inference while reducing inference time by up to 35\%. This demonstrates that sentence ranking is a simple yet effective technique for scalable and efficient zero-shot LDC.
Authors: Yiming Xu, Junfeng Jiao
Abstract: Accurately predicting travel mode choice is essential for effective transportation planning, yet traditional statistical and machine learning models are constrained by rigid assumptions, limited contextual reasoning, and reduced generalizability. This study explores the potential of Large Language Models (LLMs) as a more flexible and context-aware approach to travel mode choice prediction, enhanced by Retrieval-Augmented Generation (RAG) to ground predictions in empirical data. We develop a modular framework for integrating RAG into LLM-based travel mode choice prediction and evaluate four retrieval strategies: basic RAG, RAG with balanced retrieval, RAG with a cross-encoder for re-ranking, and RAG with balanced retrieval and cross-encoder for re-ranking. These strategies are tested across three LLM architectures (OpenAI GPT-4o, o4-mini, and o3) to examine the interaction between model reasoning capabilities and retrieval methods. Using the 2023 Puget Sound Regional Household Travel Survey data, we conduct a series of experiments to evaluate model performance. The results demonstrate that RAG substantially enhances predictive accuracy across a range of models. Notably, the GPT-4o model combined with balanced retrieval and cross-encoder re-ranking achieves the highest accuracy of 80.8%, exceeding that of conventional statistical and machine learning baselines. Furthermore, LLM-based models exhibit superior generalization abilities relative to these baselines. Findings highlight the critical interplay between LLM reasoning capabilities and retrieval strategies, demonstrating the importance of aligning retrieval strategies with model capabilities to maximize the potential of LLM-based travel behavior modeling.
Authors: Thanh Dang, Mert Gurbuzbalaban, Mohammad Rafiqul Islam, Nian Yao, Lingjiong Zhu
Abstract: Langevin algorithms are popular Markov chain Monte Carlo (MCMC) methods for large-scale sampling problems that often arise in data science. We propose Monte Carlo algorithms based on the discretizations of $P$-th order Langevin dynamics for any $P\geq 3$. Our design of $P$-th order Langevin Monte Carlo (LMC) algorithms is by combining splitting and accurate integration methods. We obtain Wasserstein convergence guarantees for sampling from distributions with log-concave and smooth densities. Specifically, the mixing time of the $P$-th order LMC algorithm scales as $O\left(d^{\frac{1}{R}}/\epsilon^{\frac{1}{2R}}\right)$ for $R=4\cdot 1_{\{ P=3\}}+ (2P-1)\cdot 1_{\{ P\geq 4\}}$, which has a better dependence on the dimension $d$ and the accuracy level $\epsilon$ as $P$ grows. Numerical experiments illustrate the efficiency of our proposed algorithms.
Authors: Weikang Wan, Jiawei Fu, Xiaodi Yuan, Yifeng Zhu, Hao Su
Abstract: Developing robotic systems capable of robustly executing long-horizon manipulation tasks with human-level dexterity is challenging, as such tasks require both physical dexterity and seamless sequencing of manipulation skills while robustly handling environment variations. While imitation learning offers a promising approach, acquiring comprehensive datasets is resource-intensive. In this work, we propose a learning framework and system LodeStar that automatically decomposes task demonstrations into semantically meaningful skills using off-the-shelf foundation models, and generates diverse synthetic demonstration datasets from a few human demos through reinforcement learning. These sim-augmented datasets enable robust skill training, with a Skill Routing Transformer (SRT) policy effectively chaining the learned skills together to execute complex long-horizon manipulation tasks. Experimental evaluations on three challenging real-world long-horizon dexterous manipulation tasks demonstrate that our approach significantly improves task performance and robustness compared to previous baselines. Videos are available at lodestar-robot.github.io.
Authors: Kairi Furui, Masahito Ohue
Abstract: In structure-based drug discovery, virtual screening using conventional molecular docking methods can be performed rapidly but suffers from limitations in prediction accuracy. Recently, Boltz-2 was proposed, achieving extremely high accuracy in binding affinity prediction, but requiring approximately 20 seconds per compound per GPU, making it difficult to apply to large-scale screening of hundreds of thousands to millions of compounds. This study proposes Boltzina, a novel framework that leverages Boltz-2's high accuracy while significantly improving computational efficiency. Boltzina achieves both accuracy and speed by omitting the rate-limiting structure prediction from Boltz-2's architecture and directly predicting affinity from AutoDock Vina docking poses. We evaluate on eight assays from the MF-PCBA dataset and show that while Boltzina performs below Boltz-2, it provides significantly higher screening performance compared to AutoDock Vina and GNINA. Additionally, Boltzina achieved up to 11.8$\times$ faster through reduced recycling iterations and batch processing. Furthermore, we investigated multi-pose selection strategies and two-stage screening combining Boltzina and Boltz-2, presenting optimization methods for accuracy and efficiency according to application requirements. This study represents the first attempt to apply Boltz-2's high-accuracy predictions to practical-scale screening, offering a pipeline that combines both accuracy and efficiency in computational biology. The Boltzina is available on github; https://github.com/ohuelab/boltzina.
Authors: Sridhar Mahadevan
Abstract: We propose a novel theory of consciousness as a functor (CF) that receives and transmits contents from unconscious memory into conscious memory. Our CF framework can be seen as a categorial formulation of the Global Workspace Theory proposed by Baars. CF models the ensemble of unconscious processes as a topos category of coalgebras. The internal language of thought in CF is defined as a Multi-modal Universal Mitchell-Benabou Language Embedding (MUMBLE). We model the transmission of information from conscious short-term working memory to long-term unconscious memory using our recently proposed Universal Reinforcement Learning (URL) framework. To model the transmission of information from unconscious long-term memory into resource-constrained short-term memory, we propose a network economic model.
Authors: Daniel Frees, Moritz Bolling, Aditri Bhagirath
Abstract: Modern computer vision models have proven to be highly useful for medical imaging classification and segmentation tasks, but the scarcity of medical imaging data often limits the efficacy of models trained from scratch. Transfer learning has emerged as a pivotal solution to this, enabling the fine-tuning of high-performance models on small data. Mei et al. (2022) found that pre-training CNNs on a large dataset of radiologist-labeled images (RadImageNet) enhanced model performance on downstream tasks compared to ImageNet pretraining. The present work extends Mei et al. (2022) by conducting a comprehensive investigation to determine optimal CNN architectures for breast lesion malignancy detection and ACL tear detection, as well as performing statistical analysis to compare the effect of RadImageNet and ImageNet pre-training on downstream model performance. Our findings suggest that 1-dimensional convolutional classifiers with skip connections, ResNet50 pre-trained backbones, and partial backbone unfreezing yields optimal downstream medical classification performance. Our best models achieve AUCs of 0.9969 for ACL tear detection and 0.9641 for breast nodule malignancy detection, competitive with the results reported by Mei et al. (2022) and surpassing other previous works. We do not find evidence confirming RadImageNet pre-training to provide superior downstream performance for ACL tear and breast lesion classification tasks.
Authors: Liane Makatura, Benjamin Jones, Siyuan Bian, Wojciech Matusik
Abstract: Metamaterials are micro-architected structures whose geometry imparts highly tunable-often counter-intuitive-bulk properties. Yet their design is difficult because of geometric complexity and a non-trivial mapping from architecture to behaviour. We address these challenges with three complementary contributions. (i) MetaDSL: a compact, semantically rich domain-specific language that captures diverse metamaterial designs in a form that is both human-readable and machine-parsable. (ii) MetaDB: a curated repository of more than 150,000 parameterized MetaDSL programs together with their derivatives-three-dimensional geometry, multi-view renderings, and simulated elastic properties. (iii) MetaBench: benchmark suites that test three core capabilities of vision-language metamaterial assistants-structure reconstruction, property-driven inverse design, and performance prediction. We establish baselines by fine-tuning state-of-the-art vision-language models and deploy an omni-model within an interactive, CAD-like interface. Case studies show that our framework provides a strong first step toward integrated design and understanding of structure-representation-property relationships.
Authors: Daniel Frees, Martin Pollack
Abstract: Despite the overwhelming performance improvements offered by recent natural language procesing (NLP) models, the decisions made by these models are largely a black box. Towards closing this gap, the field of causal NLP combines causal inference literature with modern NLP models to elucidate causal effects of text features. We replicate and extend Bansal et al's work on regularizing text classifiers to adhere to estimated effects, focusing instead on model interpretability. Specifically, we focus on developing a two-headed RieszNet-based neural network architecture which achieves better treatment effect estimation accuracy. Our framework, CausalSent, accurately predicts treatment effects in semi-synthetic IMDB movie reviews, reducing MAE of effect estimates by 2-3x compared to Bansal et al's MAE on synthetic Civil Comments data. With an ensemble of validated models, we perform an observational case study on the causal effect of the word "love" in IMDB movie reviews, finding that the presence of the word "love" causes a +2.9% increase in the probability of a positive sentiment.
Authors: Fan Nie, Ken Ziyu Liu, Zihao Wang, Rui Sun, Wei Liu, Weijia Shi, Huaxiu Yao, Linjun Zhang, Andrew Y. Ng, James Zou, Sanmi Koyejo, Yejin Choi, Percy Liang, Niklas Muennighoff
Abstract: Benchmarks shape progress in AI research. A useful benchmark should be both difficult and realistic: questions should challenge frontier models while also reflecting real-world usage. Yet, current paradigms face a difficulty-realism tension: exam-style benchmarks are often made artificially difficult with limited real-world value, while benchmarks based on real user interaction often skew toward easy, high-frequency problems. In this work, we explore a radically different paradigm: assessing models on unsolved questions. Rather than a static benchmark scored once, we curate unsolved questions and evaluate models asynchronously over time with validator-assisted screening and community verification. We introduce UQ, a testbed of 500 challenging, diverse questions sourced from Stack Exchange, spanning topics from CS theory and math to sci-fi and history, probing capabilities including reasoning, factuality, and browsing. UQ is difficult and realistic by construction: unsolved questions are often hard and naturally arise when humans seek answers, thus solving them yields direct real-world value. Our contributions are threefold: (1) UQ-Dataset and its collection pipeline combining rule-based filters, LLM judges, and human review to ensure question quality (e.g., well-defined and difficult); (2) UQ-Validators, compound validation strategies that leverage the generator-validator gap to provide evaluation signals and pre-screen candidate solutions for human review; and (3) UQ-Platform, an open platform where experts collectively verify questions and solutions. The top model passes UQ-validation on only 15% of questions, and preliminary human verification has already identified correct answers among those that passed. UQ charts a path for evaluating frontier models on real-world, open-ended challenges, where success pushes the frontier of human knowledge. We release UQ at https://uq.stanford.edu.
URLs: https://uq.stanford.edu.
Authors: Guanxing Lu, Baoxiong Jia, Puhao Li, Yixin Chen, Ziwei Wang, Yansong Tang, Siyuan Huang
Abstract: Training robot policies within a learned world model is trending due to the inefficiency of real-world interactions. The established image-based world models and policies have shown prior success, but lack robust geometric information that requires consistent spatial and physical understanding of the three-dimensional world, even pre-trained on internet-scale video sources. To this end, we propose a novel branch of world model named Gaussian World Model (GWM) for robotic manipulation, which reconstructs the future state by inferring the propagation of Gaussian primitives under the effect of robot actions. At its core is a latent Diffusion Transformer (DiT) combined with a 3D variational autoencoder, enabling fine-grained scene-level future state reconstruction with Gaussian Splatting. GWM can not only enhance the visual representation for imitation learning agent by self-supervised future prediction training, but can serve as a neural simulator that supports model-based reinforcement learning. Both simulated and real-world experiments depict that GWM can precisely predict future scenes conditioned on diverse robot actions, and can be further utilized to train policies that outperform the state-of-the-art by impressive margins, showcasing the initial data scaling potential of 3D world model.
Authors: Alireza Fallah, Michael I. Jordan, Annie Ulichney
Abstract: Machine learning models must balance accuracy and fairness, but these goals often conflict, particularly when data come from multiple demographic groups. A useful tool for understanding this trade-off is the fairness-accuracy (FA) frontier, which characterizes the set of models that cannot be simultaneously improved in both fairness and accuracy. Prior analyses of the FA frontier provide a full characterization under the assumption of complete knowledge of population distributions -- an unrealistic ideal. We study the FA frontier in the finite-sample regime, showing how it deviates from its population counterpart and quantifying the worst-case gap between them. In particular, we derive minimax-optimal estimators that depend on the designer's knowledge of the covariate distribution. For each estimator, we characterize how finite-sample effects asymmetrically impact each group's risk, and identify optimal sample allocation strategies. Our results transform the FA frontier from a theoretical construct into a practical tool for policymakers and practitioners who must often design algorithms with limited data.
Authors: Kaushik Ravi, Andreas Br\"uck
Abstract: Urban climate resilience requires more than high-resolution data; it demands systems that embed data collection, interpretation, and action within the daily lives of citizens. This chapter presents a scalable, citizen-centric framework that reimagines environmental infrastructure through participatory sensing, open analytics, and prescriptive urban planning tools. Applied in Pune, India, the framework comprises three interlinked modules: (1) a smartphone-based measurement toolkit enhanced by AI segmentation to extract tree height, canopy diameter, and trunk girth; (2) a percentile-based model using satellite-derived Land Surface Temperature to calculate localized cooling through two new metrics, Cooling Efficacy and Ambient Heat Relief; and (3) an eco-routing engine that guides mobility using a Static Environmental Quality score, based on tree density, species diversity, and cumulative carbon sequestration. Together, these modules form a closed feedback loop where citizens generate actionable data and benefit from personalized, sustainable interventions. This framework transforms open data from a passive repository into an active platform for shared governance and environmental equity. In the face of growing ecological inequality and data centralization, this chapter presents a replicable model for citizen-driven urban intelligence, reframing planning as a co-produced, climate-resilient, and radically local practice.
Authors: Minhyeong Lee, Suyoung Hwang, Seunghyun Moon, Geonho Nah, Donghyun Koh, Youngjun Cho, Johyun Park, Hojin Yoo, Jiho Park, Haneul Choi, Sungbin Moon, Taehoon Hwang, Seungwon Kim, Jaeyeong Kim, Seongjun Kim, Juneau Jung
Abstract: Recent advances in LLMs have made automated scientific research the next frontline in the path to artificial superintelligence. However, these systems are bound either to tasks of narrow scope or the limited creative capabilities of LLMs. We propose Spacer, a scientific discovery system that develops creative and factually grounded concepts without external intervention. Spacer attempts to achieve this via 'deliberate decontextualization,' an approach that disassembles information into atomic units - keywords - and draws creativity from unexplored connections between them. Spacer consists of (i) Nuri, an inspiration engine that builds keyword sets, and (ii) the Manifesting Pipeline that refines these sets into elaborate scientific statements. Nuri extracts novel, high-potential keyword sets from a keyword graph built with 180,000 academic publications in biological fields. The Manifesting Pipeline finds links between keywords, analyzes their logical structure, validates their plausibility, and ultimately drafts original scientific concepts. According to our experiments, the evaluation metric of Nuri accurately classifies high-impact publications with an AUROC score of 0.737. Our Manifesting Pipeline also successfully reconstructs core concepts from the latest top-journal articles solely from their keyword sets. An LLM-based scoring system estimates that this reconstruction was sound for over 85% of the cases. Finally, our embedding space analysis shows that outputs from Spacer are significantly more similar to leading publications compared with those from SOTA LLMs.
Authors: Qiming Guo, Jinwen Tang, Xingran Huang
Abstract: We introduce Advertisement Embedding Attacks (AEA), a new class of LLM security threats that stealthily inject promotional or malicious content into model outputs and AI agents. AEA operate through two low-cost vectors: (1) hijacking third-party service-distribution platforms to prepend adversarial prompts, and (2) publishing back-doored open-source checkpoints fine-tuned with attacker data. Unlike conventional attacks that degrade accuracy, AEA subvert information integrity, causing models to return covert ads, propaganda, or hate speech while appearing normal. We detail the attack pipeline, map five stakeholder victim groups, and present an initial prompt-based self-inspection defense that mitigates these injections without additional model retraining. Our findings reveal an urgent, under-addressed gap in LLM security and call for coordinated detection, auditing, and policy responses from the AI-safety community.
Authors: Danny Wang, Ruihong Qiu, Guangdong Bai, Zi Huang
Abstract: Out-of-distribution (OOD) detection remains challenging in text-rich networks, where textual features intertwine with topological structures. Existing methods primarily address label shifts or rudimentary domain-based splits, overlooking the intricate textual-structural diversity. For example, in social networks, where users represent nodes with textual features (name, bio) while edges indicate friendship status, OOD may stem from the distinct language patterns between bot and normal users. To address this gap, we introduce the TextTopoOOD framework for evaluating detection across diverse OOD scenarios: (1) attribute-level shifts via text augmentations and embedding perturbations; (2) structural shifts through edge rewiring and semantic connections; (3) thematically-guided label shifts; and (4) domain-based divisions. Furthermore, we propose TNT-OOD to model the complex interplay between Text aNd Topology using: 1) a novel cross-attention module to fuse local structure into node-level text representations, and 2) a HyperNetwork to generate node-specific transformation parameters. This aligns topological and semantic features of ID nodes, enhancing ID/OOD distinction across structural and textual shifts. Experiments on 11 datasets across four OOD scenarios demonstrate the nuanced challenge of TextTopoOOD for evaluating OOD detection in text-rich networks.
Authors: Nisreen Albzour, Sarah S. Lam
Abstract: Cervical cancer remains a significant global health concern and a leading cause of cancer-related deaths among women. Early detection through Pap smear tests is essential to reduce mortality rates; however, the manual examination is time consuming and prone to human error. This study proposes a deep learning framework that integrates U-Net for segmentation and a classification model to enhance diagnostic performance. The Herlev Pap Smear Dataset, a publicly available cervical cell dataset, was utilized for training and evaluation. The impact of segmentation on classification performance was evaluated by comparing the model trained on segmented images and another trained on non-segmented images. Experimental results showed that the use of segmented images marginally improved the model performance on precision (about 0.41 percent higher) and F1-score (about 1.30 percent higher), which suggests a slightly more balanced classification performance. While segmentation helps in feature extraction, the results showed that its impact on classification performance appears to be limited. The proposed framework offers a supplemental tool for clinical applications, which may aid pathologists in early diagnosis.
Authors: Guangwei Zhang, Qisheng Su, Jiateng Liu, Cheng Qian, Yanzhou Pan, Yanjie Fu, Denghui Zhang
Abstract: Large Language Models (LLMs) have revolutionized Natural Language Processing (NLP) but pose risks of inadvertently exposing copyrighted or proprietary data, especially when such data is used for training but not intended for distribution. Traditional methods address these leaks only after content is generated, which can lead to the exposure of sensitive information. This study introduces a proactive approach: examining LLMs' internal states before text generation to detect potential leaks. By using a curated dataset of copyrighted materials, we trained a neural network classifier to identify risks, allowing for early intervention by stopping the generation process or altering outputs to prevent disclosure. Integrated with a Retrieval-Augmented Generation (RAG) system, this framework ensures adherence to copyright and licensing requirements while enhancing data privacy and ethical standards. Our results show that analyzing internal states effectively mitigates the risk of copyrighted data leakage, offering a scalable solution that fits smoothly into AI workflows, ensuring compliance with copyright regulations while maintaining high-quality text generation. The implementation is available on GitHub.\footnote{https://github.com/changhu73/Internal_states_leakage}
Authors: Ryoya Fukasaku, Yutaro Kabata, Akifumi Okuno
Abstract: This paper investigates a perceptron, a simple neural network model, with ReLU activation and a ridge-regularized mean squared error (RR-MSE). Our approach leverages the fact that the RR-MSE for ReLU perceptron is piecewise polynomial, enabling a systematic analysis using tools from computational algebra. In particular, we develop a Divide-Enumerate-Merge strategy that exhaustively enumerates all local minima of the RR-MSE. By virtue of the algebraic formulation, our approach can identify not only the typical zero-dimensional minima (i.e., isolated points) obtained by numerical optimization, but also higher-dimensional minima (i.e., connected sets such as curves, surfaces, or hypersurfaces). Although computational algebraic methods are computationally very intensive for perceptrons of practical size, as a proof of concept, we apply the proposed approach in practice to minimal perceptrons with a few hidden units.
Authors: Andrea Brunello, Luca Geatti, Angelo Montanari, Nicola Saccomanno
Abstract: Monitoring is a runtime verification technique that allows one to check whether an ongoing computation of a system (partial trace) satisfies a given formula. It does not need a complete model of the system, but it typically requires the construction of a deterministic automaton doubly exponential in the size of the formula (in the worst case), which limits its practicality. In this paper, we show that, when considering finite, discrete traces, monitoring of pure past (co)safety fragments of Signal Temporal Logic (STL) can be reduced to trace checking, that is, evaluation of a formula over a trace, that can be performed in time polynomial in the size of the formula and the length of the trace. By exploiting such a result, we develop a GPU-accelerated framework for interpretable early failure detection based on vectorized trace checking, that employs genetic programming to learn temporal properties from historical trace data. The framework shows a 2-10% net improvement in key performance metrics compared to the state-of-the-art methods.
Authors: Muhammad Aqeel, Shakiba Sharifi, Marco Cristani, Francesco Setti
Abstract: Anomaly detection is fundamental for ensuring quality control and operational efficiency in industrial environments, yet conventional approaches face significant challenges when training data contains mislabeled samples-a common occurrence in real-world scenarios. This paper presents RAD, a robust anomaly detection framework that integrates Normalizing Flows with Model-Agnostic Meta-Learning to address the critical challenge of label noise in industrial settings. Our approach employs a bi-level optimization strategy where meta-learning enables rapid adaptation to varying noise conditions, while uncertainty quantification guides adaptive L2 regularization to maintain model stability. The framework incorporates multiscale feature processing through pretrained feature extractors and leverages the precise likelihood estimation capabilities of Normalizing Flows for robust anomaly scoring. Comprehensive evaluation on MVTec-AD and KSDD2 datasets demonstrates superior performance, achieving I-AUROC scores of 95.4% and 94.6% respectively under clean conditions, while maintaining robust detection capabilities above 86.8% and 92.1% even when 50% of training samples are mislabeled. The results highlight RAD's exceptional resilience to noisy training conditions and its ability to detect subtle anomalies across diverse industrial scenarios, making it a practical solution for real-world anomaly detection applications where perfect data curation is challenging.
Authors: Hanzhi Chang, Ruijie Zhu, Wenjie Chang, Mulin Yu, Yanzhe Liang, Jiahao Lu, Zhuoyuan Li, Tianzhu Zhang
Abstract: Surface reconstruction has been widely studied in computer vision and graphics. However, existing surface reconstruction works struggle to recover accurate scene geometry when the input views are extremely sparse. To address this issue, we propose MeshSplat, a generalizable sparse-view surface reconstruction framework via Gaussian Splatting. Our key idea is to leverage 2DGS as a bridge, which connects novel view synthesis to learned geometric priors and then transfers these priors to achieve surface reconstruction. Specifically, we incorporate a feed-forward network to predict per-view pixel-aligned 2DGS, which enables the network to synthesize novel view images and thus eliminates the need for direct 3D ground-truth supervision. To improve the accuracy of 2DGS position and orientation prediction, we propose a Weighted Chamfer Distance Loss to regularize the depth maps, especially in overlapping areas of input views, and also a normal prediction network to align the orientation of 2DGS with normal vectors predicted by a monocular normal estimator. Extensive experiments validate the effectiveness of our proposed improvement, demonstrating that our method achieves state-of-the-art performance in generalizable sparse-view mesh reconstruction tasks. Project Page: https://hanzhichang.github.io/meshsplat_web
Authors: Muhammad Aqeel, Danijel Skocaj, Marco Cristani, Francesco Setti
Abstract: Industrial and medical anomaly detection faces critical challenges from data scarcity and prohibitive annotation costs, particularly in evolving manufacturing and healthcare settings. To address this, we propose CoZAD, a novel zero-shot anomaly detection framework that integrates soft confident learning with meta-learning and contrastive feature representation. Unlike traditional confident learning that discards uncertain samples, our method assigns confidence-based weights to all training data, preserving boundary information while emphasizing prototypical normal patterns. The framework quantifies data uncertainty through IQR-based thresholding and model uncertainty via covariance based regularization within a Model-Agnostic Meta-Learning. Contrastive learning creates discriminative feature spaces where normal patterns form compact clusters, enabling rapid domain adaptation. Comprehensive evaluation across 10 datasets spanning industrial and medical domains demonstrates state-of-the-art performance, outperforming existing methods on 6 out of 7 industrial benchmarks with notable improvements on texture-rich datasets (99.2% I-AUROC on DTD-Synthetic, 97.2% on BTAD) and pixellevel localization (96.3% P-AUROC on MVTec-AD). The framework eliminates dependence on vision-language alignments or model ensembles, making it valuable for resourceconstrained environments requiring rapid deployment.
Authors: Maham Nazir, Muhammad Aqeel, Francesco Setti
Abstract: Medical image segmentation models struggle with rare abnormalities due to scarce annotated pathological data. We propose DiffAug a novel framework that combines textguided diffusion-based generation with automatic segmentation validation to address this challenge. Our proposed approach uses latent diffusion models conditioned on medical text descriptions and spatial masks to synthesize abnormalities via inpainting on normal images. Generated samples undergo dynamic quality validation through a latentspace segmentation network that ensures accurate localization while enabling single-step inference. The text prompts, derived from medical literature, guide the generation of diverse abnormality types without requiring manual annotation. Our validation mechanism filters synthetic samples based on spatial accuracy, maintaining quality while operating efficiently through direct latent estimation. Evaluated on three medical imaging benchmarks (CVC-ClinicDB, Kvasir-SEG, REFUGE2), our framework achieves state-of-the-art performance with 8-10% Dice improvements over baselines and reduces false negative rates by up to 28% for challenging cases like small polyps and flat lesions critical for early detection in screening applications.
Authors: Yang Chen, Yanbin Wei, Ke Jin, Yi Kong, James Kwok, Yu Zhang
Abstract: Recent advances in pre-trained vision-language models have demonstrated remarkable zero-shot generalization capabilities. To further enhance these models' adaptability to various downstream tasks, prompt tuning has emerged as a parameter-efficient fine-tuning method. However, despite its efficiency, the generalization ability of prompt remains limited. In contrast, label smoothing (LS) has been widely recognized as an effective regularization technique that prevents models from becoming over-confident and improves their generalization. This inspires us to explore the integration of LS with prompt tuning. However, we have observed that the vanilla LS even weakens the generalization ability of prompt tuning. To address this issue, we propose the Alternating Training-based Label Smoothing (ATLaS) method, which alternately trains with standard one-hot labels and soft labels generated by LS to supervise the prompt tuning. Moreover, we introduce two types of efficient offline soft labels, including Class-wise Soft Labels (CSL) and Instance-wise Soft Labels (ISL), to provide inter-class or instance-class relationships for prompt tuning. The theoretical properties of the proposed ATLaS method are analyzed. Extensive experiments demonstrate that the proposed ATLaS method, combined with CSL and ISL, consistently enhances the generalization performance of prompt tuning. Moreover, the proposed ATLaS method exhibits high compatibility with prevalent prompt tuning methods, enabling seamless integration into existing methods.
Authors: Takuhiro Kaneko, Hirokazu Kameoka, Kou Tanaka, Yuto Kondo
Abstract: A diffusion-based voice conversion (VC) model (e.g., VoiceGrad) can achieve high speech quality and speaker similarity; however, its conversion process is slow owing to iterative sampling. FastVoiceGrad overcomes this limitation by distilling VoiceGrad into a one-step diffusion model. However, it still requires a computationally intensive content encoder to disentangle the speaker's identity and content, which slows conversion. Therefore, we propose FasterVoiceGrad, a novel one-step diffusion-based VC model obtained by simultaneously distilling a diffusion model and content encoder using adversarial diffusion conversion distillation (ADCD), where distillation is performed in the conversion process while leveraging adversarial and score distillation training. Experimental evaluations of one-shot VC demonstrated that FasterVoiceGrad achieves competitive VC performance compared to FastVoiceGrad, with 6.6-6.9 and 1.8 times faster speed on a GPU and CPU, respectively.
Authors: Takuhiro Kaneko, Hirokazu Kameoka, Kou Tanaka, Yuto Kondo
Abstract: In text-to-speech (TTS) and voice conversion (VC), acoustic features, such as mel spectrograms, are typically used as synthesis or conversion targets owing to their compactness and ease of learning. However, because the ultimate goal is to generate high-quality waveforms, employing a vocoder to convert these features into waveforms and applying adversarial training in the time domain is reasonable. Nevertheless, upsampling the waveform introduces significant time and memory overheads. To address this issue, we propose a vocoder-projected feature discriminator (VPFD), which uses vocoder features for adversarial training. Experiments on diffusion-based VC distillation demonstrated that a pretrained and frozen vocoder feature extractor with a single upsampling step is necessary and sufficient to achieve a VC performance comparable to that of waveform discriminators while reducing the training time and memory consumption by 9.6 and 11.4 times, respectively.
Authors: Manlai Liang, Mandi Liu, Jiangzhou Ji, Huaijun Li, Haobo Yang, Yaohan He, Jinlong Li
Abstract: Large Language Models (LLMs) have demonstrated success across many benchmarks. However, they still exhibit limitations in long-context scenarios, primarily due to their short effective context length, quadratic computational complexity, and high memory overhead when processing lengthy inputs. To mitigate these issues, we introduce a novel context compression pipeline, called Intermediate Layer Retrieval (ILRe), which determines one intermediate decoder layer offline, encodes context by streaming chunked prefill only up to that layer, and recalls tokens by the attention scores between the input query and full key cache in that specified layer. In particular, we propose a multi-pooling kernels allocating strategy in the token recalling process to maintain the completeness of semantics. Our approach not only reduces the prefilling complexity from $O(L^2)$ to $O(L)$, but also achieves performance comparable to or better than the full context in the long context scenarios. Without additional post training or operator development, ILRe can process a single $1M$ tokens request in less than half a minute (speedup $\approx 180\times$) and scores RULER-$1M$ benchmark of $\approx 79.8$ with model Llama-3.1-UltraLong-8B-1M-Instruct on a Huawei Ascend 910B NPU.
Authors: Siddarth Srinivasan, Tao Lin, Connacher Murphy, Anish Thilagar, Yiling Chen, Ezra Karger
Abstract: Competitions are widely used to identify top performers in judgmental forecasting and machine learning, and the standard competition design ranks competitors based on their cumulative scores against a set of realized outcomes or held-out labels. However, this standard design is neither incentive-compatible nor very statistically efficient. The main culprit is noise in outcomes/labels that experts are scored against; it allows weaker competitors to often win by chance, and the winner-take-all nature incentivizes misreporting that improves win probability even if it decreases expected score. Attempts to achieve incentive-compatibility rely on randomized mechanisms that add even more noise in winner selection, but come at the cost of determinism and practical adoption. To tackle these issues, we introduce a novel deterministic mechanism: WOMAC (Wisdom of the Most Accurate Crowd). Instead of scoring experts against noisy outcomes, as is standard, WOMAC scores experts against the best ex-post aggregate of peer experts' predictions given the noisy outcomes. WOMAC is also more efficient than the standard competition design in typical settings. While the increased complexity of WOMAC makes it challenging to analyze incentives directly, we provide a clear theoretical foundation to justify the mechanism. We also provide an efficient vectorized implementation and demonstrate empirically on real-world forecasting datasets that WOMAC is a more reliable predictor of experts' out-of-sample performance relative to the standard mechanism. WOMAC is useful in any competition where there is substantial noise in the outcomes/labels.
Authors: Ana Mart\'inez-Sabiote, Michalis Skotiniotis, Jara J. Bermejo-Vega, Daniel Manzano, Carlos Cano
Abstract: This work presents a machine learning approach based on support vector machines (SVMs) for quantum entanglement detection. Particularly, we focus in bipartite systems of dimensions 3x3, 4x4, and 5x5, where the positive partial transpose criterion (PPT) provides only partial characterization. Using SVMs with quantum-inspired kernels we develop a classification scheme that distinguishes between separable states, PPT-detectable entangled states, and entangled states that evade PPT detection. Our method achieves increasing accuracy with system dimension, reaching 80%, 90%, and nearly 100% for 3x3, 4x4, and 5x5 systems, respectively. Our results show that principal component analysis significantly enhances performance for small training sets. The study reveals important practical considerations regarding purity biases in the generation of data for this problem and examines the challenges of implementing these techniques on near-term quantum hardware. Our results establish machine learning as a powerful complement to traditional entanglement detection methods, particularly for higher-dimensional systems where conventional approaches become inadequate. The findings highlight key directions for future research, including hybrid quantum-classical implementations and improved data generation protocols to overcome current limitations.
Authors: Qiwei Peng, Guimin Hu, Yekun Chai, Anders S{\o}gaard
Abstract: Debiasing techniques such as SentDebias aim to reduce bias in large language models (LLMs). Previous studies have evaluated their cross-lingual transferability by directly applying these methods to LLM representations, revealing their limited effectiveness across languages. In this work, we therefore propose to perform debiasing in a joint latent space rather than directly on LLM representations. We construct a well-aligned cross-lingual latent space using an autoencoder trained on parallel TED talk scripts. Our experiments with Aya-expanse and two debiasing techniques across four languages (English, French, German, Dutch) demonstrate that a) autoencoders effectively construct a well-aligned cross-lingual latent space, and b) applying debiasing techniques in the learned cross-lingual latent space significantly improves both the overall debiasing performance and cross-lingual transferability.
Authors: Qiwei Peng, Yekun Chai, Anders S{\o}gaard
Abstract: Large language models (LLMs) take sequences of subwords as input, requiring them to effective compose subword representations into meaningful word-level representations. In this paper, we present a comprehensive set of experiments to probe how LLMs compose subword information, focusing on three key aspects: structural similarity, semantic decomposability, and form retention. Our analysis of the experiments suggests that these five LLM families can be classified into three distinct groups, likely reflecting difference in their underlying composition strategies. Specifically, we observe (i) three distinct patterns in the evolution of structural similarity between subword compositions and whole-word representations across layers; (ii) great performance when probing layer by layer their sensitivity to semantic decompositionality; and (iii) three distinct patterns when probing sensitivity to formal features, e.g., character sequence length. These findings provide valuable insights into the compositional dynamics of LLMs and highlight different compositional pattens in how LLMs encode and integrate subword information.
Authors: Eduardo de Conto, Blaise Genest, Arvind Easwaran, Nicholas Ng, Shweta Menon
Abstract: Digital twins (DTs) are increasingly utilized to monitor, manage, and optimize complex systems across various domains, including civil engineering. A core requirement for an effective DT is to act as a fast, accurate, and maintainable surrogate of its physical counterpart, the physical twin (PT). To this end, machine learning (ML) is frequently employed to (i) construct real-time DT prototypes using efficient reduced-order models (ROMs) derived from high-fidelity simulations of the PT's nominal behavior, and (ii) specialize these prototypes into DT instances by leveraging historical sensor data from the target PT. Despite the broad applicability of ML, its use in DT engineering remains largely ad hoc. Indeed, while conventional ML pipelines often train a single model for a specific task, DTs typically require multiple, task- and domain-dependent models. Thus, a more structured approach is required to design DTs. In this paper, we introduce DesCartes Builder, an open-source tool to enable the systematic engineering of ML-based pipelines for real-time DT prototypes and DT instances. The tool leverages an open and flexible visual data flow paradigm to facilitate the specification, composition, and reuse of ML models. It also integrates a library of parameterizable core operations and ML algorithms tailored for DT design. We demonstrate the effectiveness and usability of DesCartes Builder through a civil engineering use case involving the design of a real-time DT prototype to predict the plastic strain of a structure.
Authors: Alessio Falai, Ziyao Zhang, Akos Gangoly
Abstract: In this paper we investigate cross-lingual Text-To-Speech (TTS) synthesis through the lens of adapters, in the context of lightweight TTS systems. In particular, we compare the tasks of unseen speaker and language adaptation with the goal of synthesising a target voice in a target language, in which the target voice has no recordings therein. Results from objective evaluations demonstrate the effectiveness of adapters in learning language-specific and speaker-specific information, allowing pre-trained models to learn unseen speaker identities or languages, while avoiding catastrophic forgetting of the original model's speaker or language information. Additionally, to measure how native the generated voices are in terms of accent, we propose and validate an objective metric inspired by mispronunciation detection techniques in second-language (L2) learners. The paper also provides insights into the impact of adapter placement, configuration and the number of speakers used.
Authors: Sochukwuma Nwokoye, Desmond Moru
Abstract: Neural networks in assistive technology for visually impaired leverage artificial intelligence's capacity to recognize patterns in complex data. They are used for converting visual data into auditory or tactile representations, helping the visually impaired understand their surroundings. The primary aim of this research is to explore the potential of artificial neural networks to facilitate the differentiation of various forms of cash for individuals with visual impairments. In this study, we built a custom dataset of 3,468 images, which was subsequently used to train an SSD neural network model. The proposed system can accurately identify Nigerian cash, thereby streamlining commercial transactions. The performance of the system in terms of accuracy was assessed, and the Mean Average Precision score was over 90%. We believe that our system has the potential to make a substantial contribution to the field of assistive technology while also improving the quality of life of visually challenged persons in Nigeria and beyond.
Authors: Alberto Silvio Chiappa, Boshi An, Merkourios Simos, Chengkun Li, Alexander Mathis
Abstract: Controlling high-dimensional and nonlinear musculoskeletal models of the human body is a foundational scientific challenge. Recent machine learning breakthroughs have heralded policies that master individual skills like reaching, object manipulation and locomotion in musculoskeletal systems with many degrees of freedom. However, these agents are merely "specialists", achieving high performance for a single skill. In this work, we develop Arnold, a generalist policy that masters multiple tasks and embodiments. Arnold combines behavior cloning and fine-tuning with PPO to achieve expert or super-expert performance in 14 challenging control tasks from dexterous object manipulation to locomotion. A key innovation is Arnold's sensorimotor vocabulary, a compositional representation of the semantics of heterogeneous sensory modalities, objectives, and actuators. Arnold leverages this vocabulary via a transformer architecture to deal with the variable observation and action spaces of each task. This framework supports efficient multi-task, multi-embodiment learning and facilitates rapid adaptation to novel tasks. Finally, we analyze Arnold to provide insights into biological motor control, corroborating recent findings on the limited transferability of muscle synergies across tasks.
Authors: Federico Marcuzzi, Xuefei Ning, Roy Schwartz, Iryna Gurevych
Abstract: This work presents a comprehensive evaluation of how quantization affects model bias, with particular attention to its impact on individual demographic subgroups. We focus on weight and activation quantization strategies and examine their effects across a broad range of bias types, including stereotypes, toxicity, sentiment, and fairness. We employ both probabilistic and generated text-based metrics across nine benchmarks and evaluate models varying in architecture family and reasoning ability. Our findings show that quantization has a nuanced impact on bias: while it can reduce model toxicity and does not significantly impact sentiment, it tends to slightly increase stereotypes and unfairness in generative tasks, especially under aggressive compression. These trends are generally consistent across demographic categories and model types, although their magnitude depends on the specific setting. Overall, our results highlight the importance of carefully balancing efficiency and ethical considerations when applying quantization in practice.
Authors: Zhicong Tang, Tiankai Hang, Shuyang Gu, Dong Chen, Baining Guo
Abstract: This paper aims to unify Score-based Generative Models (SGMs), also known as Diffusion models, and the Schr\"odinger Bridge (SB) problem through three reparameterization techniques: Iterative Proportional Mean-Matching (IPMM), Iterative Proportional Terminus-Matching (IPTM), and Iterative Proportional Flow-Matching (IPFM). These techniques significantly accelerate and stabilize the training of SB-based models. Furthermore, the paper introduces novel initialization strategies that use pre-trained SGMs to effectively train SB-based models. By using SGMs as initialization, we leverage the advantages of both SB-based models and SGMs, ensuring efficient training of SB-based models and further improving the performance of SGMs. Extensive experiments demonstrate the significant effectiveness and improvements of the proposed methods. We believe this work contributes to and paves the way for future research on generative models.
Authors: Jatin Nainani, Sankaran Vaidyanathan, Connor Watts, Andre N. Assis, Alice Rigg
Abstract: Modern large language models (LLMs) have demonstrated impressive performance across a wide range of multi-step reasoning tasks. Recent work suggests that LLMs may perform planning - selecting a future target token in advance and generating intermediate tokens that lead towards it - rather than merely improvising one token at a time. However, existing studies assume fixed planning horizons and often focus on single prompts or narrow domains. To distinguish planning from improvisation across models and tasks, we present formal and causally grounded criteria for detecting planning and operationalize them as a semi-automated annotation pipeline. We apply this pipeline to both base and instruction-tuned Gemma-2-2B models on the MBPP code generation benchmark and a poem generation task where Claude 3.5 Haiku was previously shown to plan. Our findings show that planning is not universal: unlike Haiku, Gemma-2-2B solves the same poem generation task through improvisation, and on MBPP it switches between planning and improvisation across similar tasks and even successive token predictions. We further show that instruction tuning refines existing planning behaviors in the base model rather than creating them from scratch. Together, these studies provide a reproducible and scalable foundation for mechanistic studies of planning in LLMs.
Authors: Farkhad Akimov, Munachiso Samuel Nwadike, Zangir Iklassov, Martin Tak\'a\v{c}
Abstract: Imagine decision-makers uploading data and, within minutes, receiving clear, actionable insights delivered straight to their fingertips. That is the promise of the AI Data Scientist, an autonomous Agent powered by large language models (LLMs) that closes the gap between evidence and action. Rather than simply writing code or responding to prompts, it reasons through questions, tests ideas, and delivers end-to-end insights at a pace far beyond traditional workflows. Guided by the scientific tenet of the hypothesis, this Agent uncovers explanatory patterns in data, evaluates their statistical significance, and uses them to inform predictive modeling. It then translates these results into recommendations that are both rigorous and accessible. At the core of the AI Data Scientist is a team of specialized LLM Subagents, each responsible for a distinct task such as data cleaning, statistical testing, validation, and plain-language communication. These Subagents write their own code, reason about causality, and identify when additional data is needed to support sound conclusions. Together, they achieve in minutes what might otherwise take days or weeks, enabling a new kind of interaction that makes deep data science both accessible and actionable.
Authors: Hung-Chun Hsu, Yuan-Ching Kuo, Chao-Han Huck Yang, Szu-Wei Fu, Hanrong Ye, Hongxu Yin, Yu-Chiang Frank Wang, Ming-Feng Tsai, Chuan-Ju Wang
Abstract: The rapid evolution of e-commerce has exposed the limitations of traditional product retrieval systems in managing complex, multi-turn user interactions. Recent advances in multimodal generative retrieval -- particularly those leveraging multimodal large language models (MLLMs) as retrievers -- have shown promise. However, most existing methods are tailored to single-turn scenarios and struggle to model the evolving intent and iterative nature of multi-turn dialogues when applied naively. Concurrently, test-time scaling has emerged as a powerful paradigm for improving large language model (LLM) performance through iterative inference-time refinement. Yet, its effectiveness typically relies on two conditions: (1) a well-defined problem space (e.g., mathematical reasoning), and (2) the model's ability to self-correct -- conditions that are rarely met in conversational product search. In this setting, user queries are often ambiguous and evolving, and MLLMs alone have difficulty grounding responses in a fixed product corpus. Motivated by these challenges, we propose a novel framework that introduces test-time scaling into conversational multimodal product retrieval. Our approach builds on a generative retriever, further augmented with a test-time reranking (TTR) mechanism that improves retrieval accuracy and better aligns results with evolving user intent throughout the dialogue. Experiments across multiple benchmarks show consistent improvements, with average gains of 14.5 points in MRR and 10.6 points in nDCG@1.
Authors: Nico Klar, Nizam Gifary, Felix P. G. Ziegler, Frank Sehnke, Anton Kaifel, Eric Price, Aamir Ahmad
Abstract: The urgent need for renewable energy expansion, particularly wind power, is hindered by conflicts with wildlife conservation. To address this, we developed BirdRecorder, an advanced AI-based anti-collision system to protect endangered birds, especially the red kite (Milvus milvus). Integrating robotics, telemetry, and high-performance AI algorithms, BirdRecorder aims to detect, track, and classify avian species within a range of 800 m to minimize bird-turbine collisions. BirdRecorder integrates advanced AI methods with optimized hardware and software architectures to enable real-time image processing. Leveraging Single Shot Detector (SSD) for detection, combined with specialized hardware acceleration and tracking algorithms, our system achieves high detection precision while maintaining the speed necessary for real-time decision-making. By combining these components, BirdRecorder outperforms existing approaches in both accuracy and efficiency. In this paper, we summarize results on field tests and performance of the BirdRecorder system. By bridging the gap between renewable energy expansion and wildlife conservation, BirdRecorder contributes to a more sustainable coexistence of technology and nature.
Authors: Syamantak Sarkar, Revoti P. Bora, Bhupender Kaushal, Sudhish N George, Kiran Raja
Abstract: Class Activation Maps (CAMs) are one of the important methods for visualizing regions used by deep learning models. Yet their robustness to different noise remains underexplored. In this work, we evaluate and report the resilience of various CAM methods for different noise perturbations across multiple architectures and datasets. By analyzing the influence of different noise types on CAM explanations, we assess the susceptibility to noise and the extent to which dataset characteristics may impact explanation stability. The findings highlight considerable variability in noise sensitivity for various CAMs. We propose a robustness metric for CAMs that captures two key properties: consistency and responsiveness. Consistency reflects the ability of CAMs to remain stable under input perturbations that do not alter the predicted class, while responsiveness measures the sensitivity of CAMs to changes in the prediction caused by such perturbations. The metric is evaluated empirically across models, different perturbations, and datasets along with complementary statistical tests to exemplify the applicability of our proposed approach.
Authors: Sara Ghazanfari, Wei-An Lin, Haitong Tian, Ersin Yumer
Abstract: Visually-guided image editing, where edits are conditioned on both visual cues and textual prompts, has emerged as a powerful paradigm for fine-grained, controllable content generation. Although recent generative models have shown remarkable capabilities, existing evaluations remain simple and insufficiently representative of real-world editing challenges. We present SpotEdit, a comprehensive benchmark designed to systematically assess visually-guided image editing methods across diverse diffusion, autoregressive, and hybrid generative models, uncovering substantial performance disparities. To address a critical yet underexplored challenge, our benchmark includes a dedicated component on hallucination, highlighting how leading models, such as GPT-4o, often hallucinate the existence of a visual cue and erroneously perform the editing task. Our code and benchmark are publicly released at https://github.com/SaraGhazanfari/SpotEdit.
Authors: Shuchismita Anwar, Sowmitra Das, Muhammad Iqbal Hossain, Jishnu Mahmud
Abstract: This study explores the challenge of improving multiclass image classification through quantum machine-learning techniques. It explores how the discarded qubit states of Noisy Intermediate-Scale Quantum (NISQ) quantum convolutional neural networks (QCNNs) can be leveraged alongside a classical classifier to improve classification performance. Current QCNNs discard qubit states after pooling; yet, unlike classical pooling, these qubits often remain entangled with the retained ones, meaning valuable correlated information is lost. We experiment with recycling this information and combining it with the conventional measurements from the retained qubits. Accordingly, we propose a hybrid quantum-classical architecture that couples a modified QCNN with fully connected classical layers. Two shallow fully connected (FC) heads separately process measurements from retained and discarded qubits, whose outputs are ensembled before a final classification layer. Joint optimisation with a classical cross-entropy loss allows both quantum and classical parameters to adapt coherently. The method outperforms comparable lightweight models on MNIST, Fashion-MNIST and OrganAMNIST. These results indicate that reusing discarded qubit information is a promising approach for future hybrid quantum-classical models and may extend to tasks beyond image classification.
Authors: Eric Alsmann, Martin Lange
Abstract: We analyse the complexity of the satisfiability problem ssmSAT for State Space Models (SSM), which asks whether an input sequence can lead the model to an accepting configuration. We find that ssmSAT is undecidable in general, reflecting the computational power of SSM. Motivated by practical settings, we identify two natural restrictions under which ssmSAT becomes decidable and establish corresponding complexity bounds. First, for SSM with bounded context length, ssmSAT is NP-complete when the input length is given in unary and in NEXPTIME (and PSPACE-hard) when the input length is given in binary. Second, for quantised SSM operating over fixed-width arithmetic, ssmSAT is PSPACE-complete resp. in EXPSPACE depending on the bit-width encoding. While these results hold for diagonal gated SSM we also establish complexity bounds for time-invariant SSM. Our results establish a first complexity landscape for formal reasoning in SSM and highlight fundamental limits and opportunities for the verification of SSM-based language models.
Authors: Bin Tan, Wangyao Ge, Yidi Wang, Xin Liu, Jeff Burtoft, Hao Fan, Hui Wang
Abstract: Modern app store recommender systems struggle with multiple-category apps, as traditional taxonomies fail to capture overlapping semantics, leading to suboptimal personalization. We propose PCR-CA (Parallel Codebook Representations with Contrastive Alignment), an end-to-end framework for improved CTR prediction. PCR-CA first extracts compact multimodal embeddings from app text, then introduces a Parallel Codebook VQ-AE module that learns discrete semantic representations across multiple codebooks in parallel -- unlike hierarchical residual quantization (RQ-VAE). This design enables independent encoding of diverse aspects (e.g., gameplay, art style), better modeling multiple-category semantics. To bridge semantic and collaborative signals, we employ a contrastive alignment loss at both the user and item levels, enhancing representation learning for long-tail items. Additionally, a dual-attention fusion mechanism combines ID-based and semantic features to capture user interests, especially for long-tail apps. Experiments on a large-scale dataset show PCR-CA achieves a +0.76% AUC improvement over strong baselines, with +2.15% AUC gains for long-tail apps. Online A/B testing further validates our approach, showing a +10.52% lift in CTR and a +16.30% improvement in CVR, demonstrating PCR-CA's effectiveness in real-world deployment. The new framework has now been fully deployed on the Microsoft Store.
Authors: Xiangxiang Wang, Xuanyu Wang, YiJia Luo, Yongbin Yu, Manping Fan, Jingtao Zhang, Liyong Ren
Abstract: This study proposes the dual technological innovation framework, including a cross-modal differ entiated quantization framework for vision-language models (VLMs) and a scene-aware vectorized memory multi-agent system for visually impaired assistance. The modular framework was developed implementing differentiated processing strategies, effectively reducing memory requirements from 38GB to 16GB while maintaining model performance. The multi-agent architecture combines scene classification, vectorized memory, and multimodal interaction, enabling persistent storage and efficient retrieval of scene memories. Through perception-memory-reasoning workflows, the system provides environmental information beyond the current view using historical memories. Experiments show the quantized 19B-parameter model only experiences a 2.05% performance drop on MMBench and maintains 63.7 accuracy on OCR-VQA (original: 64.9), outperforming smaller models with equivalent memory requirements like the Molmo-7B series. The system maintains response latency between 2.83-3.52 seconds from scene analysis to initial speech output, substantially faster than non-streaming methods. This research advances computational efficiency and assistive technology, offering visually impaired users comprehensive real-time assistance in scene perception, text recognition, and navigation.
Authors: Danielle Bednarski, Tim Roith
Abstract: These lecture notes evolve around mathematical concepts arising in inverse problems. We start by introducing inverse problems through examples such as differentiation, deconvolution, computed tomography and phase retrieval. This then leads us to the framework of well-posedness and first considerations regarding reconstruction and inversion approaches. The second chapter then first deals with classical regularization theory of inverse problems in Hilbert spaces. After introducing the pseudo-inverse, we review the concept of convergent regularization. Within this chapter we then proceed to ask the question of how to realize practical reconstruction algorithms. Here, we mainly focus on Tikhonov and sparsity promoting regularization in finite dimensional spaces. In the third chapter, we dive into modern deep-learning methods, which allow solving inverse problems in a data-dependent approach. The intersection between inverse problems and machine learning is a rapidly growing field and our exposition here restricts itself to a very limited selection of topics. Among them are learned regularization, fully-learned Bayesian estimation, post-processing strategies and plug-n-play methods.
Authors: Le Zhang, Fuping Wu, Arun Thirunavukarasu, Kevin Bronik, Thomas Nichols, Bartlomiej W. Papiez
Abstract: Large annotated datasets are vital for training segmentation models, but pixel-level labeling is time-consuming, error-prone, and often requires scarce expert annotators, especially in medical imaging. In contrast, coarse annotations are quicker, cheaper, and easier to produce, even by non-experts. In this paper, we propose to use coarse drawings from both positive (target) and negative (background) classes in the image, even with noisy pixels, to train a convolutional neural network (CNN) for semantic segmentation. We present a method for learning the true segmentation label distributions from purely noisy coarse annotations using two coupled CNNs. The separation of the two CNNs is achieved by high fidelity with the characters of the noisy training annotations. We propose to add a complementary label learning that encourages estimating negative label distribution. To illustrate the properties of our method, we first use a toy segmentation dataset based on MNIST. We then present the quantitative results of experiments using publicly available datasets: Cityscapes dataset for multi-class segmentation, and retinal images for medical applications. In all experiments, our method outperforms state-of-the-art methods, particularly in the cases where the ratio of coarse annotations is small compared to the given dense annotations.
Authors: Kushal Raj Bhandari, Pin-Yu Chen, Jianxi Gao
Abstract: Large Language Models (LLMs) have reshaped our world with significant advancements in science, engineering, and society through applications ranging from scientific discoveries and medical diagnostics to Chatbots. Despite their ubiquity and utility, the underlying mechanisms of LLM remain concealed within billions of parameters and complex structures, making their inner architecture and cognitive processes challenging to comprehend. We address this gap by adopting approaches to understanding emerging cognition in biology and developing a network-based framework that links cognitive skills, LLM architectures, and datasets, ushering in a paradigm shift in foundation model analysis. The skill distribution in the module communities demonstrates that while LLMs do not strictly parallel the focalized specialization observed in specific biological systems, they exhibit unique communities of modules whose emergent skill patterns partially mirror the distributed yet interconnected cognitive organization seen in avian and small mammalian brains. Our numerical results highlight a key divergence from biological systems to LLMs, where skill acquisition benefits substantially from dynamic, cross-regional interactions and neural plasticity. By integrating cognitive science principles with machine learning, our framework provides new insights into LLM interpretability and suggests that effective fine-tuning strategies should leverage distributed learning dynamics rather than rigid modular interventions.
Authors: Ritvik Chaturvedi
Abstract: This project implements a ResNet-based pipeline for land use and land cover (LULC) classification on Sentinel-2 imagery, benchmarked across three heterogeneous GPUs. The workflow automates data acquisition, geospatial preprocessing, tiling, model training, and visualization, and is fully containerized for reproducibility. Performance evaluation reveals up to a 2x training speed-up on an NVIDIA RTX 3060 and a Tesla T4 compared to the Apple M3 Pro baseline, while maintaining high classification accuracy on the EuroSAT dataset. These results demonstrate the feasibility of deploying deep learning LULC models on consumer and free cloud GPUs for scalable geospatial analytics.
Authors: Igor Tona Peres, Otavio T. Ranzani, Leonardo S. L. Bastos, Silvio Hamacher, Tom Edinburgh, Esteban Garcia-Gallo, Fernando Augusto Bozza
Abstract: Background. Dengue outbreaks are a major public health issue, with Brazil reporting 71% of global cases in 2024. Purpose. This study aims to describe the profile of severe dengue patients admitted to Brazilian Intensive Care units (ICUs) (2012-2024), assess trends over time, describe new onset complications while in ICU and determine the risk factors at admission to develop complications during ICU stay. Methods. We performed a prospective study of dengue patients from 253 ICUs across 56 hospitals. We used descriptive statistics to describe the dengue ICU population, logistic regression to identify risk factors for complications during the ICU stay, and a machine learning framework to predict the risk of evolving to complications. Visualisations were generated using ISARIC VERTEX. Results. Of 11,047 admissions, 1,117 admissions (10.1%) evolved to complications, including non-invasive (437 admissions) and invasive ventilation (166), vasopressor (364), blood transfusion (353) and renal replacement therapy (103). Age>80 (OR: 3.10, 95% CI: 2.02-4.92), chronic kidney disease (OR: 2.94, 2.22-3.89), liver cirrhosis (OR: 3.65, 1.82-7.04), low platelets (<50,000 cells/mm3; OR: OR: 2.25, 1.89-2.68), and high leukocytes (>7,000 cells/mm3; OR: 2.47, 2.02-3.03) were significant risk factors for complications. A machine learning tool for predicting complications was proposed, showing accurate discrimination and calibration. Conclusion. We described a large cohort of dengue patients admitted to ICUs and identified key risk factors for severe dengue complications, such as advanced age, presence of comorbidities, higher level of leukocytes and lower level of platelets. The proposed prediction tool can be used for early identification and targeted interventions to improve outcomes in dengue-endemic regions.
Authors: Vsevolod Viliuga, Leif Seute, Nicolas Wolf, Simon Wagner, Arne Elofsson, Jan St\"uhmer, Frauke Gr\"ater
Abstract: Recent advances in geometric deep learning and generative modeling have enabled the design of novel proteins with a wide range of desired properties. However, current state-of-the-art approaches are typically restricted to generating proteins with only static target properties, such as motifs and symmetries. In this work, we take a step towards overcoming this limitation by proposing a framework to condition structure generation on flexibility, which is crucial for key functionalities such as catalysis or molecular recognition. We first introduce BackFlip, an equivariant neural network for predicting per-residue flexibility from an input backbone structure. Relying on BackFlip, we propose FliPS, an SE(3)-equivariant conditional flow matching model that solves the inverse problem, that is, generating backbones that display a target flexibility profile. In our experiments, we show that FliPS is able to generate novel and diverse protein backbones with the desired flexibility, verified by Molecular Dynamics (MD) simulations. FliPS and BackFlip are available at https://github.com/graeter-group/flips .
Authors: Ran Yan, Youhe Jiang, Binhang Yuan
Abstract: Recent progress in sparse attention mechanisms has demonstrated strong potential for reducing the computational cost of long-context training and inference in large language models (LLMs). Native Sparse Attention (NSA), a state-of-the-art approach, introduces natively trainable, hardware-aligned sparse attention that delivers substantial system-level performance gains while maintaining accuracy comparable to full attention. However, the kernel implementation of NSA relies on a query-grouping strategy that is efficient only with large Grouped Query Attention (GQA) sizes, whereas modern LLMs typically adopt much smaller GQA groups, which limits the applicability of this sparse algorithmic advance. In this work, we propose Flash Sparse Attention (FSA), which includes an alternative kernel design that enables efficient NSA computation across a wide range of popular LLMs with varied smaller GQA group sizes on modern GPUs. Compared to vanilla NSA kernel implementation, our empirical evaluation demonstrates that FSA achieves (i) up to 3.5$\times$ and on average 1.6$\times$ kernel-level latency reduction, (ii) up to 1.25$\times$ and 1.09$\times$ on average end-to-end training speedup on state-of-the-art LLMs, and (iii) up to 1.36$\times$ and 1.11$\times$ on average end-to-end prefill speedup on state-of-the-art LLMs. The source code is open-sourced and publicly available at https://github.com/Relaxed-System-Lab/Flash-Sparse-Attention.
URLs: https://github.com/Relaxed-System-Lab/Flash-Sparse-Attention.
Authors: Menglin Yang, Min Zhou, Tong Zhang, Jiahong Liu, Zhihao Li, Lujia Pan, Hui Xiong, Irwin King
Abstract: Graph representation learning in Euclidean space, despite its widespread adoption and proven utility in many domains, often struggles to effectively capture the inherent hierarchical and complex relational structures prevalent in real-world data, particularly for datasets exhibiting a highly non-Euclidean latent anatomy or power-law distributions. Hyperbolic geometry, with its constant negative curvature and exponential growth property, naturally accommodates such structures, offering a promising alternative for learning rich graph representations. This survey paper provides a comprehensive review of the rapidly evolving field of Hyperbolic Graph Learning (HGL). We systematically categorize and analyze existing methods broadly dividing them into (1) hyperbolic graph embedding-based techniques, (2) graph neural network-based hyperbolic models, and (3) emerging paradigms. Beyond methodologies, we extensively discuss diverse applications of HGL across multiple domains, including recommender systems, knowledge graphs, bioinformatics, and other relevant scenarios, demonstrating the broad applicability and effectiveness of hyperbolic geometry in real-world graph learning tasks. Most importantly, we identify several key challenges that serve as directions for advancing HGL, including handling complex data structures, developing geometry-aware learning objectives, ensuring trustworthy and scalable implementations, and integrating with foundation models, e.g., large language models. We highlight promising research opportunities in this exciting interdisciplinary area. A comprehensive repository can be found at https://github.com/digailab/awesome-hyperbolic-graph-learning.
URLs: https://github.com/digailab/awesome-hyperbolic-graph-learning.
Authors: Sebastian Maldonado, Carla Vairetti, Ignacio Figueroa
Abstract: As with any task, the process of building machine learning models can benefit from prior experience. Meta-learning for classifier selection leverages knowledge about the characteristics of different datasets and/or the past performance of machine learning techniques to inform better decisions in the current modeling process. Traditional meta-learning approaches first collect metadata that describe this prior experience and then use it as input for an algorithm selection model. In this paper, however, a one-step scheme is proposed in which convolutional neural networks are trained directly on tabular datasets for binary classification. The aim is to learn the underlying structure of the data without the need to explicitly identify meta-features. Experiments with simulated datasets show that the proposed approach achieves near-perfect performance in identifying both linear and nonlinear patterns, outperforming the conventional two-step method based on meta-features. The method is further applied to real-world datasets, providing recommendations on the most suitable classifiers based on the data's inherent structure.
Authors: Jianhao Ma, Rui Ray Chen, Yinghui He, Salar Fattahi, Wei Hu
Abstract: In this paper, we study the problem of sparse mean estimation under adversarial corruptions, where the goal is to estimate the $k$-sparse mean of a heavy-tailed distribution from samples contaminated by adversarial noise. Existing methods face two key limitations: they require prior knowledge of the sparsity level $k$ and scale poorly to high-dimensional settings. We propose a simple and scalable estimator that addresses both challenges. Specifically, it learns the $k$-sparse mean without knowing $k$ in advance and operates in near-linear time and memory with respect to the ambient dimension. Under a moderate signal-to-noise ratio, our method achieves the optimal statistical rate, matching the information-theoretic lower bound. Extensive simulations corroborate our theoretical guarantees. At the heart of our approach is an incremental learning phenomenon: we show that a basic subgradient method applied to a nonconvex two-layer formulation with an $\ell_1$-loss can incrementally learn the $k$ nonzero components of the true mean while suppressing the rest. More broadly, our work is the first to reveal the incremental learning phenomenon of the subgradient method in the presence of heavy-tailed distributions and adversarial corruption.
Authors: Shengbo Wang, Nian Si, Jose Blanchet, Zhengyuan Zhou
Abstract: Motivated by the need for a robust policy in the face of environment shifts between training and deployment, we contribute to the theoretical foundation of distributionally robust reinforcement learning (DRRL). This is accomplished through a comprehensive modeling framework centered around robust Markov decision processes (RMDPs). This framework obliges the decision maker to choose an optimal policy under the worst-case distributional shift orchestrated by an adversary. By unifying and extending existing formulations, we rigorously construct RMDPs that embrace various modeling attributes for both the decision maker and the adversary. These attributes include the structure of information availability-covering history-dependent, Markov, and Markov time-homogeneous dynamics-as well as constraints on the shifts induced by the adversary, with a focus on SA- and S-rectangularity. Within this RMDP framework, we investigate conditions for the existence or absence of the dynamic programming principle (DPP). From an algorithmic standpoint, the existence of DPP holds significant implications, as the vast majority of existing data and computationally efficient DRRL algorithms are reliant on the DPP. To investigate its existence, we systematically analyze various combinations of controller and adversary attributes, presenting streamlined proofs based on a unified methodology. We then construct counterexamples for settings where a fully general DPP fails to hold and establish asymptotically optimal history-dependent policies for key scenarios where the DPP is absent.
Authors: Sagar Shrestha, Xiao Fu
Abstract: Unsupervised domain translation (UDT) aims to find functions that convert samples from one domain (e.g., sketches) to another domain (e.g., photos) without changing the high-level semantic meaning (also referred to as ``content''). The translation functions are often sought by probability distribution matching of the transformed source domain and target domain. CycleGAN stands as arguably the most representative approach among this line of work. However, it was noticed in the literature that CycleGAN and variants could fail to identify the desired translation functions and produce content-misaligned translations. This limitation arises due to the presence of multiple translation functions -- referred to as ``measure-preserving automorphism" (MPA) -- in the solution space of the learning criteria. Despite awareness of such identifiability issues, solutions have remained elusive. This study delves into the core identifiability inquiry and introduces an MPA elimination theory. Our analysis shows that MPA is unlikely to exist, if multiple pairs of diverse cross-domain conditional distributions are matched by the learning function. Our theory leads to a UDT learner using distribution matching over auxiliary variable-induced subsets of the domains -- other than over the entire data domains as in the classical approaches. The proposed framework is the first to rigorously establish translation identifiability under reasonable UDT settings, to our best knowledge. Experiments corroborate with our theoretical claims.
Authors: Maryam Ahang, Todd Charter, Mostafa Abbasi, Maziyar Khadivi, Oluwaseyi Ogunfowora, Homayoun Najjaran
Abstract: Condition monitoring is essential for ensuring the safety, reliability, and efficiency of modern industrial systems. With the increasing complexity of industrial processes, artificial intelligence (AI) has emerged as a powerful tool for fault detection and diagnosis, attracting growing interest from both academia and industry. This paper provides a comprehensive overview of intelligent condition monitoring methods, with a particular emphasis on chemical plants and the widely used Tennessee Eastman Process (TEP) benchmark. State-of-the-art machine learning (ML) and deep learning (DL) algorithms are reviewed, highlighting their strengths, limitations, and applicability to industrial fault detection and diagnosis. Special attention is given to key challenges, including imbalanced and unlabeled data, and to strategies by which models can address these issues. Furthermore, comparative analyses of algorithm performance are presented to guide method selection in practical scenarios. This survey is intended to benefit both newcomers and experienced researchers by consolidating fundamental concepts, summarizing recent advances, and outlining open challenges and promising directions for intelligent condition monitoring in industrial plants.
Authors: Emanuele Zangrando, Piero Deidda, Simone Brugiapaglia, Nicola Guglielmi, Francesco Tudisco
Abstract: Recent work in deep learning has shown strong empirical and theoretical evidence of an implicit low-rank bias: weight matrices in deep networks tend to be approximately low-rank. Moreover, removing relatively small singular values during training, or from available trained models, may significantly reduce model size while maintaining or even improving model performance. However, the majority of the theoretical investigations around low-rank bias in neural networks deal with oversimplified models, often not taking into account the impact of nonlinearity. In this work, we first of all quantify a link between the phenomenon of deep neural collapse and the emergence of low-rank weight matrices for a general class of feedforward networks with nonlinear activation. In addition, for the general class of nonlinear feedforward and residual networks, we prove the global optimality of deep neural collapsed configurations and the practical absence of a loss barrier between interpolating minima and globally optimal points, offering a possible explanation for its common occurrence. As a byproduct, our theory also allows us to forecast the final global structure of singular values before training. Our theoretical findings are supported by a range of experimental evaluations illustrating the phenomenon.
Authors: Zihang Xiang, Tianhao Wang, Chenglong Wang, Di Wang
Abstract: We study the application of differential privacy in hyper-parameter tuning, a crucial process in machine learning involving selecting the best hyper-parameter from several candidates. Unlike many private learning algorithms, including the prevalent DP-SGD, the privacy implications of tuning remain insufficiently understood or often totally ignored. Recent works propose a generic private selection solution for the tuning process, yet a fundamental question persists: is this privacy bound tight? This paper provides an in-depth examination of this question. Initially, we provide studies affirming the current privacy analysis for private selection is indeed tight in general. However, when we specifically study the hyper-parameter tuning problem in a white-box setting, such tightness no longer holds. This is first demonstrated by applying privacy audit on the tuning process. Our findings underscore a substantial gap between current theoretical privacy bound and the empirical bound derived even under strong audit setups. This gap motivates our subsequent investigations. Our further study provides improved privacy results for private hyper-parameter tuning due to its distinct properties. Our results demonstrate broader applicability compared to prior analyses, which are limited to specific parameter configurations.
Authors: Bishwamittra Ghosh, Debabrota Basu, Fu Huazhu, Wang Yuan, Renuga Kanagavelu, Jiang Jin Peng, Liu Yong, Goh Siow Mong Rick, Wei Qingsong
Abstract: Federated Learning (FL) is a collaborative machine learning (ML) approach, where multiple clients participate in training an ML model without exposing their private data. Fair and accurate assessment of client contributions facilitates incentive allocation in FL and encourages diverse clients to participate in a unified model training. Existing methods for contribution assessment adopts a co-operative game-theoretic concept, called Shapley value, but under restricted assumptions, e.g., all clients' participating in all epochs or at least in one epoch of FL. We propose a history-aware client contribution assessment framework, called FLContrib, where client-participation is dynamic, i.e., a subset of clients participates in each epoch. The theoretical underpinning of FLContrib is based on the Markovian training process of FL. Under this setting, we directly apply the linearity property of Shapley value and compute a historical timeline of client contributions. Considering the possibility of a limited computational budget, we propose a two-sided fairness criteria to schedule Shapley value computation in a subset of epochs. Empirically, FLContrib is efficient and consistently accurate in estimating contribution across multiple utility functions. As a practical application, we apply FLContrib to detect dishonest clients in FL based on historical Shaplee values.
Authors: Nicholas Zolman, Christian Lagemann, Urban Fasel, J. Nathan Kutz, Steven L. Brunton
Abstract: Deep reinforcement learning (DRL) has shown significant promise for uncovering sophisticated control policies that interact in complex environments, such as stabilizing a tokamak fusion reactor or minimizing the drag force on an object in a fluid flow. However, DRL requires an abundance of training examples and may become prohibitively expensive for many applications. In addition, the reliance on deep neural networks often results in an uninterpretable, black-box policy that may be too computationally expensive to use with certain embedded systems. Recent advances in sparse dictionary learning, such as the sparse identification of nonlinear dynamics (SINDy), have shown promise for creating efficient and interpretable data-driven models in the low-data regime. In this work we introduce SINDy-RL, a unifying framework for combining SINDy and DRL to create efficient, interpretable, and trustworthy representations of the dynamics model, reward function, and control policy. We demonstrate the effectiveness of our approaches on benchmark control environments and flow control problems, including gust mitigation on a 3D NACA 0012 airfoil at $Re=1000$. SINDy-RL achieves comparable performance to modern DRL algorithms using significantly fewer interactions in the environment and results in an interpretable control policy orders of magnitude smaller than a DRL policy.
Authors: Moshe Eliasof, Eldad Haber
Abstract: We investigate a link between Graph Neural Networks (GNNs) and Quadratic Unconstrained Binary Optimization (QUBO) problems, laying the groundwork for GNNs to approximate solutions for these computationally challenging tasks. By analyzing the sensitivity of QUBO formulations, we frame the solution of QUBO problems as a heterophilic node classification task. We then propose QUBO-GNN, an architecture that integrates graph representation learning techniques with QUBO-aware features to approximate solutions efficiently. Additionally, we introduce a self-supervised data generation mechanism to enable efficient and scalable training data acquisition even for large-scale QUBO instances. Experimental evaluations of QUBO-GNN across diverse QUBO problem sizes demonstrate its superior performance compared to exhaustive search and heuristic methods. Finally, we discuss open challenges in the emerging intersection between QUBO optimization and GNN-based learning.
Authors: Harshit Dhankhar, Kshitij Mishra, Tejas Bodas
Abstract: In the realm of multi-arm bandit problems, the Gittins index policy is known to be optimal in maximizing the expected total discounted reward obtained from pulling the Markovian arms. In most realistic scenarios however, the Markovian state transition probabilities are unknown and therefore the Gittins indices cannot be computed. One can then resort to reinforcement learning (RL) algorithms that explore the state space to learn these indices while exploiting to maximize the reward collected. In this work, we propose tabular (QGI) and Deep RL (DGN) algorithms for learning the Gittins index that are based on the retirement formulation for the multi-arm bandit problem. When compared with existing RL algorithms that learn the Gittins index, our algorithms have a lower run time, require less storage space (small Q-table size in QGI and smaller replay buffer in DGN), and illustrate better empirical convergence to the Gittins index. This makes our algorithm well suited for problems with large state spaces and is a viable alternative to existing methods. As a key application, we demonstrate the use of our algorithms in minimizing the mean flowtime in a job scheduling problem when jobs are available in batches and have an unknown service time distribution.
Authors: Kabir Dabholkar, Omri Barak
Abstract: Latent variable models serve as powerful tools to infer underlying dynamics from observed neural activity. Ideally, the inferred dynamics should align with true ones. However, due to the absence of ground truth data, prediction benchmarks are often employed as proxies. One widely-used method, $\textit{co-smoothing}$, involves jointly estimating latent variables and predicting observations along held-out channels to assess model performance. In this study, we reveal the limitations of the co-smoothing prediction framework and propose a remedy. Using a student-teacher setup, we demonstrate that models with high co-smoothing can have arbitrary extraneous dynamics in their latent representations. To address this, we introduce a secondary metric -- $\textit{few-shot co-smoothing}$, performing regression from the latent variables to held-out neurons in the data using fewer trials. Our results indicate that among models with near-optimal co-smoothing, those with extraneous dynamics underperform in the few-shot co-smoothing compared to `minimal' models that are devoid of such dynamics. We provide analytical insights into the origin of this phenomenon and further validate our findings on four standard neural datasets using a state-of-the-art method: STNDT. In the absence of ground truth, we suggest a novel measure to validate our approach. By cross-decoding the latent variables of all model pairs with high co-smoothing, we identify models with minimal extraneous dynamics. We find a correlation between few-shot co-smoothing performance and this new measure. In summary, we present a novel prediction metric designed to yield latent variables that more accurately reflect the ground truth, offering a significant improvement for latent dynamics inference.
Authors: Xuefeng Gao, Lingfei Li, Xun Yu Zhou
Abstract: We study continuous-time reinforcement learning (RL) for stochastic control in which system dynamics are governed by jump-diffusion processes. We formulate an entropy-regularized exploratory control problem with stochastic policies to capture the exploration--exploitation balance essential for RL. Unlike the pure diffusion case initially studied by Wang et al. (2020), the derivation of the exploratory dynamics under jump-diffusions calls for a careful formulation of the jump part. Through a theoretical analysis, we find that one can simply use the same policy evaluation and $q$-learning algorithms in Jia and Zhou (2022a, 2023), originally developed for controlled diffusions, without needing to check a priori whether the underlying data come from a pure diffusion or a jump-diffusion. However, we show that the presence of jumps ought to affect parameterizations of actors and critics in general. We investigate as an application the mean--variance portfolio selection problem with stock price modelled as a jump-diffusion, and show that both RL algorithms and parameterizations are invariant with respect to jumps. Finally, we present a detailed study on applying the general theory to option hedging.
Authors: Federico Errica, Giuseppe Siracusano, Davide Sanvito, Roberto Bifulco
Abstract: Large Language Models (LLMs) changed the way we design and interact with software systems. Their ability to process and extract information from text has drastically improved productivity in a number of routine tasks. Developers that want to include these models in their software stack, however, face a dreadful challenge: debugging LLMs' inconsistent behavior across minor variations of the prompt. We therefore introduce two metrics for classification tasks, namely sensitivity and consistency, which are complementary to task performance. First, sensitivity measures changes of predictions across rephrasings of the prompt, and does not require access to ground truth labels. Instead, consistency measures how predictions vary across rephrasings for elements of the same class. We perform an empirical comparison of these metrics on text classification tasks, using them as guideline for understanding failure modes of the LLM. Our hope is that sensitivity and consistency will be helpful to guide prompt engineering and obtain LLMs that balance robustness with performance.
Authors: Menglin Yang, Harshit Verma, Delvin Ce Zhang, Jiahong Liu, Irwin King, Rex Ying
Abstract: Hyperbolic geometry have shown significant potential in modeling complex structured data, particularly those with underlying tree-like and hierarchical structures. Despite the impressive performance of various hyperbolic neural networks across numerous domains, research on adapting the Transformer to hyperbolic space remains limited. Previous attempts have mainly focused on modifying self-attention modules in the Transformer. However, these efforts have fallen short of developing a complete hyperbolic Transformer. This stems primarily from: (i) the absence of well-defined modules in hyperbolic space, including linear transformation layers, LayerNorm layers, activation functions, dropout operations, etc. (ii) the quadratic time complexity of the existing hyperbolic self-attention module w.r.t the number of input tokens, which hinders its scalability. To address these challenges, we propose, Hypformer, a novel hyperbolic Transformer based on the Lorentz model of hyperbolic geometry. In Hypformer, we introduce two foundational blocks that define the essential modules of the Transformer in hyperbolic space. Furthermore, we develop a linear self-attention mechanism in hyperbolic space, enabling hyperbolic Transformer to process billion-scale graph data and long-sequence inputs for the first time. Our experimental results confirm the effectiveness and efficiency of Hypformer across various datasets, demonstrating its potential as an effective and scalable solution for large-scale data representation and large models.
Authors: Nasir Ahmad
Abstract: Herein the topics of (natural) gradient descent, data decorrelation, and approximate methods for backpropagation are brought into a common discussion. Natural gradient descent illuminates how gradient vectors, pointing at directions of steepest descent, can be improved by considering the local curvature of loss landscapes. We extend this perspective and show that to fully solve the problem illuminated by natural gradients in neural networks, one must recognise that correlations in the data at any linear transformation, including node responses at every layer of a neural network, cause a non-orthonormal relationship between the model's parameters. To solve this requires a method for decorrelating inputs at each individual layer of a neural network. We describe a range of methods which have been proposed for decorrelation and whitening of node output, and expand on these to provide a novel method specifically useful for distributed computing and computational neuroscience. Implementing decorrelation within multi-layer neural networks, we can show that not only is training via backpropagation sped up significantly but also existing approximations of backpropagation, which have failed catastrophically in the past, benefit significantly in their accuracy and convergence speed. This has the potential to provide a route forward for approximate gradient descent methods which have previously been discarded, training approaches for analogue and neuromorphic hardware, and potentially insights as to the efficacy and utility of decorrelation processes in the brain.
Authors: Jiaxing Miao, Liang Hu, Qi Zhang, Longbing Cao
Abstract: Graph data in real-world scenarios undergo rapid and frequent changes, making it challenging for existing graph models to effectively handle the continuous influx of new data and accommodate data withdrawal requests. The approach to frequently retraining graph models is resource intensive and impractical. To address this pressing challenge, this paper introduces a new concept of graph memory learning. Its core idea is to enable a graph model to selectively remember new knowledge but forget old knowledge. Building on this approach, the paper presents a novel graph memory learning framework - Brain-inspired Graph Memory Learning (BGML), inspired by brain network dynamics and function-structure coupling strategies. BGML incorporates a multi-granular hierarchical progressive learning mechanism rooted in feature graph grain learning to mitigate potential conflict between memorization and forgetting in graph memory learning. This mechanism allows for a comprehensive and multi-level perception of local details within evolving graphs. In addition, to tackle the issue of unreliable structures in newly added incremental information, the paper introduces an information self-assessment ownership mechanism. This mechanism not only facilitates the propagation of incremental information within the model but also effectively preserves the integrity of past experiences. We design five types of graph memory learning tasks: regular, memory, unlearning, data-incremental, and class-incremental to evaluate BGML. Its excellent performance is confirmed through extensive experiments on multiple real-world node classification datasets.
Authors: Saeed Mohammadi Dashtaki, Reza Mohammadi Dashtaki, Mehdi Hosseini Chagahi, Behzad Moshiri, Md. Jalil Piran
Abstract: Predicting cryptocurrency price trends remains a major challenge due to the volatility and complexity of digital asset markets. Artificial intelligence (AI) has emerged as a powerful tool to address this problem. This study proposes a multisource fusion framework that integrates quantitative financial indicators, such as historical prices and technical indicators, with qualitative sentiment signals derived from X (formerly Twitter). Sentiment analysis is performed using Financial Bidirectional Encoder Representations from Transformers (FinBERT), a domain-specific BERT-based model optimized for financial text, while sequential dependencies are captured through a Bidirectional Long Short-Term Memory (BiLSTM) network. Experimental results on a large-scale Bitcoin dataset demonstrate that the proposed approach substantially outperforms single-source models, achieving an accuracy of approximately 96.8\%. The findings underscore the importance of incorporating real-time social sentiment alongside traditional indicators, thereby enhancing predictive accuracy and supporting more informed investment decisions.
Authors: Md Rejwanur Rahman, Adrian Rodriguez-Marek, Nina Stark, Grace Massey, Carl Friedrichs, Kelly M. Dorgan
Abstract: The geotechnical evaluation of seabed sediments is important for engineering projects and naval applications, offering valuable insights into sediment properties, behavior, and strength. Obtaining high-quality seabed samples can be a challenging task, making in situ testing an essential part of site characterization. Free-fall penetrometers (FFPs) are robust tools for rapidly profiling seabed surface sediments, even in energetic nearshore or estuarine conditions and shallow as well as deep depths. Although methods for interpretation of traditional offshore cone penetration testing (CPT) data are well-established, their adaptation to FFP data is still an area of research. This study introduces an innovative approach that utilizes machine learning algorithms to create a sediment behavior classification system based on portable free- fall penetrometer (PFFP) data. The proposed model leverages PFFP measurements obtained from multiple locations, such as Sequim Bay (Washington), the Potomac River, and the York River (Virginia). The results show 91.1% accuracy in the class prediction, with the classes representing cohesionless sediment with little to no plasticity (Class 1), cohesionless sediment with some plasticity (Class 2), cohesive sediment with low plasticity (Class 3), and cohesive sediment with high plasticity (Class 4). The model prediction not only predicts classes but also yields an estimate of inherent uncertainty associated with the prediction, which can provide valuable insight into different sediment behaviors. Lower uncertainties are more common, but they can increase significantly depending on variations in sediment composition, environmental conditions, and operational techniques. By quantifying uncertainty, the model offers a more comprehensive and informed approach to sediment classification
Authors: Eshika Saxena, Alberto Alfarano, Fran\c{c}ois Charton, Zeyuan Allen-Zhu, Emily Wenger, Kristin Lauter
Abstract: Recent work showed that ML-based attacks on Learning with Errors (LWE), a hard problem used in post-quantum cryptography, outperform classical algebraic attacks in certain settings. Although promising, ML attacks struggle to scale to more complex LWE settings. Prior work connected this issue to the difficulty of training ML models to do modular arithmetic, a core feature of the LWE problem. To address this, we develop techniques that significantly boost the performance of ML models on modular arithmetic tasks, enabling the models to sum up to $N=128$ elements modulo $q \le 974269$. Our core innovation is the use of custom training data distributions and a carefully designed loss function that better represents the problem structure. We apply an initial proof of concept of our techniques to LWE specifically and find that they allow recovery of 2x harder secrets than prior work. Our techniques also help ML models learn other well-studied problems better, including copy, associative recall, and parity, motivating further study.
Authors: Qidong Yang, Jonathan Giezendanner, Daniel Salles Civitarese, Johannes Jakubik, Eric Schmitt, Anirban Chandra, Jeremy Vila, Detlef Hohl, Chris Hill, Campbell Watson, Sherrie Wang
Abstract: Urgent applications like wildfire management and renewable energy generation require precise, localized weather forecasts near the Earth's surface. However, forecasts produced by machine learning models or numerical weather prediction systems are typically generated on large-scale regular grids, where direct downscaling fails to capture fine-grained, near-surface weather patterns. In this work, we propose a multi-modal transformer model trained end-to-end to downscale gridded forecasts to off-grid locations of interest. Our model directly combines local historical weather observations (e.g., wind, temperature, dewpoint) with gridded forecasts to produce locally accurate predictions at various lead times. Multiple data modalities are collected and concatenated at station-level locations, treated as a token at each station. Using self-attention, the token corresponding to the target location aggregates information from its neighboring tokens. Experiments using weather stations across the Northeastern United States show that our model outperforms a range of data-driven and non-data-driven off-grid forecasting methods. They also reveal that direct input of station data provides a phase shift in local weather forecasting accuracy, reducing the prediction error by up to 80% compared to pure gridded data based models. This approach demonstrates how to bridge the gap between large-scale weather models and locally accurate forecasts to support high-stakes, location-sensitive decision-making.
Authors: Xinrui He, Yikun Ban, Jiaru Zou, Tianxin Wei, Curtiss B. Cook, Jingrui He
Abstract: Missing data imputation is a critical challenge in various domains, such as healthcare and finance, where data completeness is vital for accurate analysis. Large language models (LLMs), trained on vast corpora, have shown strong potential in data generation, making them a promising tool for data imputation. However, challenges persist in designing effective prompts for a finetuning-free process and in mitigating biases and uncertainty in LLM outputs. To address these issues, we propose a novel framework, LLM-Forest, which introduces a "forest" of few-shot prompt learning LLM "trees" with their outputs aggregated via confidence-based weighted voting based on LLM self-assessment, inspired by the ensemble learning (Random Forest). This framework is established on a new concept of bipartite information graphs to identify high-quality relevant neighboring entries with both feature and value granularity. Extensive experiments on 9 real-world datasets demonstrate the effectiveness and efficiency of LLM-Forest.
Authors: Jingge Xiao, Yile Chen, Gao Cong, Wolfgang Nejdl, Simon Gottschalk
Abstract: Forecasting time series with irregular temporal structures remains challenging for universal pre-trained models. Existing approaches often assume regular sampling or depend heavily on imputation, limiting their applicability in real-world scenarios where irregularities are prevalent due to diverse sensing devices and recording practices. We introduce FlexTSF, a flexible forecasting model specifically designed for time series data with variable temporal regularities. At its foundation lies the IVP Patcher, a continuous-time patching module leveraging Initial Value Problems (IVPs) to inherently support uneven time intervals, variable sequence lengths, and missing values. FlexTSF employs a decoder-only architecture that integrates normalized timestamp inputs and domain-specific statistics through a specialized causal self-attention mechanism, enabling adaptability across domains. Extensive experiments on 16 datasets demonstrate FlexTSF's effectiveness, significantly outperforming existing models in classic forecasting scenarios, zero-shot generalization, and low-resource fine-tuning conditions. Ablation studies confirm the contributions of each design component and the advantage of not relying on predefined fixed patch lengths.
Authors: Vincent Mai, Quang Hung Pham, Arthur Favrel, Jean-Philippe Gauthier, Martin Gagnon
Abstract: Hydro-generating units (HGUs) play a crucial role in integrating intermittent renewable energy sources into the power grid due to their flexible operational capabilities. This evolving role has led to an increase in transient events, such as startups, which impose significant stresses on turbines, leading to increased turbine fatigue and a reduced operational lifespan. Consequently, optimizing startup sequences to minimize stresses is vital for hydropower utilities. However, this task is challenging, as stress measurements on prototypes can be expensive and time-consuming. To tackle this challenge, we propose an innovative automated approach to optimize the startup parameters of HGUs with a limited budget of measured startup sequences. Our method combines active learning and black-box optimization techniques, utilizing virtual strain sensors and dynamic simulations of HGUs. This approach was tested in real-time during an on-site measurement campaign on an instrumented Francis turbine prototype. The results demonstrate that our algorithm successfully identified an optimal startup sequence using only seven measured sequences. It achieves a remarkable 42% reduction in the maximum strain cycle amplitude compared to the standard startup sequence. This study paves the way for more efficient HGU startup optimization, potentially extending their operational lifespans.
Authors: Ruofan Jia, Weiying Xie, Jie Lei, Jitao Ma, Haonan Qin, Leyuan Fang
Abstract: While large pre-trained models have achieved impressive performance across AI tasks, their deployment in privacy-sensitive and distributed environments remains challenging. Federated learning (FL) offers a viable solution by enabling decentralized fine-tuning without data sharing, but real-world applications face significant obstacles due to heterogeneous client resources in compute and memory. To address this, we propose HeteroTune, a novel federated fine-tuning paradigm for large, heterogeneous models operating under limited communication and computation budgets. The core of our method lies in a novel architecture, DeMA (Dense Mixture of Adapters), which enables flexible and efficient aggregation of heterogeneous models by preserving their full representational capacity while facilitating seamless cross-model knowledge fusion. We further introduce CMGA (Cross-Model Gradient Alignment), a lightweight yet effective mechanism that enhances training stability by harmonizing gradient directions across heterogeneous client models during aggregation, mitigating update conflicts and promoting more consistent convergence in federated settings. We provide both theoretical analysis and empirical evidence showing that HeteroTune achieves state-of-the-art performance and efficiency across diverse tasks and model architectures. For example, on LLaMA models, it reduces communication overhead by 99.5%, cuts peak memory usage by ~50%, and improves performance by 4.61%.
Authors: Tomer Borreda, Daniel Freedman, Or Litany
Abstract: We present ReHub, a novel graph transformer architecture that achieves linear complexity through an efficient reassignment technique between nodes and virtual nodes. Graph transformers have become increasingly important in graph learning for their ability to utilize long-range node communication explicitly, addressing limitations such as oversmoothing and oversquashing found in message-passing graph networks. However, their dense attention mechanism scales quadratically with the number of nodes, limiting their applicability to large-scale graphs. ReHub draws inspiration from the airline industry's hub-and-spoke model, where flights are assigned to optimize operational efficiency. In our approach, graph nodes (spokes) are dynamically reassigned to a fixed number of virtual nodes (hubs) at each model layer. Recent work, Neural Atoms (Li et al., 2024), has demonstrated impressive and consistent improvements over GNN baselines by utilizing such virtual nodes; their findings suggest that the number of hubs strongly influences performance. However, increasing the number of hubs typically raises complexity, requiring a trade-off to maintain linear complexity. Our key insight is that each node only needs to interact with a small subset of hubs to achieve linear complexity, even when the total number of hubs is large. To leverage all hubs without incurring additional computational costs, we propose a simple yet effective adaptive reassignment technique based on hub-hub similarity scores, eliminating the need for expensive node-hub computations. Our experiments on LRGB indicate a consistent improvement in results over the base method, Neural Atoms, while maintaining a linear complexity. Remarkably, our sparse model achieves performance on par with its non-sparse counterpart. Furthermore, ReHub outperforms competitive baselines and consistently ranks among top performers across various benchmarks.
Authors: Xiaoyu Luo, Qiongxiu Li
Abstract: Adversarial robustness, the ability of a model to withstand manipulated inputs that cause errors, is essential for ensuring the trustworthiness of machine learning models in real-world applications. However, previous studies have shown that enhancing adversarial robustness through adversarial training increases vulnerability to privacy attacks. While differential privacy can mitigate these attacks, it often compromises robustness against both natural and adversarial samples. Our analysis reveals that differential privacy disproportionately impacts low-risk samples, causing an unintended performance drop. To address this, we propose DeMem, which selectively targets high-risk samples, achieving a better balance between privacy protection and model robustness. DeMem is versatile and can be seamlessly integrated into various adversarial training techniques. Extensive evaluations across multiple training methods and datasets demonstrate that DeMem significantly reduces privacy leakage while maintaining robustness against both natural and adversarial samples. These results confirm DeMem's effectiveness and broad applicability in enhancing privacy without compromising robustness.
Authors: Chao Feng, Yuanzhe Gao, Alberto Huertas Celdran, Gerome Bovet, Burkhard Stiller
Abstract: Federated Learning (FL) is widely recognized as a privacy-preserving Machine Learning paradigm due to its model-sharing mechanism that avoids direct data exchange. Nevertheless, model training leaves exploitable traces that can be used to infer sensitive information. In Decentralized FL (DFL), the topology, defining how participants are connected, plays a crucial role in shaping the model's privacy, robustness, and convergence. However, the topology introduces an unexplored vulnerability: attackers can exploit it to infer participant relationships and launch targeted attacks. This work uncovers the hidden risks of DFL topologies by proposing a novel Topology Inference Attack that infers the topology solely from model behavior. A taxonomy of topology inference attacks is introduced, categorizing them by the attacker's capabilities and knowledge. Practical attack strategies are designed for various scenarios, and experiments are conducted to identify key factors influencing attack success. The results demonstrate that analyzing only the model of each node can accurately infer the DFL topology, highlighting a critical privacy risk in DFL systems. These findings offer insights for improving privacy preservation in DFL environments.
Authors: Tim Grams, Patrick Betz, Sascha Marton, Stefan L\"udtke, Christian Bartelt
Abstract: Exploration is a crucial skill for in-context reinforcement learning in unknown environments. However, it remains unclear if large language models can effectively explore a partially hidden state space. This work isolates exploration as the sole objective, tasking an agent with gathering information that enhances future returns. Within this framework, we argue that measuring agent returns is not sufficient for a fair evaluation. Hence, we decompose missing rewards into their exploration and exploitation components based on the optimal achievable return. Experiments with various models reveal that most struggle to explore the state space, and weak exploration is insufficient. Nevertheless, we found a positive correlation between exploration performance and reasoning capabilities. Our decomposition can provide insights into differences in behaviors driven by prompt engineering, offering a valuable tool for refining performance in exploratory tasks.
Authors: Elham Kiyani, Khemraj Shukla, Jorge F. Urb\'an, J\'er\^ome Darbon, George Em Karniadakis
Abstract: Physics-Informed Neural Networks (PINNs) have revolutionized the computation of PDE solutions by integrating partial differential equations (PDEs) into the neural network's training process as soft constraints, becoming an important component of the scientific machine learning (SciML) ecosystem. More recently, physics-informed Kolmogorv-Arnold networks (PIKANs) have also shown to be effective and comparable in accuracy with PINNs. In their current implementation, both PINNs and PIKANs are mainly optimized using first-order methods like Adam, as well as quasi-Newton methods such as BFGS and its low-memory variant, L-BFGS. However, these optimizers often struggle with highly non-linear and non-convex loss landscapes, leading to challenges such as slow convergence, local minima entrapment, and (non)degenerate saddle points. In this study, we investigate the performance of Self-Scaled BFGS (SSBFGS), Self-Scaled Broyden (SSBroyden) methods and other advanced quasi-Newton schemes, including BFGS and L-BFGS with different line search strategies. These methods dynamically rescale updates based on historical gradient information, thus enhancing training efficiency and accuracy. We systematically compare these optimizers using both PINNs and PIKANs on key challenging PDEs, including the Burgers, Allen-Cahn, Kuramoto-Sivashinsky, Ginzburg-Landau, and Stokes equations. Additionally, we evaluate the performance of SSBFGS and SSBroyden for Deep Operator Network (DeepONet) architectures, demonstrating their effectiveness for data-driven operator learning. Our findings provide state-of-the-art results with orders-of-magnitude accuracy improvements without the use of adaptive weights or any other enhancements typically employed in PINNs.
Authors: Jiwoo Kim, Joonhyung Lee, Gunho Park, Byeongwook Kim, Se Jung Kwon, Dongsoo Lee, Youngjoo Lee
Abstract: As large language models (LLMs) continue to scale, the high power consumption of AI accelerators in datacenters presents significant challenges, substantially increasing the total cost of ownership (TCO) for cloud service providers (CSPs) that provide LLM inference. In this work, we analyze the computational characteristics of LLM inference from a TCO perspective and present a generalizable framework to compare AI accelerators across diverse operational requirements. Using this model, we investigate key workload characteristics influencing TCO for AI accelerators from Intel (Gaudi 2 & 3) and NVIDIA (H100 & H200), especially thin GEMM utilization and FP8 quantization. In particular, as FP8 emerges as the baseline precision for next-generation LLMs, understanding how different architectures implement and benefit from low-precision computation is increasingly critical. Throughput on thin GEMMs has a greater impact on TCO than theoretical hardware peak throughput because the memory-bound decode phase is dominated by GEMV-like computations. We find that Gaudi HPUs achieve superior utilization on thin GEMMs compared to their counterparts, especially in FP8-quantized models. Our result underscores the importance of empirical, workload-level analysis in evaluating accelerator performance, rather than relying solely on theoretical hardware specifications. By studying the interaction between power consumption, quantization strategies, and hardware architecture, we provide insights to support informed deployment decisions and guide future accelerator designs aimed at improving the TCO of LLM inference workloads.
Authors: Alexander Kolesov, Manukhov Stepan, Vladimir V. Palyulin, Alexander Korotin
Abstract: We propose Electrostatic Field Matching (EFM), a novel method that is suitable for both generative modeling and distribution transfer tasks. Our approach is inspired by the physics of an electrical capacitor. We place source and target distributions on the capacitor plates and assign them positive and negative charges, respectively. Then we learn the electrostatic field of the capacitor using a neural network approximator. To map the distributions to each other, we start at one plate of the capacitor and move the samples along the learned electrostatic field lines until they reach the other plate. We theoretically justify that this approach provably yields the distribution transfer. In practice, we demonstrate the performance of our EFM in toy and image data experiments. Our code is available at https://github.com/justkolesov/FieldMatching
Authors: Keisuke Kamahori, Jungo Kasai, Noriyuki Kojima, Baris Kasikci
Abstract: Modern automatic speech recognition (ASR) models, such as OpenAI's Whisper, rely on deep encoder-decoder architectures, and their encoders are a critical bottleneck for efficient deployment due to high computational intensity. We introduce LiteASR, a low-rank compression scheme for ASR encoders that significantly reduces inference costs while maintaining transcription accuracy. Our approach leverages the strong low-rank properties observed in intermediate activations: by applying principal component analysis (PCA) with a small calibration dataset, we approximate linear transformations with a chain of low-rank matrix multiplications, and further optimize self-attention to work in reduced dimensionality. Evaluation results show that our method can compress Whisper large-v3's encoder size by over 50%, matching Whisper medium's size with better transcription accuracy, thereby establishing a new Pareto frontier of accuracy and efficiency. The code of LiteASR is available at https://github.com/efeslab/LiteASR.
Authors: Aditya Shankar, Lydia Y. Chen, Arie van Deursen, Rihan Hai
Abstract: Generating temporal data under conditions is crucial for forecasting, imputation, and generative tasks. Such data often has metadata and partially observed signals that jointly influence the generated values. However, existing methods face three key limitations: (1) they condition on either the metadata or observed values, but rarely both together; (2) they adopt either training-time approaches that fail to generalize to unseen scenarios, or inference-time approaches that ignore metadata; and (3) they suffer from trade-offs between generation speed and temporal coherence across time windows--choosing either slow but coherent autoregressive methods or fast but incoherent parallel ones. We propose WaveStitch, a novel diffusion-based method to overcome these hurdles through: (1) dual-sourced conditioning on both metadata and partially observed signals; (2) a hybrid training-inference architecture, incorporating metadata during training and observations at inference via gradient-based guidance; and (3) a novel pipeline-style paradigm that generates time windows in parallel while preserving coherence through an inference-time conditional loss and a stitching mechanism. Across diverse datasets, WaveStitch demonstrates adaptability to arbitrary patterns of observed signals, achieving 1.81x lower mean-squared-error compared to the state-of-the-art, and generates data up to 166.48x faster than autoregressive methods while maintaining coherence. Our code is available at: https://github.com/adis98/WaveStitch
Authors: Liane Xu, Amit Singer
Abstract: Laplacian-based methods are popular for dimensionality reduction of data lying in $\mathbb{R}^N$. Several theoretical results for these algorithms depend on the fact that the Euclidean distance locally approximates the geodesic distance on the underlying submanifold which the data are assumed to lie on. However, for some applications, other metrics, such as the Wasserstein distance, may provide a more appropriate notion of distance than the Euclidean distance. We provide a framework that generalizes the problem of manifold learning to metric spaces and study when a metric satisfies sufficient conditions for the pointwise convergence of the graph Laplacian.
Authors: Jihwan Oh, Minchan Jeong, Jongwoo Ko, Se-Young Yun
Abstract: Large Language Models $($LLMs$)$ solve complex problems using training-free methods like prompt engineering and in-context learning, yet ensuring reasoning correctness remains challenging. While self-correction methods such as self-consistency and self-refinement aim to improve reliability, they often reinforce biases due to the lack of effective feedback mechanisms. Multi-Agent Debate $($MAD$)$ has emerged as an alternative, but we identify two key limitations: bias reinforcement, where debate amplifies model biases instead of correcting them, and lack of perspective diversity, as all agents share the same model and reasoning patterns, limiting true debate effectiveness. To systematically evaluate these issues, we introduce $\textit{MetaNIM Arena}$, a benchmark designed to assess LLMs in adversarial strategic decision-making, where dynamic interactions influence optimal decisions. To overcome MAD's limitations, we propose $\textbf{DReaMAD}$ $($$\textbf{D}$iverse $\textbf{Rea}$soning via $\textbf{M}$ulti-$\textbf{A}$gent $\textbf{D}$ebate with Refined Prompt$)$, a novel framework that $(1)$ refines LLM's strategic prior knowledge to improve reasoning quality and $(2)$ promotes diverse viewpoints within a single model by systematically modifying prompts, reducing bias. Empirical results show that $\textbf{DReaMAD}$ significantly improves decision accuracy, reasoning diversity, and bias mitigation across multiple strategic tasks, establishing it as a more effective approach for LLM-based decision-making.
Authors: Arik Ermshaus, Patrick Sch\"afer, Ulf Leser
Abstract: The ever-growing amount of sensor data from machines, smart devices, and the environment leads to an abundance of high-resolution, unannotated time series (TS). These recordings encode recognizable properties of latent states and transitions from physical phenomena that can be modelled as abstract processes. The unsupervised localization and identification of these states and their transitions is the task of time series state detection (TSSD). Current TSSD algorithms employ classical unsupervised learning techniques, to infer state membership directly from feature space. This limits their predictive power, compared to supervised learning methods, which can exploit additional label information. We introduce CLaP, a new, highly accurate and efficient algorithm for TSSD. It leverages the predictive power of time series classification for TSSD in an unsupervised setting by applying novel self-supervision techniques to detect whether data segments emerge from the same state. To this end, CLaP cross-validates a classifier with segment-labelled subsequences to quantify confusion between segments. It merges labels from segments with high confusion, representing the same latent state, if this leads to an increase in overall classification quality. We conducted an experimental evaluation using 405 TS from five benchmarks and found CLaP to be significantly more precise in detecting states than six state-of-the-art competitors. It achieves the best accuracy-runtime tradeoff and is scalable to large TS. We provide a Python implementation of CLaP, which can be deployed in TS analysis workflows.
Authors: Akira Tamamori
Abstract: Hopfield networks using Hebbian learning suffer from limited storage capacity. While supervised methods like Linear Logistic Regression (LLR) offer some improvement, kernel methods like Kernel Logistic Regression (KLR) significantly enhance storage capacity and noise robustness. However, KLR requires computationally expensive iterative learning. We propose Kernel Ridge Regression (KRR) as an efficient kernel-based alternative for learning high-capacity Hopfield networks. KRR utilizes the kernel trick and predicts bipolar states via regression, crucially offering a non-iterative, closed-form solution for learning dual variables. We evaluate KRR and compare its performance against Hebbian, LLR, and KLR. Our results demonstrate that KRR achieves state-of-the-art storage capacity (reaching a storage load of 1.5) and noise robustness, comparable to KLR. Crucially, KRR drastically reduces training time, being orders of magnitude faster than LLR and significantly faster than KLR, especially at higher storage loads. This establishes KRR as a potent and highly efficient method for building high-performance associative memories, providing comparable performance to KLR with substantial training speed advantages. This work provides the first empirical comparison between KRR and KLR in the context of Hopfield network learning.
Authors: Manar D. Samad, Kazi Fuad B. Akhter, Shourav B. Rabbani, Ibna Kowsar
Abstract: Tabular data sets with varying missing values are prepared for machine learning using an arbitrary imputation strategy. Synthetic values generated by imputation models often raise concerns about computational complexity, data quality, and data-driven outcomes. To address these concerns, this article proposes a no-imputation incremental attention learning (NIAL) method for tabular data. A pair of attention masks is derived and retrofitted to a transformer to directly streamline tabular data without imputing or initializing missing values. The proposed method incrementally learns partitions of overlapping and fixed-size feature sets to enhance the efficiency and performance of the transformer. The average classification performance rank order across 15 diverse tabular data sets highlights the superiority of NIAL over 11 state-of-the-art learning methods with or without missing value imputations. Further experiments substantiate the robustness of NIAL against varying missing value types and rates compared to methods involving missing value imputation. Our analysis reveals that a feature partition size of half the original feature space is, both computationally and in terms of accuracy, the best choice for the proposed incremental learning. The proposed method is one of the first solutions to enable deep attention learning of tabular data without requiring missing-value imputation.
Authors: Stefan Jonas, Angela Meyer
Abstract: Intelligent condition monitoring of wind turbines is essential for reducing downtimes. Machine learning models trained on wind turbine operation data are commonly used to detect anomalies and, eventually, operation faults. However, data-driven normal behavior models (NBMs) require a substantial amount of training data, as NBMs trained with scarce data may result in unreliable fault detection. To overcome this limitation, we present a novel generative deep transfer learning approach to make SCADA samples from one wind turbine lacking training data resemble SCADA data from wind turbines with representative training data. Through CycleGAN-based domain mapping, our method enables the application of an NBM trained on an existing wind turbine to a new one with severely limited data. We demonstrate our approach on field data mapping SCADA samples across 7 substantially different WTs. Our findings show significantly improved fault detection in wind turbines with scarce data. Our method achieves the most similar anomaly scores to an NBM trained with abundant data, outperforming NBMs trained on scarce training data with improvements of +10.3% in F1-score when 1 month of training data is available and +16.8% when 2 weeks are available. The domain mapping approach outperforms conventional fine-tuning at all considered degrees of data scarcity, ranging from 1 to 8 weeks of training data. The proposed technique enables earlier and more reliable fault detection in newly installed wind farms, demonstrating a novel and promising research direction to improve anomaly detection when faced with training data scarcity.
Authors: Chris Child, Lam Ngo
Abstract: The DeeP-Mod framework builds an environment model using features from a Deep Dynamic Programming Network (DDPN), trained via a Deep Q-Network (DQN). While Deep Q-Learning is effective in decision-making, state information is lost in deeper DQN layers due to mixed state-action representations. We address this by using Dynamic Programming (DP) to train a DDPN, where Value Iteration ensures the output represents state values, not state-action pairs. Extracting features from the DDPN preserves state information, enabling task and action set independence. We show that a reduced DDPN can be trained using features extracted from the original DDPN trained on an identical problem. This reduced DDPN achieves faster convergence under noise and outperforms the original DDPN. Finally, we introduce the DeeP-Mod framework, which creates an environment model using the evolution of features extracted from a DDPN in response to actions. A second DDPN, which learns directly from this feature model rather than raw states, can learn an effective feature-value representation and thus optimal policy. A key advantage of DeeP-Mod is that an externally defined environment model is not needed at any stage, making DDPN applicable to a wide range of environments.
Authors: Suk Ki Lee, Hyunwoong Ko
Abstract: Dynamic manufacturing processes exhibit complex characteristics defined by time-varying parameters, nonlinear behaviors, and uncertainties. These characteristics require sophisticated in-situ monitoring techniques utilizing multimodal sensor data and adaptive control systems that can respond to real-time feedback while maintaining product quality. Recently, generative machine learning (ML) has emerged as a powerful tool for modeling complex distributions and generating synthetic data while handling these manufacturing uncertainties. However, adopting these generative technologies in dynamic manufacturing systems lacks a functional control-oriented perspective to translate their probabilistic understanding into actionable process controls while respecting constraints. This review presents a functional classification of Prediction-Based, Direct Policy, Quality Inference, and Knowledge-Integrated approaches, offering a perspective for understanding existing ML-enhanced control systems and incorporating generative ML. The analysis of generative ML architectures within this framework demonstrates control-relevant properties and potential to extend current ML-enhanced approaches where conventional methods prove insufficient. We show generative ML's potential for manufacturing control through decision-making applications, process guidance, simulation, and digital twins, while identifying critical research gaps: separation between generation and control functions, insufficient physical understanding of manufacturing phenomena, and challenges adapting models from other domains. To address these challenges, we propose future research directions aimed at developing integrated frameworks that combine generative ML and control technologies to address the dynamic complexities of modern manufacturing systems.
Authors: Xinlin Li, Osama Hanna, Christina Fragouli, Suhas Diggavi
Abstract: The rapid deployment of Large Language Models (LLMs) highlights the need for efficient low-bit post-training quantization (PTQ), due to their high memory costs. A key challenge in weight quantization is the presence of outliers, which inflate quantization ranges and lead to large errors. While a number of outlier suppression techniques have been proposed, they either: fail to effectively shrink the quantization range, or incur (relatively) high bit overhead. In this paper, we present ICQuant, a novel framework that leverages outlier statistics to design an efficient index coding scheme for outlier-aware weight-only quantization. Compared to existing outlier suppression techniques requiring $\approx 1$ bit overhead to halve the quantization range, ICQuant requires only $\approx 0.3$ bits; a significant saving in extreme compression regimes (e.g., 2-3 bits per weight). ICQuant can be used on top of any existing quantizers to eliminate outliers, improving the quantization quality. Using just 2.3 bits per weight and simple scalar quantizers, ICQuant improves the zero-shot accuracy of the 2-bit Llama3-70B model by up to 130% and 150% relative to QTIP and QuIP#; and it achieves comparable performance to the best-known fine-tuned quantizer (PV-tuning) without fine-tuning.
Authors: Haoyue Bai, Yiyou Sun, Wei Cheng, Haifeng Chen
Abstract: The recent proliferation of photorealistic images created by generative models has sparked both excitement and concern, as these images are increasingly indistinguishable from real ones to the human eye. While offering new creative and commercial possibilities, the potential for misuse, such as in misinformation and fraud, highlights the need for effective detection methods. Current detection approaches often rely on access to model weights or require extensive collections of real image datasets, limiting their scalability and practical application in real world scenarios. In this work, we introduce a novel black box detection framework that requires only API access, sidestepping the need for model weights or large auxiliary datasets. Our approach leverages a corrupt and recover strategy: by masking part of an image and assessing the model ability to reconstruct it, we measure the likelihood that the image was generated by the model itself. For black-box models that do not support masked image inputs, we incorporate a cost efficient surrogate model trained to align with the target model distribution, enhancing detection capability. Our framework demonstrates strong performance, outperforming baseline methods by 4.31% in mean average precision across eight diffusion model variant datasets.
Authors: Drew Prinster, Xing Han, Anqi Liu, Suchi Saria
Abstract: Responsibly deploying artificial intelligence (AI) / machine learning (ML) systems in high-stakes settings arguably requires not only proof of system reliability, but also continual, post-deployment monitoring to quickly detect and address any unsafe behavior. Methods for nonparametric sequential testing -- especially conformal test martingales (CTMs) and anytime-valid inference -- offer promising tools for this monitoring task. However, existing approaches are restricted to monitoring limited hypothesis classes or ``alarm criteria'' (e.g., detecting data shifts that violate certain exchangeability or IID assumptions), do not allow for online adaptation in response to shifts, and/or cannot diagnose the cause of degradation or alarm. In this paper, we address these limitations by proposing a weighted generalization of conformal test martingales (WCTMs), which lay a theoretical foundation for online monitoring for any unexpected changepoints in the data distribution while controlling false-alarms. For practical applications, we propose specific WCTM algorithms that adapt online to mild covariate shifts (in the marginal input distribution), quickly detect harmful shifts, and diagnose those harmful shifts as concept shifts (in the conditional label distribution) or extreme (out-of-support) covariate shifts that cannot be easily adapted to. On real-world datasets, we demonstrate improved performance relative to state-of-the-art baselines.
Authors: Chuanbo Hua, Federico Berto, Zhikai Zhao, Jiwoo Son, Changhyun Kwon, Jinkyoo Park
Abstract: The Profiled Vehicle Routing Problem (PVRP) extends the classical VRP by incorporating vehicle-client-specific preferences and constraints, reflecting real-world requirements such as zone restrictions and service-level preferences. While recent reinforcement-learning solvers have shown promising performance, they require retraining for each new profile distribution, suffer from poor representation ability, and struggle to generalize to out-of-distribution instances. In this paper, we address these limitations by introducing Unified Solver for Profiled Routing (USPR), a novel framework that natively handles arbitrary profile types. USPR introduces on three key innovations: (i) Profile Embeddings (PE) to encode any combination of profile types; (ii) Multi-Head Profiled Attention (MHPA), an attention mechanism that models rich interactions between vehicles and clients; (iii) Profile-aware Score Reshaping (PSR), which dynamically adjusts decoder logits using profile scores to improve generalization. Empirical results on diverse PVRP benchmarks demonstrate that USPR achieves state-of-the-art results among learning-based methods while offering significant gains in flexibility and computational efficiency. We make our source code publicly available to foster future research.
Authors: Zhihao Dou, Dongfei Cui, Jun Yan, Weida Wang, Benteng Chen, Haoming Wang, Zeke Xie, Shufei Zhang
Abstract: Although Reinforcement Learning (RL) agents are effective in well-defined environments, they often struggle to generalize their learned policies to dynamic settings due to their reliance on trial-and-error interactions. Recent work has explored applying Large Language Models (LLMs) or Vision Language Models (VLMs) to boost the generalization of RL agents through policy optimization guidance or prior knowledge. However, these approaches often lack seamless coordination between the RL agent and the foundation model, leading to unreasonable decision-making in unfamiliar environments and efficiency bottlenecks. Making full use of the inferential capabilities of foundation models and the rapid response capabilities of RL agents and enhancing the interaction between the two to form a dual system is still a lingering scientific question. To address this problem, we draw inspiration from Kahneman's theory of fast thinking (System 1) and slow thinking (System 2), demonstrating that balancing intuition and deep reasoning can achieve nimble decision-making in a complex world. In this study, we propose a Dual-System Adaptive Decision Framework (DSADF), integrating two complementary modules: System 1, comprising an RL agent and a memory space for fast and intuitive decision making, and System 2, driven by a VLM for deep and analytical reasoning. DSADF facilitates efficient and adaptive decision-making by combining the strengths of both systems. The empirical study in the video game environment: Crafter and Housekeep demonstrates the effectiveness of our proposed method, showing significant improvements in decision abilities for both unseen and known tasks.
Authors: Varun Raaghav, Dimitrios Bikos, Antonio Rago, Francesca Toni, Maria Charalambides
Abstract: Composites are amongst the most important materials manufactured today, as evidenced by their use in countless applications. In order to establish the suitability of composites in specific applications, finite element (FE) modelling, a numerical method based on partial differential equations, is the industry standard for assessing their mechanical properties. However, FE modelling is exceptionally costly from a computational viewpoint, a limitation which has led to efforts towards applying AI models to this task. However, in these approaches: the chosen model architectures were rudimentary, feed-forward neural networks giving limited accuracy; the studies focused on predicting elastic mechanical properties, without considering material strength limits; and the models lacked transparency, hindering trustworthiness by users. In this paper, we show that convolutional neural networks (CNNs) equipped with methods from explainable AI (XAI) can be successfully deployed to solve this problem. Our approach uses customised CNNs trained on a dataset we generate using transverse tension tests in FE modelling to predict composites' mechanical properties, i.e., Young's modulus and yield strength. We show empirically that our approach achieves high accuracy, outperforming a baseline, ResNet-34, in estimating the mechanical properties. We then use SHAP and Integrated Gradients, two post-hoc XAI methods, to explain the predictions, showing that the CNNs use the critical geometrical features that influence the composites' behaviour, thus allowing engineers to verify that the models are trustworthy by representing the science of composites.
Authors: Benedikt H\"oltgen, Nuria Oliver
Abstract: Fairness through Unawareness (FtU) describes the idea that discrimination against demographic groups can be avoided by not considering group membership in the decisions or predictions. This idea has long been criticized in the machine learning literature as not being sufficient to ensure fairness. In addition, the use of additional features is typically thought to increase the accuracy of the predictions for all groups, so that FtU is sometimes thought to be detrimental to all groups. In this paper, we show both theoretically and empirically that FtU can reduce algorithmic discrimination without necessarily reducing accuracy. We connect this insight with the literature on Model Multiplicity, to which we contribute with novel theoretical and empirical results. Furthermore, we illustrate how, in a real-life application, FtU can contribute to the deployment of more equitable policies without losing efficacy. Our findings suggest that FtU is worth considering in practical applications, particularly in high-risk scenarios, and that the use of protected attributes such as gender in predictive models should be accompanied by a clear and well-founded justification.
Authors: Liangwei Nathan Zheng, Wei Emma Zhang, Mingyu Guo, Miao Xu, Olaf Maennel, Weitong Chen
Abstract: Effectively managing missing modalities is a fundamental challenge in real-world multimodal learning scenarios, where data incompleteness often results from systematic collection errors or sensor failures. Sparse Mixture-of-Experts (SMoE) architectures have the potential to naturally handle multimodal data, with individual experts specializing in different modalities. However, existing SMoE approach often lacks proper ability to handle missing modality, leading to performance degradation and poor generalization in real-world applications. We propose ConfSMoE to introduce a two-stage imputation module to handle the missing modality problem for the SMoE architecture by taking the opinion of experts and reveal the insight of expert collapse from theoretical analysis with strong empirical evidence. Inspired by our theoretical analysis, ConfSMoE propose a novel expert gating mechanism by detaching the softmax routing score to task confidence score w.r.t ground truth signal. This naturally relieves expert collapse without introducing additional load balance loss function. We show that the insights of expert collapse aligns with other gating mechanism such as Gaussian and Laplacian gate. The proposed method is evaluated on four different real world dataset with three distinct experiment settings to conduct comprehensive analysis of ConfSMoE on resistance to missing modality and the impacts of proposed gating mechanism.
Authors: Junyi An, Xinyu Lu, Chao Qu, Yunfei Shi, Peijia Lin, Qianwei Tang, Licheng Xu, Fenglei Cao, Yuan Qi
Abstract: SE(3)-equivariant Graph Neural Networks (GNNs) have significantly advanced molecular system modeling by employing group representations. However, their message passing processes, which rely on tensor product-based convolutions, are limited by insufficient non-linearity and incomplete group representations, thereby restricting expressiveness. To overcome these limitations, we introduce the Equivariant Spherical Transformer (EST), a novel framework that leverages a Transformer structure within the spatial domain of group representations after Fourier transform. We theoretically and empirically demonstrate that EST can encompass the function space of tensor products while achieving superior expressiveness. Furthermore, EST's equivariant inductive bias is guaranteed through a uniform sampling strategy for the Fourier transform. Our experiments demonstrate state-of-the-art performance by EST on various molecular benchmarks, including OC20 and QM9.
Authors: Shichang Zhang, Hongzhe Du, Jiaqi W. Ma, Himabindu Lakkaraju
Abstract: Modern AI systems are typically developed through multiple stages-pretraining, fine-tuning rounds, and subsequent adaptation or alignment, where each stage builds on the previous ones and updates the model in distinct ways. This raises a critical question of accountability: when a deployed model succeeds or fails, which stage is responsible, and to what extent? We pose the accountability attribution problem for tracing model behavior back to specific stages of the model development process. To address this challenge, we propose a general framework that answers counterfactual questions about stage effects: how would the model's behavior have changed if the updates from a particular stage had not occurred? Within this framework, we introduce estimators that efficiently quantify stage effects without retraining the model, accounting for both the data and key aspects of model optimization dynamics, including learning rate schedules, momentum, and weight decay. We demonstrate that our approach successfully quantifies the accountability of each stage to the model's behavior. Based on the attribution results, our method can identify and remove spurious correlations learned during image classification and text toxicity detection tasks that were developed across multiple stages. Our approach provides a practical tool for model analysis and represents a significant step toward more accountable AI development.
Authors: Andr\'es Guzm\'an-Cordero, Floor Eijkelboom, Jan-Willem van de Meent
Abstract: While denoising diffusion and flow matching have driven major advances in generative modeling, their application to tabular data remains limited, despite its ubiquity in real-world applications. To this end, we develop TabbyFlow, a variational Flow Matching (VFM) method for tabular data generation. To apply VFM to data with mixed continuous and discrete features, we introduce Exponential Family Variational Flow Matching (EF-VFM), which represents heterogeneous data types using a general exponential family distribution. We hereby obtain an efficient, data-driven objective based on moment matching, enabling principled learning of probability paths over mixed continuous and discrete variables. We also establish a connection between variational flow matching and generalized flow matching objectives based on Bregman divergences. Evaluation on tabular data benchmarks demonstrates state-of-the-art performance compared to baselines.
Authors: Elie Kadoche, Pascal Bianchi, Florence Carton, Philippe Ciblat, Damien Ernst
Abstract: Within wind farms, wake effects between turbines can significantly reduce overall energy production. Wind farm flow control encompasses methods designed to mitigate these effects through coordinated turbine control. Wake steering, for example, consists in intentionally misaligning certain turbines with the wind to optimize airflow and increase power output. However, designing a robust wake steering controller remains challenging, and existing machine learning approaches are limited to quasi-static wind conditions or small wind farms. This work presents a new deep reinforcement learning methodology to develop a wake steering policy that overcomes these limitations. Our approach introduces a novel architecture that combines graph attention networks and multi-head self-attention blocks, alongside a novel reward function and training strategy. The resulting model computes the yaw angles of each turbine, optimizing energy production in time-varying wind conditions. An empirical study conducted on steady-state, low-fidelity simulation, shows that our model requires approximately 10 times fewer training steps than a fully connected neural network and achieves more robust performance compared to a strong optimization baseline, increasing energy production by up to 14 %. To the best of our knowledge, this is the first deep reinforcement learning-based wake steering controller to generalize effectively across any time-varying wind conditions in a low-fidelity, steady-state numerical simulation setting.
Authors: Paul Fogel (Data Services, Forvis Mazars, Levallois, France), Christophe Geissler (Data Services, Forvis Mazars, Levallois, France), George Luta (Department of Biostatistics, Bioinformatics and Biomathematics, Georgetown University Medical Center, Washington, DC, USA)
Abstract: The interpretation of the results of survival analysis often benefits from latent factor representations of baseline covariates. However, existing methods, such as Nonnegative Matrix Factorization (NMF), do not incorporate survival information, limiting their predictive power. We present CoxNTF, a novel approach that uses non-negative tensor factorization (NTF) to derive meaningful latent representations that are closely associated with survival outcomes. CoxNTF constructs a weighted covariate tensor in which survival probabilities derived from the Coxnet model are used to guide the tensorization process. Our results show that CoxNTF achieves survival prediction performance comparable to using Coxnet with the original covariates, while providing a structured and interpretable clustering framework. In addition, the new approach effectively handles feature redundancy, making it a powerful tool for joint clustering and prediction in survival analysis.
Authors: Yuan Yuan, Yukun Liu, Chonghua Han, Jie Feng, Yong Li
Abstract: Foundation models have revolutionized fields such as natural language processing and computer vision by enabling general-purpose learning across diverse tasks and datasets. However, building analogous models for human mobility remains challenging due to the privacy-sensitive nature of mobility data and the resulting data silos across institutions. To bridge this gap, we propose MoveGCL, a scalable and privacy-preserving framework for training mobility foundation models via generative continual learning. Without sharing raw data, MoveGCL enables decentralized and progressive model evolution by replaying synthetic trajectories generated from a frozen teacher model, and reinforces knowledge retention through a tailored distillation strategy that mitigates catastrophic forgetting. To address the heterogeneity of mobility patterns, MoveGCL incorporates a Mixture-of-Experts Transformer with a mobility-aware expert routing mechanism, and employs a layer-wise progressive adaptation strategy to stabilize continual updates. Experiments on six real-world urban datasets demonstrate that MoveGCL achieves performance comparable to joint training and significantly outperforms federated learning baselines, while offering strong privacy protection. MoveGCL marks a crucial step toward unlocking foundation models for mobility, offering a practical blueprint for open, scalable, and privacy-preserving model development in the era of foundation models. To facilitate reproducibility and future research, we have released the code and models at https://github.com/tsinghua-fib-lab/MoveGCL.
Authors: Milad Hoseinpour, Vladimir Dvorkin
Abstract: High-quality power flow datasets are essential for training machine learning models in power systems. However, security and privacy concerns restrict access to real-world data, making statistically accurate and physically consistent synthetic datasets a viable alternative. We develop a diffusion model for generating synthetic power flow datasets from real-world power grids that both replicate the statistical properties of the real-world data and ensure AC power flow feasibility. To enforce the constraints, we incorporate gradient guidance based on the power flow constraints to steer diffusion sampling toward feasible samples. For computational efficiency, we further leverage insights from the fast decoupled power flow method and propose a variable decoupling strategy for the training and sampling of the diffusion model. These solutions lead to a physics-informed diffusion model, generating power flow datasets that outperform those from the standard diffusion in terms of feasibility and statistical similarity, as shown in experiments across IEEE benchmark systems.
Authors: Jiahui Zheng, Cole Jahnke, Wei "Wayne" Chen
Abstract: This paper introduces GUST (Generative Uncertainty learning via Self-supervised pretraining and Transfer learning), a framework for quantifying free-form geometric uncertainties inherent in the manufacturing of metamaterials. GUST leverages the representational power of deep generative models to learn a high-dimensional conditional distribution of as-fabricated unit cell geometries given nominal designs, thereby enabling uncertainty quantification. To address the scarcity of real-world manufacturing data, GUST employs a two-stage learning process. First, it leverages self-supervised pretraining on a large-scale synthetic dataset to capture the structure variability inherent in metamaterial geometries and an approximated distribution of as-fabricated geometries given nominal designs. Subsequently, GUST employs transfer learning by fine-tuning the pretrained model on limited real-world manufacturing data, allowing it to adapt to specific manufacturing processes and nominal designs. With only 960 unit cells additively manufactured in only two passes, GUST can capture the variability in geometry and effective material properties. In contrast, directly training a generative model on the same amount of real-world data proves insufficient, as demonstrated through both qualitative and quantitative comparisons. This scalable and cost-effective approach significantly reduces data requirements while maintaining the effectiveness in learning complex, real-world geometric uncertainties, offering an affordable method for free-form geometric uncertainty quantification in the manufacturing of metamaterials. The capabilities of GUST hold significant promise for high-precision industries such as aerospace and biomedical engineering, where understanding and mitigating manufacturing uncertainties are critical.
Authors: Haiyang Guo, Fanhu Zeng, Fei Zhu, Jiayi Wang, Xukai Wang, Jingang Zhou, Hongbo Zhao, Wenzhuo Liu, Shijie Ma, Da-Han Wang, Xu-Yao Zhang, Cheng-Lin Liu
Abstract: The rapid advancement of generative models has empowered modern AI systems to comprehend and produce highly sophisticated content, even achieving human-level performance in specific domains. However, these models are fundamentally constrained by \emph{catastrophic forgetting}, \ie~a persistent challenge where models experience performance degradation on previously learned tasks when adapting to new tasks. To address this practical limitation, numerous approaches have been proposed to enhance the adaptability and scalability of generative AI in real-world applications. In this work, we present a comprehensive survey of continual learning methods for mainstream generative AI models, encompassing large language models, multimodal large language models, vision-language-action models, and diffusion models. Drawing inspiration from the memory mechanisms of the human brain, we systematically categorize these approaches into three paradigms: architecture-based, regularization-based, and replay-based methods, while elucidating their underlying methodologies and motivations. We further analyze continual learning setups for different generative models, including training objectives, benchmarks, and core backbones, thereby providing deeper insights into the field. The project page of this paper is available at https://github.com/Ghy0501/Awesome-Continual-Learning-in-Generative-Models.
URLs: https://github.com/Ghy0501/Awesome-Continual-Learning-in-Generative-Models.
Authors: Francesco Carzaniga, Michael Hersche, Abu Sebastian, Kaspar Schindler, Abbas Rahimi
Abstract: Learning from multi-variate time-series with heterogeneous channel configurations remains a fundamental challenge for deep neural networks, particularly in clinical domains such as intracranial electroencephalography (iEEG), where channel setups vary widely across subjects. In this work, we introduce multi-variate parallel attention (MVPA), a novel self-attention mechanism that disentangles content, temporal, and spatial attention, enabling flexible, generalizable, and efficient modeling of time-series data with varying channel counts and configurations. We use MVPA to build MVPFormer, a generative foundation model for human electrophysiology, trained to predict the evolution of iEEG signals across diverse subjects. To support this and future efforts by the community, we release the SWEC iEEG dataset, the largest publicly available iEEG dataset to date, comprising nearly 10,000 hours of recordings from heterogeneous clinical sources. MVPFormer leverages MVPA to achieve strong generalization across subjects, demonstrating expert-level performance in several iEEG tasks. MVPFormer surpasses state-of-the-art Transformer baselines in seizure detection across the SWEC, the MAYO, and the FNUSA datasets, while also achieving state-of-the-art performance on four Brain TreeBank iEEG decoding tasks. We further validate MVPA on standard time-series forecasting and classification tasks, where it matches or exceeds the performance of existing attention-based models. Together, our contributions establish MVPA as a general-purpose attention mechanism for heterogeneous time-series and MVPFormer as the first open-source, open-weights, and open-data iEEG foundation model with SOTA clinical performance. The code is available at https://github.com/IBM/multi-variate-parallel-transformer. The SWEC iEEG dataset is available at https://huggingface.co/datasets/NeuroTec/SWEC_iEEG_Dataset.
URLs: https://github.com/IBM/multi-variate-parallel-transformer., https://huggingface.co/datasets/NeuroTec/SWEC_iEEG_Dataset.
Authors: XiaYu Liu, Chao Fan, Yang Liu, Hou-biao Li
Abstract: Accurate prediction of molecular properties is essential in drug discovery and related fields. However, existing graph neural networks (GNNs) often struggle to simultaneously capture both local and global molecular structures. In this work, we propose a Multi-Level Fusion Graph Neural Network (MLFGNN) that integrates Graph Attention Networks and a novel Graph Transformer to jointly model local and global dependencies. In addition, we incorporate molecular fingerprints as a complementary modality and introduce a mechanism of interaction between attention to adaptively fuse information across representations. Extensive experiments on multiple benchmark datasets demonstrate that MLFGNN consistently outperforms state-of-the-art methods in both classification and regression tasks. Interpretability analysis further reveals that the model effectively captures task-relevant chemical patterns, supporting the usefulness of multi-level and multi-modal fusion in molecular representation learning.
Authors: Yudan Song (Co-first author), Yuecen Wei (Co-first author), Yuhang Lu, Qingyun Sun, Minglai Shao, Li-e Wang, Chunming Hu, Xianxian Li, Xingcheng Fu
Abstract: Graph representation learning has become a mainstream method for fraud detection due to its strong expressive power, which focuses on enhancing node representations through improved neighborhood knowledge capture. However, the focus on local interactions leads to imbalanced transmission of global topological information and increased risk of node-specific information being overwhelmed during aggregation due to the imbalance between fraud and benign nodes. In this paper, we first summarize the impact of topology and class imbalance on downstream tasks in GNN-based fraud detection, as the problem of imbalanced supervisory messages is caused by fraudsters' topological behavior obfuscation and identity feature concealment. Based on statistical validation, we propose a novel dual-view graph representation learning method to mitigate Message imbalance in Fraud Detection (MimbFD). Specifically, we design a topological message reachability module for high-quality node representation learning to penetrate fraudsters' camouflage and alleviate insufficient propagation. Then, we introduce a local confounding debiasing module to adjust node representations, enhancing the stable association between node representations and labels to balance the influence of different classes. Finally, we conducted experiments on three public fraud datasets, and the results demonstrate that MimbFD exhibits outstanding performance in fraud detection.
Authors: Paul Fogel (Data Services, Forvis Mazars, Levallois, France), Christophe Geissler (Data Services, Forvis Mazars, Levallois, France), George Luta (Department of Biostatistics, Bioinformatics and Biomathematics, Georgetown University Medical Center, Washington, DC, USA)
Abstract: This paper introduces the "Target Polish," a robust and computationally efficient framework for Non-Negative Matrix Factorization (NMF). Although conventional weighted NMF approaches are resistant to outliers, they converge slowly due to the use of multiplicative updates to minimize the objective criterion. In contrast, the Target Polish approach remains compatible with the Fast-HALS algorithm, which is renowned for its speed, by adaptively "polishing" the data with a weighted median-based transformation. This innovation provides outlier resistance while maintaining the highly efficient additive update structure of Fast-HALS. Empirical evaluations using image datasets corrupted with structured (block) and unstructured (salt) noise demonstrate that the Target Polish approach matches or exceeds the accuracy of state-of-the-art robust NMF methods while reducing computational time by an order of magnitude in the studied scenarios.
Authors: Guokai Li, Pin Gao, Stefanus Jasin, Zizhuo Wang
Abstract: Assortment optimization involves selecting a subset of substitutable products (subject to certain constraints) to maximize the expected revenue. It is a classic problem in revenue management and finds applications across various industries. However, the problem is usually NP-hard due to its combinatorial and non-linear nature. In this work, we explore how graph convolutional networks (GCNs) can be leveraged to efficiently solve constrained assortment optimization under the mixed multinomial logit choice model. We first develop a graph representation of the assortment problem, then train a GCN to learn the patterns of optimal assortments, and lastly propose two inference policies based on the GCN's output. Due to the GCN's inherent ability to generalize across inputs of varying sizes, we can use a GCN trained on small-scale instances to facilitate large-scale instances. Extensive numerical experiments demonstrate that given a GCN trained on small-scale instances (e.g., with 20 products), the proposed policies can achieve superior performance (90%+ optimality) on large-scale instances (with up to 2,000 products) within seconds, which outperform existing heuristic policies in both performance and efficiency. Furthermore, we extend our framework to a model-free setting where the underlying choice model is unknown but transaction data is available. We also conduct numerical experiments to demonstrate the effectiveness and efficiency of our proposed policies in this setting.
Authors: Annalisa Barbara, Joseph Lazzaro, Ciara Pike-Burke
Abstract: This work investigates the impact of ensuring local differential privacy in the thresholding bandit problem. We consider both the fixed budget and fixed confidence settings. We propose methods that utilize private responses, obtained through a Bernoulli-based differentially private mechanism, to identify arms with expected rewards exceeding a predefined threshold. We show that this procedure provides strong privacy guarantees and derive theoretical performance bounds on the proposed algorithms. Additionally, we present general lower bounds that characterize the additional loss incurred by any differentially private mechanism, and show that the presented algorithms match these lower bounds up to poly-logarithmic factors. Our results provide valuable insights into privacy-preserving decision-making frameworks in bandit problems.
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: 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: 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: 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: Jiansheng Rao, Jiayi Li, Zhizhi Gong, Soummya Kar, Haoxuan Li
Abstract: Federated learning (FL) is a widely adopted paradigm for privacy-preserving model training, but FedAvg optimise for the majority while under-serving minority clients. Existing methods such as federated multi-objective learning (FMOL) attempts to import multi-objective optimisation (MOO) into FL. However, it merely delivers task-wise Pareto-stationary points, leaving client fairness to chance. In this paper, we introduce Conically-Regularised FMOL (CR-FMOL), the first federated MOO framework that enforces client-wise Pareto optimality through a novel preference-cone constraint. After local federated multi-gradient descent averaging (FMGDA) / federated stochastic multi-gradient descent averaging (FSMGDA) steps, each client transmits its aggregated task-loss vector as an implicit preference; the server then solves a cone-constrained Pareto-MTL sub-problem centred at the uniform vector, producing a descent direction that is Pareto-stationary for every client within its cone. Experiments on non-IID benchmarks show that CR-FMOL enhances client fairness, and although the early-stage performance is slightly inferior to FedAvg, it is expected to achieve comparable accuracy given sufficient training rounds.
Authors: Vincent-Daniel Yun
Abstract: Low-precision training has become crucial for reducing the computational and memory costs of large-scale deep learning. However, quantizing gradients introduces magnitude shrinkage, which can change how stochastic gradient descent (SGD) converges. In this study, we explore SGD convergence under a gradient shrinkage model, where each stochastic gradient is scaled by a factor \( q_k \in (0,1] \). We show that this shrinkage affect the usual stepsize \( \mu_k \) with an effective stepsize \( \mu_k q_k \), slowing convergence when \( q_{\min} < 1 \). With typical smoothness and bounded-variance assumptions, we prove that low-precision SGD still converges, but at a slower pace set by \( q_{\min} \), and with a higher steady error level due to quantization effects. We analyze theoretically how lower numerical precision slows training by treating it as gradient shrinkage within the standard SGD convergence setup.
Authors: Vincent Perreault, Katsumi Inoue, Richard Labib, Alain Hertz
Abstract: Traditional neural networks have an impressive classification performance, but what they learn cannot be inspected, verified or extracted. Neural Logic Networks on the other hand have an interpretable structure that enables them to learn a logical mechanism relating the inputs and outputs with AND and OR operations. We generalize these networks with NOT operations and biases that take into account unobserved data and develop a rigorous logical and probabilistic modeling in terms of concept combinations to motivate their use. We also propose a novel factorized IF-THEN rule structure for the model as well as a modified learning algorithm. Our method improves the state-of-the-art in Boolean networks discovery and is able to learn relevant, interpretable rules in tabular classification, notably on examples from the medical and industrial fields where interpretability has tangible value.
Authors: Yunshan Peng, Wenzheng Shu, Jiahao Sun, Yanxiang Zeng, Jinan Pang, Wentao Bai, Yunke Bai, Xialong Liu, Peng Jiang
Abstract: Auto-bidding is widely used in advertising systems, serving a diverse range of advertisers. Generative bidding is increasingly gaining traction due to its strong planning capabilities and generalizability. Unlike traditional reinforcement learning-based bidding, generative bidding does not depend on the Markov Decision Process (MDP), thereby exhibiting superior planning performance in long-horizon scenarios. Conditional diffusion modeling approaches have shown significant promise in the field of auto-bidding. However, relying solely on return as the optimality criterion is insufficient to guarantee the generation of truly optimal decision sequences, as it lacks personalized structural information. Moreover, the auto-regressive generation mechanism of diffusion models inherently introduces timeliness risks. To address these challenges, we introduce a novel conditional diffusion modeling approach that integrates expert trajectory guidance with a skip-step sampling strategy to improve generation efficiency. The efficacy of this method has been demonstrated through comprehensive offline experiments and further substantiated by statistically significant outcomes in online A/B testing, yielding an 11.29% increase in conversions and a 12.36% growth in revenue relative to the baseline.
Authors: Sameep Chattopadhyay, Vinay Sutar, Jaya Prakash Champati, Sharayu Moharir
Abstract: This work focuses on Hierarchical Inference (HI) in edge intelligence systems, where a compact Local-ML model on an end-device works in conjunction with a high-accuracy Remote-ML model on an edge-server. HI aims to reduce latency, improve accuracy, and lower bandwidth usage by first using the Local-ML model for inference and offloading to the Remote-ML only when the local inference is likely incorrect. A critical challenge in HI is estimating the likelihood of the local inference being incorrect, especially when data distributions and offloading costs change over time -- a problem we term Hierarchical Inference Learning (HIL). We introduce a novel approach to HIL by modeling the probability of correct inference by the Local-ML as an increasing function of the model's confidence measure, a structure motivated by empirical observations but previously unexploited. We propose two policies, HI-LCB and HI-LCB-lite, based on the Upper Confidence Bound (UCB) framework. We demonstrate that both policies achieve order-optimal regret of $O(\log T)$, a significant improvement over existing HIL policies with $O(T^{2/3})$ regret guarantees. Notably, HI-LCB-lite has an $O(1)$ per-sample computational complexity, making it well-suited for deployment on devices with severe resource limitations. Simulations using real-world datasets confirm that our policies outperform existing state-of-the-art HIL methods.
Authors: Konstantinos Vasili, Zachery T. Dahm, Stylianos Chatzidakis
Abstract: Next generation advanced nuclear reactors are expected to be smaller both in size and power output, relying extensively on fully digital instrumentation and control systems. These reactors will generate a large flow of information in the form of multivariate time series data, conveying simultaneously various non linear cyber physical, process, control, sensor, and operational states. Ensuring data integrity against deception attacks is becoming increasingly important for networked communication and a requirement for safe and reliable operation. Current efforts to address replay attacks, almost universally focus on watermarking or supervised anomaly detection approaches without further identifying and characterizing the root cause of the anomaly. In addition, these approaches rely mostly on synthetic data with uncorrelated Gaussian process and measurement noise and full state feedback or are limited to univariate signals, signal stationarity, linear quadratic regulators, or other linear-time invariant state-space which may fail to capture any unmodeled system dynamics. In the realm of regulated nuclear cyber-physical systems, additional work is needed on characterization of replay attacks and explainability of predictions using real data. Here, we propose an unsupervised explainable AI framework based on a combination of autoencoder and customized windowSHAP algorithm to fully characterize real-time replay attacks, i.e., detection, source identification, timing and type, of increasing complexity during a dynamic time evolving reactor process. The proposed XAI framework was benchmarked on several real world datasets from Purdue's nuclear reactor PUR-1 with up to six signals concurrently being replayed. In all cases, the XAI framework was able to detect and identify the source and number of signals being replayed and the duration of the falsification with 95 percent or better accuracy.
Authors: Bing Han, Feifei Zhao, Dongcheng Zhao, Guobin Shen, Ping Wu, Yu Shi, Yi Zeng
Abstract: Fine-tuning as service injects domain-specific knowledge into large language models (LLMs), while challenging the original alignment mechanisms and introducing safety risks. A series of defense strategies have been proposed for the alignment, fine-tuning, and post-fine-tuning phases, where most post-fine-tuning defenses rely on coarse-grained safety layer mapping. These methods lack a comprehensive consideration of both safety layers and fine-grained neurons, limiting their ability to efficiently balance safety and utility. To address this, we propose the Fine-Grained Safety Neurons (FGSN) with Training-Free Continual Projection method to reduce the fine-tuning safety risks. FGSN inherently integrates the multi-scale interactions between safety layers and neurons, localizing sparser and more precise fine-grained safety neurons while minimizing interference with downstream task neurons. We then project the safety neuron parameters onto safety directions, improving model safety while aligning more closely with human preferences. Extensive experiments across multiple fine-tuned LLM models demonstrate that our method significantly reduce harmfulness scores and attack success rates with minimal parameter modifications, while preserving the model's utility. Furthermore, by introducing a task-specific, multi-dimensional heterogeneous safety neuron cluster optimization mechanism, we achieve continual defense and generalization capability against unforeseen emerging safety concerns.
Authors: Bilal Faye, Hanane Azzag, Mustapha Lebbah
Abstract: Diffusion models have emerged as a leading framework for high-quality image generation, offering stable training and strong performance across diverse domains. However, they remain computationally intensive, particularly during the iterative denoising process. Latent-space models like Stable Diffusion alleviate some of this cost by operating in compressed representations, though at the expense of fine-grained detail. More recent approaches such as Retrieval-Augmented Diffusion Models (RDM) address efficiency by conditioning denoising on similar examples retrieved from large external memory banks. While effective, these methods introduce drawbacks: they require costly storage and retrieval infrastructure, depend on static vision-language models like CLIP for similarity, and lack adaptability during training. We propose the Prototype Diffusion Model (PDM), a method that integrates prototype learning directly into the diffusion process for efficient and adaptive visual conditioning - without external memory. Instead of retrieving reference samples, PDM constructs a dynamic set of compact visual prototypes from clean image features using contrastive learning. These prototypes guide the denoising steps by aligning noisy representations with semantically relevant visual patterns, enabling efficient generation with strong semantic grounding. Experiments show that PDM maintains high generation quality while reducing computational and storage overhead, offering a scalable alternative to retrieval-based conditioning in diffusion models.
Authors: Qingbin Li, Rongkun Xue, Jie Wang, Ming Zhou, Zhi Li, Xiaofeng Ji, Yongqi Wang, Miao Liu, Zheming Yang, Minghui Qiu, Jing Yang
Abstract: Recent advances in Reinforcement Learning with Verified Reward (RLVR) have driven the emergence of more sophisticated cognitive behaviors in large language models (LLMs), thereby enhancing their reasoning capabilities. However, in prior RLVR pipelines, the repeated use of static initial-state sampling drawn exactly from the dataset distribution during each sampling phase produced overly deterministic, low diversity model behavior, which manifested as rapid entropy collapse and hindered sustained performance gains during prolonged training. To address this issue, we introduce CURE (Critical-token-gUided Re concatenation for Entropy-collapse prevention), a two-stage framework that balances exploration and exploitation. Specifically, in the first stage, to deliberately steer the model toward novel yet coherent contexts, we re-generate at high-entropy critical tokens and jointly optimize the original and the branched trajectories. The further comparison with vanilla DAPO shows that the regeneration process achieves a better performance on math reasoning tasks while sustaining a high-level entropy degree for exploration. In the second stage, we continue training with static initial-state sampling by DAPO, intentionally placing the model in a familiar state to gradually strengthen exploitation. Extensive experiments on Qwen-2.5-Math-7B show that, compared to other RLVR methods, CURE achieves a 5% performance gain across six math benchmarks, establishing state-of-the-art performance in both entropy and accuracy. A series of experiments further validate the effectiveness of our approach. Code is available at https://github.com/bytedance/CURE.
Authors: Jayanta Mandi, Ali \.Irfan Mahmuto\u{g}ullar{\i}, Senne Berden, Tias Guns
Abstract: Decision-focused learning (DFL) trains a machine learning (ML) model to predict parameters of an optimization problem, to directly minimize decision regret, i.e., maximize decision quality. Gradient-based DFL requires computing the derivative of the solution to the optimization problem with respect to the predicted parameters. However, for many optimization problems, such as linear programs (LPs), the gradient of the regret with respect to the predicted parameters is zero almost everywhere. Existing gradient-based DFL approaches for LPs try to circumvent this issue in one of two ways: (a) smoothing the LP into a differentiable optimization problem by adding a quadratic regularizer and then minimizing the regret directly or (b) minimizing surrogate losses that have informative (sub)gradients. In this paper, we show that the former approach still results in zero gradients, because even after smoothing the regret remains constant across large regions of the parameter space. To address this, we propose minimizing surrogate losses -- even when a differentiable optimization layer is used and regret can be minimized directly. Our experiments demonstrate that minimizing surrogate losses allows differentiable optimization layers to achieve regret comparable to or better than surrogate-loss based DFL methods. Further, we demonstrate that this also holds for DYS-Net, a recently proposed differentiable optimization technique for LPs, that computes approximate solutions and gradients through operations that can be performed using feedforward neural network layers. Because DYS-Net executes the forward and the backward pass very efficiently, by minimizing surrogate losses using DYS-Net, we are able to attain regret on par with the state-of-the-art while reducing training time by a significant margin.
Authors: Lorenzo Livi
Abstract: We study how gating mechanisms in recurrent neural networks (RNNs) implicitly induce adaptive learning-rate behavior, even when training is carried out with a fixed, global learning rate. This effect arises from the coupling between state-space time scales--parametrized by the gates--and parameter-space dynamics during gradient descent. By deriving exact Jacobians for leaky-integrator and gated RNNs, we obtain a first-order expansion that makes explicit how constant, scalar, and multi-dimensional gates reshape gradient propagation, modulate effective step sizes, and introduce anisotropy in parameter updates. These findings reveal that gates not only control information flow, but also act as data-driven preconditioners that adapt optimization trajectories in parameter space. We further draw formal analogies with learning-rate schedules, momentum, and adaptive methods such as Adam. Empirical simulations corroborate these claims: in several sequence tasks, we show that gates induce lag-dependent effective learning rates and directional concentration of gradient flow, with multi-gate models matching or exceeding the anisotropic structure produced by Adam. These results highlight that optimizer-driven and gate-driven adaptivity are complementary but not equivalent mechanisms. Overall, this work provides a unified dynamical systems perspective on how gating couples state evolution with parameter updates, explaining why gated architectures achieve robust trainability and stability in practice.
Authors: Hansol Lim, Jongseong Brad Choi, Jee Won Lee, Haeseong Jeoung, Minkyu Han
Abstract: We present a hybrid surrogate model for electric vehicle parameter estimation and power consumption. We combine our novel architecture Spectral Parameter Operator built on a Fourier Neural Operator backbone for global context and a differentiable physics module in the forward pass. From speed and acceleration alone, it outputs time-varying motor and regenerative braking efficiencies, as well as aerodynamic drag, rolling resistance, effective mass, and auxiliary power. These parameters drive a physics-embedded estimate of battery power, eliminating any separate physics-residual loss. The modular design lets representations converge to physically meaningful parameters that reflect the current state and condition of the vehicle. We evaluate on real-world logs from a Tesla Model 3, Tesla Model S, and the Kia EV9. The surrogate achieves a mean absolute error of 0.2kW (about 1% of average traction power at highway speeds) for Tesla vehicles and about 0.8kW on the Kia EV9. The framework is interpretable, and it generalizes well to unseen conditions, and sampling rates, making it practical for path optimization, eco-routing, on-board diagnostics, and prognostics health management.
Authors: Emmanouil Kritharakis, Dusan Jakovetic, Antonios Makris, Konstantinos Tserpes
Abstract: Federated Learning (FL) enables collaborative model training across multiple clients without sharing private data. We consider FL scenarios wherein FL clients are subject to adversarial (Byzantine) attacks, while the FL server is trusted (honest) and has a trustworthy side dataset. This may correspond to, e.g., cases where the server possesses trusted data prior to federation, or to the presence of a trusted client that temporarily assumes the server role. Our approach requires only two honest participants, i.e., the server and one client, to function effectively, without prior knowledge of the number of malicious clients. Theoretical analysis demonstrates bounded optimality gaps even under strong Byzantine attacks. Experimental results show that our algorithm significantly outperforms standard and robust FL baselines such as Mean, Trimmed Mean, Median, Krum, and Multi-Krum under various attack strategies including label flipping, sign flipping, and Gaussian noise addition across MNIST, FMNIST, and CIFAR-10 benchmarks using the Flower framework.
Authors: Chuyun Deng, Huiwen Jia
Abstract: We consider contextual linear bandits over networks, a class of sequential decision-making problems where learning occurs simultaneously across multiple locations and the reward distributions share structural similarities while also exhibiting local differences. While classical contextual bandits assume either fully centralized data or entirely isolated learners, much remains unexplored in networked environments when information is partially shared. In this paper, we address this gap by developing two network-aware Upper Confidence Bound (UCB) algorithms, NetLinUCB and Net-SGD-UCB, which enable adaptive information sharing guided by dynamically updated network weights. Our approach decompose learning into global and local components and as a result allow agents to benefit from shared structure without full synchronization. Both algorithms incur lighter communication costs compared to a fully centralized setting as agents only share computed summaries regarding the homogeneous features. We establish regret bounds showing that our methods reduce the learning complexity associated with the shared structure from $O(N)$ to sublinear $O(\sqrt{N})$, where $N$ is the size of the network. The two algorithms reveal complementary strengths: NetLinUCB excels in low-noise regimes with fine-grained heterogeneity, while Net-SGD-UCB is robust to high-dimensional, high-variance contexts. We further demonstrate the effectiveness of our methods across simulated pricing environments compared to standard benchmarks.
Authors: Yishun Lu, Wesley Armour
Abstract: Modern GPUs are equipped with large amounts of high-bandwidth memory, enabling them to support mini-batch sizes of up to tens of thousands of training samples. However, most existing optimizers struggle to perform effectively at such a large batch size. As batch size increases, gradient noise decreases due to averaging over many samples, limiting the ability of first-order methods to escape sharp or suboptimal minima and reach the global minimum. Meanwhile, second-order methods like the natural gradient with Kronecker-Factored Approximate Curvature (KFAC) often require excessively high damping to remain stable at large batch sizes. This high damping effectively washes out the curvature information that gives these methods their advantage, reducing their performance to that of simple gradient descent. In this paper, we introduce Fisher-Orthogonal Projection (FOP), a novel technique that restores the effectiveness of the second-order method at very large batch sizes, enabling scalable training with improved generalization and faster convergence. FOP constructs a variance-aware update direction by leveraging gradients from two sub-batches, enhancing the average gradient with a component of the gradient difference that is orthogonal to the average under the Fisher-metric.
Authors: Jiho Park, Huiwen Jia
Abstract: We introduce a Multi-User Contextual Cascading Bandit model, a new combinatorial bandit framework that captures realistic online advertising scenarios where multiple users interact with sequentially displayed items simultaneously. Unlike classical contextual bandits, MCCB integrates three key structural elements: (i) cascading feedback based on sequential arm exposure, (ii) parallel context sessions enabling selective exploration, and (iii) heterogeneous arm-level rewards. We first propose Upper Confidence Bound with Backward Planning (UCBBP), a UCB-style algorithm tailored to this setting, and prove that it achieves a regret bound of $\widetilde{O}(\sqrt{THN})$ over $T$ episodes, $H$ session steps, and $N$ contexts per episode. Motivated by the fact that many users interact with the system simultaneously, we introduce a second algorithm, termed Active Upper Confidence Bound with Backward Planning (AUCBBP), which shows a strict efficiency improvement in context scaling, i.e., user scaling, with a regret bound of $\widetilde{O}(\sqrt{T+HN})$. We validate our theoretical findings via numerical experiments, demonstrating the empirical effectiveness of both algorithms under various settings.
Authors: Lian Lian, Yilin Li, Song Han, Renzi Meng, Sibo Wang, Ming Wang
Abstract: This study proposes an anomaly detection method based on the Transformer architecture with integrated multiscale feature perception, aiming to address the limitations of temporal modeling and scale-aware feature representation in cloud service environments. The method first employs an improved Transformer module to perform temporal modeling on high-dimensional monitoring data, using a self-attention mechanism to capture long-range dependencies and contextual semantics. Then, a multiscale feature construction path is introduced to extract temporal features at different granularities through downsampling and parallel encoding. An attention-weighted fusion module is designed to dynamically adjust the contribution of each scale to the final decision, enhancing the model's robustness in anomaly pattern modeling. In the input modeling stage, standardized multidimensional time series are constructed, covering core signals such as CPU utilization, memory usage, and task scheduling states, while positional encoding is used to strengthen the model's temporal awareness. A systematic experimental setup is designed to evaluate performance, including comparative experiments and hyperparameter sensitivity analysis, focusing on the impact of optimizers, learning rates, anomaly ratios, and noise levels. Experimental results show that the proposed method outperforms mainstream baseline models in key metrics, including precision, recall, AUC, and F1-score, and maintains strong stability and detection performance under various perturbation conditions, demonstrating its superior capability in complex cloud environments.
Authors: Preston Fu, Oleh Rybkin, Zhiyuan Zhou, Michal Nauman, Pieter Abbeel, Sergey Levine, Aviral Kumar
Abstract: As models grow larger and training them becomes expensive, it becomes increasingly important to scale training recipes not just to larger models and more data, but to do so in a compute-optimal manner that extracts maximal performance per unit of compute. While such scaling has been well studied for language modeling, reinforcement learning (RL) has received less attention in this regard. In this paper, we investigate compute scaling for online, value-based deep RL. These methods present two primary axes for compute allocation: model capacity and the update-to-data (UTD) ratio. Given a fixed compute budget, we ask: how should resources be partitioned across these axes to maximize sample efficiency? Our analysis reveals a nuanced interplay between model size, batch size, and UTD. In particular, we identify a phenomenon we call TD-overfitting: increasing the batch quickly harms Q-function accuracy for small models, but this effect is absent in large models, enabling effective use of large batch size at scale. We provide a mental model for understanding this phenomenon and build guidelines for choosing batch size and UTD to optimize compute usage. Our findings provide a grounded starting point for compute-optimal scaling in deep RL, mirroring studies in supervised learning but adapted to TD learning.
Authors: Rui Wang, Qianguo Sun, Chao Song, Junlong Wu, Tianrong Chen, Zhiyun Zeng, Yu Li
Abstract: DPO (Direct Preference Optimization) has become a widely used offline preference optimization algorithm due to its simplicity and training stability. However, DPO is prone to overfitting and collapse. To address these challenges, we propose Linear Preference Optimization (LPO), a novel alignment framework featuring three key innovations. First, we introduce gradient decoupling by replacing the log-sigmoid function with an absolute difference loss, thereby isolating the optimization dynamics. Second, we improve stability through an offset constraint combined with a positive regularization term to preserve the chosen response quality. Third, we implement controllable rejection suppression using gradient separation with straightforward estimation and a tunable coefficient that linearly regulates the descent of the rejection probability. Through extensive experiments, we demonstrate that LPO consistently improves performance on various tasks, including general text tasks, math tasks, and text-to-speech (TTS) tasks. These results establish LPO as a robust and tunable paradigm for preference alignment, and we release the source code, models, and training data publicly.
Authors: Sami Alabed, Dominik Grewe, Norman Alexander Rink, Masha Samsikova, Timur Sitdikov, Agnieszka Swietlik, Dimitrios Vytiniotis, Daniel Belov
Abstract: Partitioning large machine learning models across distributed accelerator systems is a complex process, requiring a series of interdependent decisions that are further complicated by internal sharding ambiguities. Consequently, existing auto-partitioners often suffer from out-of-memory errors or are prohibitively slow when exploring the exponentially large space of possible partitionings. To mitigate this, they artificially restrict the search space, but this approach frequently yields infeasible solutions that violate device memory constraints or lead to sub-optimal performance. We propose a system that combines a novel static compiler analysis with a Monte Carlo Tree Search. Our analysis constructs an efficient decision space by identifying (i) tensor dimensions requiring identical sharding, and (ii) partitioning "conflicts" that require resolution. Our system significantly outperforms state-of-the-art industrial methods across diverse hardware platforms and model architectures, discovering previously unknown, superior solutions, and the process is fully automated even for complex and large models.
Authors: Omkar Tupe, Max Hartman, Lav R. Varshney, Saurav Prakash
Abstract: We consider federated learning of linearly-parameterized nonlinear systems. We establish theoretical guarantees on the effectiveness of federated nonlinear system identification compared to centralized approaches, demonstrating that the convergence rate improves as the number of clients increases. Although the convergence rates in the linear and nonlinear cases differ only by a constant, this constant depends on the feature map $\phi$, which can be carefully chosen in the nonlinear setting to increase excitation and improve performance. We experimentally validate our theory in physical settings where client devices are driven by i.i.d. control inputs and control policies exhibiting i.i.d. random perturbations, ensuring non-active exploration. Experiments use trajectories from nonlinear dynamical systems characterized by real-analytic feature functions, including polynomial and trigonometric components, representative of physical systems including pendulum and quadrotor dynamics. We analyze the convergence behavior of the proposed method under varying noise levels and data distributions. Results show that federated learning consistently improves convergence of any individual client as the number of participating clients increases.
Authors: Shaswati Saha, Sourajit Saha, Manas Gaur, Tejas Gokhale
Abstract: Concerns about text-to-image (T2I) generative models infringing on privacy, copyright, and safety have led to the development of Concept Erasure Techniques (CETs). The goal of an effective CET is to prohibit the generation of undesired "target" concepts specified by the user, while preserving the ability to synthesize high-quality images of the remaining concepts. In this work, we demonstrate that CETs can be easily circumvented and present several side effects of concept erasure. For a comprehensive measurement of the robustness of CETs, we present Side Effect Evaluation (SEE), an evaluation benchmark that consists of hierarchical and compositional prompts that describe objects and their attributes. This dataset and our automated evaluation pipeline quantify side effects of CETs across three aspects: impact on neighboring concepts, evasion of targets, and attribute leakage. Our experiments reveal that CETs can be circumvented by using superclass-subclass hierarchy and semantically similar prompts, such as compositional variants of the target. We show that CETs suffer from attribute leakage and counterintuitive phenomena of attention concentration or dispersal. We release our dataset, code, and evaluation tools to aid future work on robust concept erasure.
Authors: Lei Bai, Zhongrui Cai, Yuhang Cao, Maosong Cao, Weihan Cao, Chiyu Chen, Haojiong Chen, Kai Chen, Pengcheng Chen, Ying Chen, Yongkang Chen, Yu Cheng, Pei Chu, Tao Chu, Erfei Cui, Ganqu Cui, Long Cui, Ziyun Cui, Nianchen Deng, Ning Ding, Nanqing Dong, Peijie Dong, Shihan Dou, Sinan Du, Haodong Duan, Caihua Fan, Ben Gao, Changjiang Gao, Jianfei Gao, Songyang Gao, Yang Gao, Zhangwei Gao, Jiaye Ge, Qiming Ge, Lixin Gu, Yuzhe Gu, Aijia Guo, Qipeng Guo, Xu Guo, Conghui He, Junjun He, Yili Hong, Siyuan Hou, Caiyu Hu, Hanglei Hu, Jucheng Hu, Ming Hu, Zhouqi Hua, Haian Huang, Junhao Huang, Xu Huang, Zixian Huang, Zhe Jiang, Lingkai Kong, Linyang Li, Peiji Li, Pengze Li, Shuaibin Li, Tianbin Li, Wei Li, Yuqiang Li, Dahua Lin, Junyao Lin, Tianyi Lin, Zhishan Lin, Hongwei Liu, Jiangning Liu, Jiyao Liu, Junnan Liu, Kai Liu, Kaiwen Liu, Kuikun Liu, Shichun Liu, Shudong Liu, Wei Liu, Xinyao Liu, Yuhong Liu, Zhan Liu, Yinquan Lu, Haijun Lv, Hongxia Lv, Huijie Lv, Qitan Lv, Ying Lv, Chengqi Lyu, Chenglong Ma, Jianpeng Ma, Ren Ma, Runmin Ma, Runyuan Ma, Xinzhu Ma, Yichuan Ma, Zihan Ma, Sixuan Mi, Junzhi Ning, Wenchang Ning, Xinle Pang, Jiahui Peng, Runyu Peng, Yu Qiao, Jiantao Qiu, Xiaoye Qu, Yuan Qu, Yuchen Ren, Fukai Shang, Wenqi Shao, Junhao Shen, Shuaike Shen, Chunfeng Song, Demin Song, Diping Song, Chenlin Su, Weijie Su, Weigao Sun, Yu Sun, Qian Tan, Cheng Tang, Huanze Tang, Kexian Tang, Shixiang Tang, Jian Tong, Aoran Wang, Bin Wang, Dong Wang, Lintao Wang, Rui Wang, Weiyun Wang, Wenhai Wang, Jiaqi Wang, Yi Wang, Ziyi Wang, Ling-I Wu, Wen Wu, Yue Wu, Zijian Wu, Linchen Xiao, Shuhao Xing, Chao Xu, Huihui Xu, Jun Xu, Ruiliang Xu, Wanghan Xu, GanLin Yang, Yuming Yang, Haochen Ye, Jin Ye, Shenglong Ye, Jia Yu, Jiashuo Yu, Jing Yu, Fei Yuan, Yuhang Zang, Bo Zhang, Chao Zhang, Chen Zhang, Hongjie Zhang, Jin Zhang, Qiaosheng Zhang, Qiuyinzhe Zhang, Songyang Zhang, Taolin Zhang, Wenlong Zhang, Wenwei Zhang, Yechen Zhang, Ziyang Zhang, Haiteng Zhao, Qian Zhao, Xiangyu Zhao, Xiangyu Zhao, Bowen Zhou, Dongzhan Zhou, Peiheng Zhou, Yuhao Zhou, Yunhua Zhou, Dongsheng Zhu, Lin Zhu, Yicheng Zou
Abstract: In recent years, a plethora of open-source foundation models have emerged, achieving remarkable progress in some widely attended fields, with performance being quite close to that of closed-source models. However, in high-value but more challenging scientific professional fields, either the fields still rely on expert models, or the progress of general foundation models lags significantly compared to those in popular areas, far from sufficient for transforming scientific research and leaving substantial gap between open-source models and closed-source models in these scientific domains. To mitigate this gap and explore a step further toward Artificial General Intelligence (AGI), we introduce Intern-S1, a specialized generalist equipped with general understanding and reasoning capabilities with expertise to analyze multiple science modal data. Intern-S1 is a multimodal Mixture-of-Experts (MoE) model with 28 billion activated parameters and 241 billion total parameters, continually pre-trained on 5T tokens, including over 2.5T tokens from scientific domains. In the post-training stage, Intern-S1 undergoes offline and then online reinforcement learning (RL) in InternBootCamp, where we propose Mixture-of-Rewards (MoR) to synergize the RL training on more than 1000 tasks simultaneously. Through integrated innovations in algorithms, data, and training systems, Intern-S1 achieved top-tier performance in online RL training. On comprehensive evaluation benchmarks, Intern-S1 demonstrates competitive performance on general reasoning tasks among open-source models and significantly outperforms open-source models in scientific domains, surpassing closed-source state-of-the-art models in professional tasks, such as molecular synthesis planning, reaction condition prediction, predicting thermodynamic stabilities for crystals. Our models are available at https://huggingface.co/internlm/Intern-S1.
Authors: Xiaojuan Tang, Fanxu Meng, Pingzhi Tang, Yuxuan Wang, Di Yin, Xing Sun, Muhan Zhang
Abstract: Multi-Head Latent Attention (MLA), introduced in DeepSeek-V2, compresses key-value states into a low-rank latent vector, caching only this vector to reduce memory. In tensor parallelism (TP), however, attention heads are computed across multiple devices, and each device must load the full cache, eroding the advantage of MLA over Grouped Query Attention (GQA). We propose Tensor-Parallel Latent Attention (TPLA): a scheme that partitions both the latent representation and each head's input dimension across devices, performs attention independently per shard, and then combines results with an all-reduce. TPLA preserves the benefits of a compressed KV cache while unlocking TP efficiency. Unlike Grouped Latent Attention (GLA), every head in TPLA still leverages the full latent representation, maintaining stronger representational capacity. TPLA is drop-in compatible with models pre-trained using MLA: it supports MLA-style prefilling and enables efficient tensor-parallel decoding without retraining. Applying simple orthogonal transforms -- e.g., the Hadamard transform or PCA -- before TP slicing further mitigates cross-shard interference, yielding minimal accuracy degradation. By reducing the per-device KV cache for DeepSeek-V3 and Kimi-K2, we achieve 1.79x and 1.93x speedups, respectively, at a 32K-token context length while maintaining performance on commonsense and LongBench benchmarks. TPLA can be implemented with FlashAttention-3, enabling practical end-to-end acceleration.
Authors: Huichi Zhou, Yihang Chen, Siyuan Guo, Xue Yan, Kin Hei Lee, Zihan Wang, Ka Yiu Lee, Guchun Zhang, Kun Shao, Linyi Yang, Jun Wang
Abstract: In this paper, we introduce a novel learning paradigm for Adaptive Large Language Model (LLM) agents that eliminates the need for fine-tuning the underlying LLMs. Existing approaches are often either rigid, relying on static, handcrafted reflection workflows, or computationally intensive, requiring gradient updates of LLM model parameters. In contrast, our method enables low-cost continual adaptation via memory-based online reinforcement learning. We formalise this as a Memory-augmented Markov Decision Process (M-MDP), equipped with a neural case-selection policy to guide action decisions. Past experiences are stored in an episodic memory, either differentiable or non-parametric. The policy is continually updated based on environmental feedback through a memory rewriting mechanism, whereas policy improvement is achieved through efficient memory reading (retrieval). We instantiate our agent model in the deep research setting, namely \emph{Memento}, which attains top-1 on GAIA validation ($87.88\%$ Pass@$3$) and $79.40\%$ on the test set. It reaches $66.6\%$ F1 and $80.4\%$ PM on the DeepResearcher dataset, outperforming the state-of-the-art training-based method, while case-based memory adds $4.7\%$ to $9.6\%$ absolute points on out-of-distribution tasks. Our approach offers a scalable and efficient pathway for developing generalist LLM agents capable of continuous, real-time learning without gradient updates, advancing machine learning towards open-ended skill acquisition and deep research scenarios. The code is available at https://github.com/Agent-on-the-Fly/Memento.
Authors: Nathan Bigaud, Vincent Cabeli, Meltem G\"urel, Arthur Pignet, John Klein, Gilles Wainrib, Eric Durand
Abstract: While large language models (LLMs) are rapidly advancing scientific research, they continue to struggle with core biological reasoning tasks essential for translational and biomedical discovery. To address this limitation, we created and curated eight comprehensive benchmark datasets comprising over 300,000 verifiable question-and-answer pairs, each targeting critical challenges in drug discovery including target druggability, modality suitability, and drug perturbation effects. Using this resource, we developed the OwkinZero models by post-training open-source LLMs through a Reinforcement Learning from Verifiable Rewards strategy. Our results demonstrate that specialized 8-32B OwkinZero models substantially outperform larger, state-of-the-art commercial LLMs on these biological benchmarks. Remarkably, we uncover evidence of a key aspect of generalization: specialist models trained on a single task consistently outperform their base models on previously unseen tasks. This generalization effect is further amplified in our comprehensive OwkinZero models, which were trained on a mixture of datasets and achieve even broader cross-task improvements. This study represents a significant step toward addressing the biological reasoning blind spot in current LLMs, demonstrating that targeted reinforcement learning on carefully curated data can unlock generalizable performance in specialized models, thereby accelerating AI-driven biological discovery.
Authors: Xingchao Peng, Zijun Huang, Yizhe Zhu, Kate Saenko
Abstract: Federated learning improves data privacy and efficiency in machine learning performed over networks of distributed devices, such as mobile phones, IoT and wearable devices, etc. Yet models trained with federated learning can still fail to generalize to new devices due to the problem of domain shift. Domain shift occurs when the labeled data collected by source nodes statistically differs from the target node's unlabeled data. In this work, we present a principled approach to the problem of federated domain adaptation, which aims to align the representations learned among the different nodes with the data distribution of the target node. Our approach extends adversarial adaptation techniques to the constraints of the federated setting. In addition, we devise a dynamic attention mechanism and leverage feature disentanglement to enhance knowledge transfer. Empirically, we perform extensive experiments on several image and text classification tasks and show promising results under unsupervised federated domain adaptation setting.
Authors: Mang Li
Abstract: Unexpected advertising items in sponsored search may reduce users' reliance on organic search, resulting in hidden cost for the e-commerce platform. To address this problem and promote sustainable growth, we propose a dynamic reserve price design that incorporates the hidden cost into the auction mechanism to determine whether to sell the traffic, thereby ensuring a balanced relationship between revenue and user experience. Our dynamic reserve price design framework optimizes traffic sales by minimizing impacts on user experience while maintaining long-term incentives for advertisers to reveal their valuations truthfully. Furthermore, we introduce a distributed algorithm capable of computing reserve prices with billion-scale data in the production environment. Experiments involving offline evaluations and online A/B testing demonstrate that this method is simple and efficient, making it suitable for use in industrial production. This method has already been fully deployed in the production environment.
Authors: Sannara EK, Fran\c{c}ois Portet, Philippe Lalanda
Abstract: Human Activity Recognition (HAR) on mobile devices has been demonstrated to be possible using neural models trained on data collected from the device's inertial measurement units. These models have used Convolutional Neural Networks (CNNs), Long Short-Term Memory (LSTMs), Transformers or a combination of these to achieve state-of-the-art results with real-time performance. However, these approaches have not been extensively evaluated in real-world situations where the input data may be different from the training data. This paper highlights the issue of data heterogeneity in machine learning applications and how it can hinder their deployment in pervasive settings. To address this problem, we propose and publicly release the code of two sensor-wise Transformer architectures called HART and MobileHART for Human Activity Recognition Transformer. Our experiments on several publicly available datasets show that these HART architectures outperform previous architectures with fewer floating point operations and parameters than conventional Transformers. The results also show they are more robust to changes in mobile position or device brand and hence better suited for the heterogeneous environments encountered in real-life settings. Finally, the source code has been made publicly available.
Authors: Jiayang Ren, Ningning You, Kaixun Hua, Chaojie Ji, Yankai Cao
Abstract: This paper presents a practical global optimization algorithm for the K-center clustering problem, which aims to select K samples as the cluster centers to minimize the maximum within-cluster distance. This algorithm is based on a reduced-space branch and bound scheme and guarantees convergence to the global optimum in a finite number of steps by only branching on the regions of centers. To improve efficiency, we have designed a two-stage decomposable lower bound, the solution of which can be derived in a closed form. In addition, we also propose several acceleration techniques to narrow down the region of centers, including bounds tightening, sample reduction, and parallelization. Extensive studies on synthetic and real-world datasets have demonstrated that our algorithm can solve the K-center problems to global optimal within 4 hours for ten million samples in the serial mode and one billion samples in the parallel mode. Moreover, compared with the state-of-the-art heuristic methods, the global optimum obtained by our algorithm can averagely reduce the objective function by 25.8% on all the synthetic and real-world datasets.
Authors: Ren Wang, Yuxuan Li, Can Chen, Dakuo Wang, Jinjun Xiong, Pin-Yu Chen, Sijia Liu, Mohammad Shahidehpour, Alfred Hero
Abstract: Adversarial robustness is a critical measure of a neural network's ability to withstand adversarial attacks at inference time. While robust training techniques have improved defenses against individual $\ell_p$-norm attacks (e.g., $\ell_2$ or $\ell_\infty$), models remain vulnerable to diversified $\ell_p$ perturbations. To address this challenge, we propose a novel Robust Mode Connectivity (RMC)-oriented adversarial defense framework comprising two population-based learning phases. In Phase I, RMC searches the parameter space between two pre-trained models to construct a continuous path containing models with high robustness against multiple $\ell_p$ attacks. To improve efficiency, we introduce a Self-Robust Mode Connectivity (SRMC) module that accelerates endpoint generation in RMC. Building on RMC, Phase II presents RMC-based optimization, where RMC modules are composed to further enhance diversified robustness. To increase Phase II efficiency, we propose Efficient Robust Mode Connectivity (ERMC), which leverages $\ell_1$- and $\ell_\infty$-adversarially trained models to achieve robustness across a broad range of $p$-norms. An ensemble strategy is employed to further boost ERMC's performance. Extensive experiments across diverse datasets and architectures demonstrate that our methods significantly improve robustness against $\ell_\infty$, $\ell_2$, $\ell_1$, and hybrid attacks. Code is available at https://github.com/wangren09/MCGR.
Authors: Tingwei Zhang, Rishi Jha, Eugene Bagdasaryan, Vitaly Shmatikov
Abstract: Multi-modal embeddings encode texts, images, thermal images, sounds, and videos into a single embedding space, aligning representations across different modalities (e.g., associate an image of a dog with a barking sound). In this paper, we show that multi-modal embeddings can be vulnerable to an attack we call "adversarial illusions." Given an image or a sound, an adversary can perturb it to make its embedding close to an arbitrary, adversary-chosen input in another modality. These attacks are cross-modal and targeted: the adversary can align any image or sound with any target of his choice. Adversarial illusions exploit proximity in the embedding space and are thus agnostic to downstream tasks and modalities, enabling a wholesale compromise of current and future tasks, as well as modalities not available to the adversary. Using ImageBind and AudioCLIP embeddings, we demonstrate how adversarially aligned inputs, generated without knowledge of specific downstream tasks, mislead image generation, text generation, zero-shot classification, and audio retrieval. We investigate transferability of illusions across different embeddings and develop a black-box version of our method that we use to demonstrate the first adversarial alignment attack on Amazon's commercial, proprietary Titan embedding. Finally, we analyze countermeasures and evasion attacks.
Authors: Ding Huang, Jian Huang, Ting Li, Guohao Shen
Abstract: We propose a conditional stochastic interpolation (CSI) method for learning conditional distributions. CSI is based on estimating probability flow equations or stochastic differential equations that transport a reference distribution to the target conditional distribution. This is achieved by first learning the conditional drift and score functions based on CSI, which are then used to construct a deterministic process governed by an ordinary differential equation or a diffusion process for conditional sampling. In our proposed approach, we incorporate an adaptive diffusion term to address the instability issues arising in the diffusion process. We derive explicit expressions of the conditional drift and score functions in terms of conditional expectations, which naturally lead to an nonparametric regression approach to estimating these functions. Furthermore, we establish nonasymptotic error bounds for learning the target conditional distribution. We illustrate the application of CSI on image generation using a benchmark image dataset.
Authors: M. Cerezo, Martin Larocca, Diego Garc\'ia-Mart\'in, N. L. Diaz, Paolo Braccia, Enrico Fontana, Manuel S. Rudolph, Pablo Bermejo, Aroosa Ijaz, Supanut Thanasilp, Eric R. Anschuetz, Zo\"e Holmes
Abstract: A large amount of effort has recently been put into understanding the barren plateau phenomenon. In this perspective article, we face the increasingly loud elephant in the room and ask a question that has been hinted at by many but not explicitly addressed: Can the structure that allows one to avoid barren plateaus also be leveraged to efficiently simulate the loss classically? We collect evidence-on a case-by-case basis-that many commonly used models whose loss landscapes avoid barren plateaus can also admit classical simulation, provided that one can collect some classical data from quantum devices during an initial data acquisition phase. This follows from the observation that barren plateaus result from a curse of dimensionality, and that current approaches for solving them end up encoding the problem into some small, classically simulable, subspaces. Thus, while stressing that quantum computers can be essential for collecting data, our analysis sheds doubt on the information processing capabilities of many parametrized quantum circuits with provably barren plateau-free landscapes. We end by discussing the (many) caveats in our arguments including the limitations of average case arguments, the role of smart initializations, models that fall outside our assumptions, the potential for provably superpolynomial advantages and the possibility that, once larger devices become available, parametrized quantum circuits could heuristically outperform our analytic expectations.
Authors: Roi Naveiro, Becky Tang
Abstract: Bayesian Optimization (BO) is a powerful method for optimizing black-box functions by combining prior knowledge with ongoing function evaluations. BO constructs a probabilistic surrogate model of the objective function given the covariates, which is in turn used to inform the selection of future evaluation points through an acquisition function. For smooth continuous search spaces, Gaussian Processes (GPs) are commonly used as the surrogate model as they offer analytical access to posterior predictive distributions, thus facilitating the computation and optimization of acquisition functions. However, in complex scenarios involving optimization over categorical or mixed covariate spaces, GPs may not be ideal. This paper introduces Simulation Based Bayesian Optimization (SBBO) as a novel approach to optimizing acquisition functions that only requires sampling-based access to posterior predictive distributions. SBBO allows the use of surrogate probabilistic models tailored for combinatorial spaces with discrete variables. Any Bayesian model in which posterior inference is carried out through Markov chain Monte Carlo can be selected as the surrogate model in SBBO. We demonstrate empirically the effectiveness of SBBO using various choices of surrogate models in applications involving combinatorial optimization.
Authors: Ashok Khanna, Olivier Francon, Risto Miikkulainen
Abstract: Diabetes, a chronic condition that impairs how the body turns food into energy, i.e. blood glucose, affects 38 million people in the US alone. The standard treatment is to supplement carbohydrate intake with an artificial pancreas, i.e. a continuous insulin pump (basal shots), as well as occasional insulin injections (bolus shots). The goal of the treatment is to keep blood glucose at the center of an acceptable range, as measured through a continuous glucose meter. A secondary goal is to minimize injections, which are unpleasant and difficult for some patients to implement. In this study, neuroevolution was used to discover an optimal strategy for the treatment. Based on a dataset of 30 days of treatment and measurements of a single patient, a random forest was first trained to predict future glucose levels. A neural network was then evolved to prescribe carbohydrates, basal pumping levels, and bolus injections. Evolution discovered a Pareto front that reduced deviation from the target and number of injections compared to the original data, thus improving patients' quality of life. To make the system easier to adopt, a language interface was developed with a large language model. Thus, these technologies not only improve patient care but also adoption in a broader population.
Authors: Ayan Banerjee, Nityanand Mathur, Josep Llad\'os, Umapada Pal, Anjan Dutta
Abstract: Generating VectorArt from text prompts is a challenging vision task, requiring diverse yet realistic depictions of the seen as well as unseen entities. However, existing research has been mostly limited to the generation of single objects, rather than comprehensive scenes comprising multiple elements. In response, this work introduces SVGCraft, a novel end-to-end framework for the creation of vector graphics depicting entire scenes from textual descriptions. Utilizing a pre-trained LLM for layout generation from text prompts, this framework introduces a technique for producing masked latents in specified bounding boxes for accurate object placement. It introduces a fusion mechanism for integrating attention maps and employs a diffusion U-Net for coherent composition, speeding up the drawing process. The resulting SVG is optimized using a pre-trained encoder and LPIPS loss with opacity modulation to maximize similarity. Additionally, this work explores the potential of primitive shapes in facilitating canvas completion in constrained environments. Through both qualitative and quantitative assessments, SVGCraft is demonstrated to surpass prior works in abstraction, recognizability, and detail, as evidenced by its performance metrics (CLIP-T: 0.4563, Cosine Similarity: 0.6342, Confusion: 0.66, Aesthetic: 6.7832). The code will be available at https://github.com/ayanban011/SVGCraft.
Authors: Noah Lewis, Jean Luca Bez, Surendra Byna
Abstract: Growing interest in Artificial Intelligence (AI) has resulted in a surge in demand for faster methods of Machine Learning (ML) model training and inference. This demand for speed has prompted the use of high performance computing (HPC) systems that excel in managing distributed workloads. Because data is the main fuel for AI applications, the performance of the storage and I/O subsystem of HPC systems is critical. In the past, HPC applications accessed large portions of data written by simulations or experiments or ingested data for visualizations or analysis tasks. ML workloads perform small reads spread across a large number of random files. This shift of I/O access patterns poses several challenges to modern parallel storage systems. In this paper, we survey I/O in ML applications on HPC systems, and target literature within a 6-year time window from 2019 to 2024. We define the scope of the survey, provide an overview of the common phases of ML, review available profilers and benchmarks, examine the I/O patterns encountered during offline data preparation, training, and inference, and explore I/O optimizations utilized in modern ML frameworks and proposed in recent literature. Lastly, we seek to expose research gaps that could spawn further R&D
Authors: Jun Zhuang, Jack Cunningham, Chaowen Guan
Abstract: In the era of noisy intermediate-scale quantum (NISQ), variational quantum circuits (VQCs) have been widely applied in various domains, demonstrating the potential advantages of quantum circuits over classical models. Similar to classic models, VQCs can be optimized by various gradient-based methods. However, the optimization may get stuck in barren plateaus initially or trapped in saddle points during training. These gradient-related issues can severely impact the trainability of VQCs. In this work, we propose a strategy that regularizes model parameters with prior knowledge of the training data and Gaussian noise diffusion. We conduct ablation studies to verify the effectiveness of our strategy across four public datasets and demonstrate that our method can improve the trainability of VQCs against the above-mentioned gradient issues.
Authors: Jiancong Xiao, Ziniu Li, Xingyu Xie, Emily Getzen, Cong Fang, Qi Long, Weijie J. Su
Abstract: Accurately aligning large language models (LLMs) with human preferences is crucial for informing fair, economically sound, and statistically efficient decision-making processes. However, we argue that the predominant approach for aligning LLMs with human preferences through a reward model -- reinforcement learning from human feedback (RLHF) -- suffers from an inherent algorithmic bias due to its Kullback--Leibler-based regularization in optimization. In extreme cases, this bias could lead to a phenomenon we term preference collapse, where minority preferences are virtually disregarded. To mitigate this algorithmic bias, we introduce preference matching (PM) RLHF, a novel approach that provably aligns LLMs with the preference distribution of the reward model under the Bradley--Terry--Luce/Plackett--Luce model. Central to our approach is a PM regularizer that takes the form of the negative logarithm of the LLM's policy probability distribution over responses, which helps the LLM balance response diversification and reward maximization. Notably, we obtain this regularizer by solving an ordinary differential equation that is necessary for the PM property. For practical implementation, we introduce a conditional variant of PM RLHF that is tailored to natural language generation. Finally, we empirically validate the effectiveness of conditional PM RLHF through experiments on the OPT and Llama-family models, demonstrating a 29% to 41% improvement in alignment with human preferences, as measured by a certain metric, compared to standard RLHF.
Authors: Arvi Jonnarth, Ola Johansson, Jie Zhao, Michael Felsberg
Abstract: Coverage path planning (CPP) is the problem of finding a path that covers the entire free space of a confined area, with applications ranging from robotic lawn mowing to search-and-rescue. While for known environments, offline methods can find provably complete paths, and in some cases optimal solutions, unknown environments need to be planned online during mapping. We investigate the suitability of continuous-space reinforcement learning (RL) for this challenging problem, and propose a computationally feasible egocentric map representation based on frontiers, as well as a novel reward term based on total variation to promote complete coverage. Compared to existing classical methods, this approach allows for a flexible path space, and enables the agent to adapt to specific environment characteristics. Meanwhile, the deployment of RL models on real robot systems is difficult. Training from scratch may be infeasible due to slow convergence times, while transferring from simulation to reality, i.e. sim-to-real transfer, is a key challenge in itself. We bridge the sim-to-real gap through a semi-virtual environment, including a real robot and real-time aspects, while utilizing a simulated sensor and obstacles to enable environment randomization and automated episode resetting. We investigate what level of fine-tuning is needed for adapting to a realistic setting. Through extensive experiments, we show that our approach surpasses the performance of both previous RL-based approaches and highly specialized methods across multiple CPP variations in simulation. Meanwhile, our method successfully transfers to a real robot. Our code implementation can be found online.
Authors: Krishna Kanth Nakka, Ahmed Frikha, Ricardo Mendes, Xue Jiang, Xuebing Zhou
Abstract: The latest and most impactful advances in large models stem from their increased size. Unfortunately, this translates into an improved memorization capacity, raising data privacy concerns. Specifically, it has been shown that models can output personal identifiable information (PII) contained in their training data. However, reported PIII extraction performance varies widely, and there is no consensus on the optimal methodology to evaluate this risk, resulting in underestimating realistic adversaries. In this work, we empirically demonstrate that it is possible to improve the extractability of PII by over ten-fold by grounding the prefix of the manually constructed extraction prompt with in-domain data. Our approach, PII-Compass, achieves phone number extraction rates of 0.92%, 3.9%, and 6.86% with 1, 128, and 2308 queries, respectively, i.e., the phone number of 1 person in 15 is extractable.
Authors: Zhenhua Xu, Qichen Liu, Zhebo Wang, Wenpeng Xing, Dezhang Kong, Mohan Li, Meng Han
Abstract: Backdoor-based fingerprinting has emerged as an effective technique for tracing the ownership of large language models. However, in real-world deployment scenarios, developers often instantiate multiple downstream models from a shared base model, and applying fingerprinting to each variant individually incurs prohibitive computational overhead. While inheritance-based approaches -- where fingerprints are embedded into the base model and expected to persist through fine-tuning -- appear attractive, they suffer from three key limitations: late-stage fingerprinting, fingerprint instability, and interference with downstream adaptation. To address these challenges, we propose a novel mechanism called the Fingerprint Vector. Our method first embeds a fingerprint into the base model via backdoor-based fine-tuning, then extracts a task-specific parameter delta as a fingerprint vector by computing the difference between the fingerprinted and clean models. This vector can be directly added to any structurally compatible downstream model, allowing the fingerprint to be transferred post hoc without additional fine-tuning. Extensive experiments show that Fingerprint Vector achieves comparable or superior performance to direct injection across key desiderata. It maintains strong effectiveness across diverse model architectures as well as mainstream downstream variants within the same family. It also preserves harmlessness and robustness in most cases. Even when slight robustness degradation is observed, the impact remains within acceptable bounds and is outweighed by the scalability benefits of our approach.
Authors: Tetiana Parshakova, Trevor Hastie, Stephen Boyd
Abstract: We examine a special case of the multilevel factor model, with covariance given by multilevel low rank (MLR) matrix~\cite{parshakova2023factor}. We develop a novel, fast implementation of the expectation-maximization algorithm, tailored for multilevel factor models, to maximize the likelihood of the observed data. This method accommodates any hierarchical structure and maintains linear time and storage complexities per iteration. This is achieved through a new efficient technique for computing the inverse of the positive definite MLR matrix. We show that the inverse of positive definite MLR matrix is also an MLR matrix with the same sparsity in factors, and we use the recursive Sherman-Morrison-Woodbury matrix identity to obtain the factors of the inverse. Additionally, we present an algorithm that computes the Cholesky factorization of an expanded matrix with linear time and space complexities, yielding the covariance matrix as its Schur complement. This paper is accompanied by an open-source package that implements the proposed methods.
Authors: Chenxu Yang, Ruipeng Jia, Naibin Gu, Zheng Lin, Siyuan Chen, Chao Pang, Weichong Yin, Yu Sun, Hua Wu, Weiping Wang
Abstract: DPO is an effective preference optimization algorithm. However, the DPO-tuned models tend to overfit on the dispreferred samples, manifested as overly long generations lacking diversity. While recent regularization approaches have endeavored to alleviate this issue by modifying the objective function, they achieved that at the cost of alignment performance degradation. In this paper, we innovatively incorporate regularization from the perspective of weight updating to curb alignment overfitting. Through the pilot experiment, we discovered that there exists a positive correlation between overfitting and the hyperspherical energy fluctuation. Hence, we introduce orthogonal finetuning for DPO via a weight-Rotated Preference Optimization (RoPO) method, which merely conducts rotational and magnitude-stretching updates on the weight parameters to maintain the hyperspherical energy invariant, thereby preserving the knowledge encoded in the angle between neurons. Extensive experiments demonstrate that our model aligns perfectly with human preferences while retaining the original expressive capacity using only 0.0086% of the trainable parameters, suggesting an effective regularization against overfitting. Specifically, RoPO outperforms DPO by up to 10 points on MT-Bench and by up to 2.8 points on AlpacaEval 2, while enhancing the generation diversity by an average of 6 points.
Authors: Harish Karthikeyan, Antigoni Polychroniadou
Abstract: Our work aims to minimize interaction in secure computation due to the high cost and challenges associated with communication rounds, particularly in scenarios with many clients. In this work, we revisit the problem of secure aggregation in the single-server setting where a single evaluation server can securely aggregate client-held individual inputs. Our key contribution is the introduction of One-shot Private Aggregation ($\mathsf{OPA}$) where clients speak only once (or even choose not to speak) per aggregation evaluation. Since each client communicates only once per aggregation, this simplifies managing dropouts and dynamic participation, contrasting with multi-round protocols and aligning with plaintext secure aggregation, where clients interact only once. We construct $\mathsf{OPA}$ based on LWR, LWE, class groups, DCR and demonstrate applications to privacy-preserving Federated Learning (FL) where clients \emph{speak once}. This is a sharp departure from prior multi-round FL protocols whose study was initiated by Bonawitz et al. (CCS, 2017). Moreover, unlike the YOSO (You Only Speak Once) model for general secure computation, $\mathsf{OPA}$ eliminates complex committee selection protocols to achieve adaptive security. Beyond asymptotic improvements, $\mathsf{OPA}$ is practical, outperforming state-of-the-art solutions. We benchmark logistic regression classifiers for two datasets, while also building an MLP classifier to train on MNIST, CIFAR-10, and CIFAR-100 datasets. We build two flavors of $\caps$ (1) from (threshold) key homomorphic PRF and (2) from seed homomorphic PRG and secret sharing.
Authors: Atharva Mehta, Shivam Chauhan, Monojit Choudhury
Abstract: Recent advances in generative AI have sparked renewed interest and expanded possibilities for music generation. However, the performance and versatility of these systems across musical genres are heavily influenced by the availability of training data. We conducted an extensive analysis of over one million hours of audio datasets used in AI music generation research and manually reviewed more than 200 papers from eleven prominent AI and music conferences and organizations (AAAI, ACM, EUSIPCO, EURASIP, ICASSP, ICML, IJCAI, ISMIR, NeurIPS, NIME, SMC) to identify a critical gap in the fair representation and inclusion of the musical genres of the Global South in AI research. Our findings reveal a stark imbalance: approximately 86% of the total dataset hours and over 93% of researchers focus primarily on music from the Global North. However, around 40% of these datasets include some form of non-Western music, genres from the Global South account for only 14.6% of the data. Furthermore, approximately 51% of the papers surveyed concentrate on symbolic music generation, a method that often fails to capture the cultural nuances inherent in music from regions such as South Asia, the Middle East, and Africa. As AI increasingly shapes the creation and dissemination of music, the significant underrepresentation of music genres in datasets and research presents a serious threat to global musical diversity. We also propose some important steps to mitigate these risks and foster a more inclusive future for AI-driven music generation.
Authors: Yuqi Wu, Wenzhao Zheng, Sicheng Zuo, Yuanhui Huang, Jie Zhou, Jiwen Lu
Abstract: 3D occupancy prediction provides a comprehensive description of the surrounding scenes and has become an essential task for 3D perception. Most existing methods focus on offline perception from one or a few views and cannot be applied to embodied agents that demand to gradually perceive the scene through progressive embodied exploration. In this paper, we formulate an embodied 3D occupancy prediction task to target this practical scenario and propose a Gaussian-based EmbodiedOcc framework to accomplish it. We initialize the global scene with uniform 3D semantic Gaussians and progressively update local regions observed by the embodied agent. For each update, we extract semantic and structural features from the observed image and efficiently incorporate them via deformable cross-attention to refine the regional Gaussians. Finally, we employ Gaussian-to-voxel splatting to obtain the global 3D occupancy from the updated 3D Gaussians. Our EmbodiedOcc assumes an unknown (i.e., uniformly distributed) environment and maintains an explicit global memory of it with 3D Gaussians. It gradually gains knowledge through the local refinement of regional Gaussians, which is consistent with how humans understand new scenes through embodied exploration. We reorganize an EmbodiedOcc-ScanNet benchmark based on local annotations to facilitate the evaluation of the embodied 3D occupancy prediction task. Our EmbodiedOcc outperforms existing methods by a large margin and accomplishes the embodied occupancy prediction with high accuracy and efficiency. Code: https://github.com/YkiWu/EmbodiedOcc.
Authors: Tianxin Xie, Yan Rong, Pengfei Zhang, Wenwu Wang, Li Liu
Abstract: Text-to-speech (TTS) has advanced from generating natural-sounding speech to enabling fine-grained control over attributes like emotion, timbre, and style. Driven by rising industrial demand and breakthroughs in deep learning, e.g., diffusion and large language models (LLMs), controllable TTS has become a rapidly growing research area. This survey provides the first comprehensive review of controllable TTS methods, from traditional control techniques to emerging approaches using natural language prompts. We categorize model architectures, control strategies, and feature representations, while also summarizing challenges, datasets, and evaluations in controllable TTS. This survey aims to guide researchers and practitioners by offering a clear taxonomy and highlighting future directions in this fast-evolving field. One can visit https://github.com/imxtx/awesome-controllabe-speech-synthesis for a comprehensive paper list and updates.
URLs: https://github.com/imxtx/awesome-controllabe-speech-synthesis
Authors: Yuta Shikuri
Abstract: Gaussian process regression is a powerful Bayesian nonlinear regression method. Recent research has enabled the capture of many types of observations using non-Gaussian likelihoods. To deal with various tasks in spatial modeling, we benefit from this development. Difficulties still arise when we can only access summarized data consisting of representative features, summary statistics, and data point counts. Such situations frequently occur primarily due to concerns about confidentiality and management costs associated with spatial data. This study tackles learning and inference using only summarized data within the framework of Gaussian process regression. To address this challenge, we analyze the approximation errors in the marginal likelihood and posterior distribution that arise from utilizing representative features. We also introduce the concept of sample quasi-likelihood, which facilitates learning and inference using only summarized data. Non-Gaussian likelihoods satisfying certain assumptions can be captured by specifying a variance function that characterizes a sample quasi-likelihood function. Theoretical and experimental results demonstrate that the approximation performance is influenced by the granularity of summarized data relative to the length scale of covariance functions. Experiments on a real-world dataset highlight the practicality of our method for spatial modeling.
Authors: Tenghui Li, Guoxu Zhou, Xuyang Zhao, Qibin Zhao
Abstract: Large language models have demonstrated predictable scaling behaviors with respect to model parameters and training data. This study investigates whether a similar scaling relationship exist for vision-language models with respect to the number of vision tokens. A mathematical framework is developed to characterize a relationship between vision token number and the expected divergence of distance between vision-referencing sequences. The theoretical analysis reveals two distinct scaling regimes: sublinear scaling for less vision tokens and linear scaling for more vision tokens. This aligns with model performance relationships of the form \(S(n) \approx c / n^{\alpha(n)}\), where the scaling exponent relates to the correlation structure between vision token representations. Empirical validations across multiple vision-language benchmarks show that model performance matches the prediction from scaling relationship. The findings contribute to understanding vision token scaling in transformers through a theoretical framework that complements empirical observations.
Authors: Zhenyu Lei, Yushun Dong, Weiyu Li, Rong Ding, Qi Wang, Jundong Li
Abstract: Large language models (LLMs) have revolutionized scientific research with their exceptional capabilities and transformed various fields. Among their practical applications, LLMs have been playing a crucial role in mitigating threats to human life, infrastructure, and the environment. Despite growing research in disaster LLMs, there remains a lack of systematic review and in-depth analysis of LLMs for natural disaster management. To address the gap, this paper presents a comprehensive survey of existing LLMs in natural disaster management, along with a taxonomy that categorizes existing works based on disaster phases and application scenarios. By collecting public datasets and identifying key challenges and opportunities, this study aims to guide the professional community in developing advanced LLMs for disaster management to enhance the resilience against natural disasters.
Authors: Akshay Krishnan, Xinchen Yan, Vincent Casser, Abhijit Kundu
Abstract: We introduce Orchid, a unified latent diffusion model that learns a joint appearance-geometry prior to generate color, depth, and surface normal images in a single diffusion process. This unified approach is more efficient and coherent than current pipelines that use separate models for appearance and geometry. Orchid is versatile - it directly generates color, depth, and normal images from text, supports joint monocular depth and normal estimation with color-conditioned finetuning, and seamlessly inpaints large 3D regions by sampling from the joint distribution. It leverages a novel Variational Autoencoder (VAE) that jointly encodes RGB, relative depth, and surface normals into a shared latent space, combined with a latent diffusion model that denoises these latents. Our extensive experiments demonstrate that Orchid delivers competitive performance against SOTA task-specific methods for geometry prediction, even surpassing them in normal-prediction accuracy and depth-normal consistency. It also inpaints color-depth-normal images jointly, with more qualitative realism than existing multi-step methods.
Authors: Huayu Chen, Kai Jiang, Kaiwen Zheng, Jianfei Chen, Hang Su, Jun Zhu
Abstract: Classifier-Free Guidance (CFG) has been a default technique in various visual generative models, yet it requires inference from both conditional and unconditional models during sampling. We propose to build visual models that are free from guided sampling. The resulting algorithm, Guidance-Free Training (GFT), matches the performance of CFG while reducing sampling to a single model, halving the computational cost. Unlike previous distillation-based approaches that rely on pretrained CFG networks, GFT enables training directly from scratch. GFT is simple to implement. It retains the same maximum likelihood objective as CFG and differs mainly in the parameterization of conditional models. Implementing GFT requires only minimal modifications to existing codebases, as most design choices and hyperparameters are directly inherited from CFG. Our extensive experiments across five distinct visual models demonstrate the effectiveness and versatility of GFT. Across domains of diffusion, autoregressive, and masked-prediction modeling, GFT consistently achieves comparable or even lower FID scores, with similar diversity-fidelity trade-offs compared with CFG baselines, all while being guidance-free. Code will be available at https://github.com/thu-ml/GFT.
Authors: Yannis Montreuil, Axel Carlier, Lai Xing Ng, Wei Tsang Ooi
Abstract: Two-stage Learning-to-Defer (L2D) enables optimal task delegation by assigning each input to either a fixed main model or one of several offline experts, supporting reliable decision-making in complex, multi-agent environments. However, existing L2D frameworks assume clean inputs and are vulnerable to adversarial perturbations that can manipulate query allocation--causing costly misrouting or expert overload. We present the first comprehensive study of adversarial robustness in two-stage L2D systems. We introduce two novel attack strategie--untargeted and targeted--which respectively disrupt optimal allocations or force queries to specific agents. To defend against such threats, we propose SARD, a convex learning algorithm built on a family of surrogate losses that are provably Bayes-consistent and $(\mathcal{R}, \mathcal{G})$-consistent. These guarantees hold across classification, regression, and multi-task settings. Empirical results demonstrate that SARD significantly improves robustness under adversarial attacks while maintaining strong clean performance, marking a critical step toward secure and trustworthy L2D deployment.
Authors: Nina Corvelo Benz, Stratis Tsirtsis, Eleni Straitouri, Ivi Chatzi, Ander Artola Velasco, Suhas Thejaswi, Manuel Gomez-Rodriguez
Abstract: State of the art large language models rely on randomization to respond to a prompt. As an immediate consequence, a model may respond differently to the same prompt if asked multiple times. In this work, we argue that the evaluation and ranking of large language models should control for the randomization underpinning their functioning. Our starting point is the development of a causal model for coupled autoregressive generation, which allows different large language models to sample responses with the same source of randomness. Building upon our causal model, we first show that, on evaluations based on benchmark datasets, coupled autoregressive generation leads to the same conclusions as vanilla autoregressive generation but using provably fewer samples. However, we further show that, on evaluations based on (human) pairwise comparisons, coupled and vanilla autoregressive generation can surprisingly lead to different rankings when comparing more than two models, even with an infinite amount of samples. This suggests that the apparent advantage of a model over others in existing evaluation protocols may not be genuine but rather confounded by the randomness inherent to the generation process. To illustrate and complement our theoretical results, we conduct experiments with several large language models from the Llama, Mistral and Qwen families. We find that, across multiple benchmark datasets, coupled autoregressive generation requires up to 75% fewer samples to reach the same conclusions as vanilla autoregressive generation. Further, we find that the win-rates derived from pairwise comparisons by a strong large language model to prompts from the LMSYS Chatbot Arena platform differ under coupled and vanilla autoregressive generation.
Authors: Lancelot F. James, Juho Lee, Abhinav Pandey
Abstract: We introduce the Poisson Hierarchical Indian Buffet Process (PHIBP), a new class of species sampling models designed to address the challenges of complex, sparse count data by facilitating information sharing across and within groups. Our theoretical developments enable a tractable Bayesian nonparametric framework with machine learning elements, accommodating a potentially infinite number of species (taxa) whose parameters are learned from data. Focusing on microbiome analysis, we address key gaps by providing a flexible multivariate count model that accounts for overdispersion and robustly handles diverse data types (OTUs, ASVs). We introduce novel parameters reflecting species abundance and diversity. The model borrows strength across groups while explicitly distinguishing between technical and biological zeros to interpret sparse co-occurrence patterns. This results in a framework with tractable posterior inference, exact generative sampling, and a principled solution to the unseen species problem. We describe extensions where domain experts can incorporate knowledge through covariates and structured priors, with potential for strain-level analysis. While motivated by ecology, our work provides a broadly applicable methodology for hierarchical count modeling in genetics, commerce, and text analysis, and has significant implications for the broader theory of species sampling models arising in probability and statistics.
Authors: Wenbo Sun, Qiming Guo, Wenlu Wang, Rihan Hai
Abstract: Deploying Large Language Models (LLMs) on resource-constrained devices remains challenging due to limited memory, lack of GPUs, and the complexity of existing runtimes. In this paper, we introduce TranSQL+, a template-based code generator that translates LLM computation graphs into pure SQL queries for execution in relational databases. Without relying on external libraries, TranSQL+, leverages mature database features, such as vectorized execution and out-of-core processing, for efficient inference. We further propose a row-to-column (ROW2COL) optimization that improves join efficiency in matrix operations. Evaluated on Llama3-8B and DeepSeekMoE models, TranSQL+ achieves up to 20x lower prefill latency and 4x higher decoding speed compared to DeepSpeed Inference and Llama.cpp in low-memory and CPU-only configurations. Our results highlight relational databases as a practical environment for LLMs on low-resource hardware.
Authors: Yuxuan Han, Han Zhong, Miao Lu, Jose Blanchet, Zhengyuan Zhou
Abstract: We study the fundamental problem of offline assortment optimization under the Multinomial Logit (MNL) model, where sellers must determine the optimal subset of the products to offer based solely on historical customer choice data. While most existing approaches to learning-based assortment optimization focus on the online learning of the optimal assortment through repeated interactions with customers, such exploration can be costly or even impractical in many real-world settings. In this paper, we consider the offline learning paradigm and investigate the minimal data requirements for efficient offline assortment optimization. To this end, we introduce Pessimistic Rank-Breaking (PRB), an algorithm that combines rank-breaking with pessimistic estimation. We prove that PRB is nearly minimax optimal by establishing the tight suboptimality upper bound and a nearly matching lower bound. This further shows that "optimal item coverage" - where each item in the optimal assortment appears sufficiently often in the historical data - is both sufficient and necessary for efficient offline learning. This significantly relaxes the previous requirement of observing the complete optimal assortment in the data. Our results provide fundamental insights into the data requirements for offline assortment optimization under the MNL model.
Authors: Aakash Patel, Tianqing Zhang, Camille Avestruz, Jeffrey Regier, the LSST Dark Energy Science Collaboration
Abstract: Neural posterior estimation (NPE), a type of amortized variational inference, is a computationally efficient means of constructing probabilistic catalogs of light sources from astronomical images. To date, NPE has not been used to perform inference in models with spatially varying covariates. However, ground-based astronomical images have spatially varying sky backgrounds and point spread functions (PSFs), and accounting for this variation is essential for constructing accurate catalogs of imaged light sources. In this work, we introduce a method of performing NPE with spatially varying backgrounds and PSFs. In this method, we generate synthetic catalogs and semi-synthetic images for these catalogs using randomly sampled PSF and background estimates from existing surveys. Using this data, we train a neural network, which takes an astronomical image and representations of its background and PSF as input, to output a probabilistic catalog. Our experiments with Sloan Digital Sky Survey data demonstrate the effectiveness of NPE in the presence of spatially varying backgrounds and PSFs for light source detection, star/galaxy separation, and flux measurement.
Authors: Hanjiang Hu, Alexander Robey, Changliu Liu
Abstract: Large language models (LLMs) are shown to be vulnerable to jailbreaking attacks where adversarial prompts are designed to elicit harmful responses. While existing defenses effectively mitigate single-turn attacks by detecting and filtering unsafe inputs, they fail against multi-turn jailbreaks that exploit contextual drift over multiple interactions, gradually leading LLMs away from safe behavior. To address this challenge, we propose a safety steering framework grounded in safe control theory, ensuring invariant safety in multi-turn dialogues. Our approach models the dialogue with LLMs using state-space representations and introduces a novel neural barrier function (NBF) to detect and filter harmful queries emerging from evolving contexts proactively. Our method achieves invariant safety at each turn of dialogue by learning a safety predictor that accounts for adversarial queries, preventing potential context drift toward jailbreaks. Extensive experiments under multiple LLMs show that our NBF-based safety steering outperforms safety alignment, prompt-based steering and lightweight LLM guardrails baselines, offering stronger defenses against multi-turn jailbreaks while maintaining a better trade-off among safety, helpfulness and over-refusal. Check out the website here https://sites.google.com/view/llm-nbf/home . Our code is available on https://github.com/HanjiangHu/NBF-LLM .
URLs: https://sites.google.com/view/llm-nbf/home, https://github.com/HanjiangHu/NBF-LLM
Authors: Yunfan Jiang, Ruohan Zhang, Josiah Wong, Chen Wang, Yanjie Ze, Hang Yin, Cem Gokmen, Shuran Song, Jiajun Wu, Li Fei-Fei
Abstract: Real-world household tasks present significant challenges for mobile manipulation robots. An analysis of existing robotics benchmarks reveals that successful task performance hinges on three key whole-body control capabilities: bimanual coordination, stable and precise navigation, and extensive end-effector reachability. Achieving these capabilities requires careful hardware design, but the resulting system complexity further complicates visuomotor policy learning. To address these challenges, we introduce the BEHAVIOR Robot Suite (BRS), a comprehensive framework for whole-body manipulation in diverse household tasks. Built on a bimanual, wheeled robot with a 4-DoF torso, BRS integrates a cost-effective whole-body teleoperation interface for data collection and a novel algorithm for learning whole-body visuomotor policies. We evaluate BRS on five challenging household tasks that not only emphasize the three core capabilities but also introduce additional complexities, such as long-range navigation, interaction with articulated and deformable objects, and manipulation in confined spaces. We believe that BRS's integrated robotic embodiment, data collection interface, and learning framework mark a significant step toward enabling real-world whole-body manipulation for everyday household tasks. BRS is open-sourced at https://behavior-robot-suite.github.io/
Authors: Massa Baali, Xiang Li, Hao Chen, Syed Abdul Hannan, Rita Singh, Bhiksha Raj
Abstract: Speaker verification is a typical zero-shot learning task, where inference of unseen classes is performed by comparing embeddings of test instances to known examples. The models performing inference must hence naturally generate embeddings that cluster same-class instances compactly, while maintaining separation across classes. In order to learn to do so, they are typically trained on a large number of classes (speakers), often using specialized losses. However real-world speaker datasets often lack the class diversity needed to effectively learn this in a generalizable manner. We introduce CAARMA, a class augmentation framework that addresses this problem by generating synthetic classes through data mixing in the embedding space, expanding the number of training classes. To ensure the authenticity of the synthetic classes we adopt a novel adversarial refinement mechanism that minimizes categorical distinctions between synthetic and real classes. We evaluate CAARMA on multiple speaker verification tasks, as well as other representative zero-shot comparison-based speech analysis tasks and obtain consistent improvements: our framework demonstrates a significant improvement of 8\% over all baseline models. The code is available at: https://github.com/massabaali7/CAARMA/
Authors: Taufiq Ahmed, Abhishek Kumar, Constantino \'Alvarez Casado, Anlan Zhang, Tuomo H\"anninen, Lauri Loven, Miguel Bordallo L\'opez, Sasu Tarkoma
Abstract: Object detection models often struggle with class imbalance, where rare categories appear significantly less frequently than common ones. Existing sampling-based rebalancing strategies, such as Repeat Factor Sampling (RFS) and Instance-Aware Repeat Factor Sampling (IRFS), mitigate this issue by adjusting sample frequencies based on image and instance counts. However, these methods are based on linear adjustments, which limit their effectiveness in long-tailed distributions. This work introduces Exponentially Weighted Instance-Aware Repeat Factor Sampling (E-IRFS), an extension of IRFS that applies exponential scaling to better differentiate between rare and frequent classes. E-IRFS adjusts sampling probabilities using an exponential function applied to the geometric mean of image and instance frequencies, ensuring a more adaptive rebalancing strategy. We evaluate E-IRFS on a dataset derived from the Fireman-UAV-RGBT Dataset and four additional public datasets, using YOLOv11 object detection models to identify fire, smoke, people and lakes in emergency scenarios. The results show that E-IRFS improves detection performance by 22\% over the baseline and outperforms RFS and IRFS, particularly for rare categories. The analysis also highlights that E-IRFS has a stronger effect on lightweight models with limited capacity, as these models rely more on data sampling strategies to address class imbalance. The findings demonstrate that E-IRFS improves rare object detection in resource-constrained environments, making it a suitable solution for real-time applications such as UAV-based emergency monitoring. The code is available at: https://github.com/futurians/E-IRFS.
Authors: Huan Zhao, Yiming Liu, Jina Yao, Ling Xiong, Zexin Zhou, Zixing Zhang
Abstract: Recent breakthroughs in single-cell technology have ushered in unparalleled opportunities to decode the molecular intricacy of intricate biological systems, especially those linked to diseases unique to humans. However, these progressions have also ushered in novel obstacles-specifically, the efficient annotation of extensive, long-tailed single-cell data pertaining to disease conditions. To effectively surmount this challenge, we introduce Celler, a state-of-the-art generative pre-training model crafted specifically for the annotation of single-cell data. Celler incorporates two groundbreaking elements: First, we introduced the Gaussian Inflation (GInf) Loss function. By dynamically adjusting sample weights, GInf Loss significantly enhances the model's ability to learn from rare categories while reducing the risk of overfitting for common categories. Secondly, we introduce an innovative Hard Data Mining (HDM) strategy into the training process, specifically targeting the challenging-to-learn minority data samples, which significantly improved the model's predictive accuracy. Additionally, to further advance research in this field, we have constructed a large-scale single-cell dataset: Celler-75, which encompasses 40 million cells distributed across 80 human tissues and 75 specific diseases. This dataset provides critical support for comprehensively exploring the potential of single-cell technology in disease research. Our code is available at https://github.com/AI4science-ym/HiCeller.
Authors: Anas Barakat, Wayne Lin, John Lazarsfeld, Antonios Varvitsiotis
Abstract: We introduce symmetric cone games (SCGs), a broad class of multi-player games where each player's strategy lies in a generalized simplex (the trace-one slice of a symmetric cone). This framework unifies a wide spectrum of settings, including normal-form games (simplex strategies), quantum games (density matrices), and continuous games with ball-constrained strategies. It also captures several structured machine learning and optimization problems, such as distance metric learning and Fermat-Weber facility location, as two-player zero-sum SCGs. To compute approximate Nash equilibria in two-player zero-sum SCGs, we propose a single online learning algorithm: Optimistic Symmetric Cone Multiplicative Weights Updates (OSCMWU). Unlike prior methods tailored to specific geometries, OSCMWU provides closed-form, projection-free updates over any symmetric cone and achieves an optimal $\tilde{\mathcal{O}}(1/\epsilon)$ iteration complexity for computing $\epsilon$-saddle points. Our analysis builds on the Optimistic Follow-the-Regularized-Leader framework and hinges on a key technical contribution: We prove that the symmetric cone negative entropy is strongly convex with respect to the trace-one norm. This result extends known results for the simplex and spectraplex to all symmetric cones, and may be of independent interest.
Authors: Xiuyuan Cheng, Zheng Dong, Yao Xie
Abstract: Spatio-temporal point processes (STPPs) model discrete events distributed in time and space, with important applications in areas such as criminology, seismology, epidemiology, and social networks. Traditional models often rely on parametric kernels, limiting their ability to capture heterogeneous, nonstationary dynamics. Recent innovations integrate deep neural architectures -- either by modeling the conditional intensity function directly or by learning flexible, data-driven influence kernels, substantially broadening their expressive power. This article reviews the development of the deep influence kernel approach, which enjoys statistical explainability, since the influence kernel remains in the model to capture the spatiotemporal propagation of event influence and its impact on future events, while also possessing strong expressive power, thereby benefiting from both worlds. We explain the main components in developing deep kernel point processes, leveraging tools such as functional basis decomposition and graph neural networks to encode complex spatial or network structures, as well as estimation using both likelihood-based and likelihood-free methods, and address computational scalability for large-scale data. We also discuss the theoretical foundation of kernel identifiability. Simulated and real-data examples highlight applications to crime analysis, earthquake aftershock prediction, and sepsis prediction modeling, and we conclude by discussing promising directions for the field.
Authors: Salman Rahman, Liwei Jiang, James Shiffer, Genglin Liu, Sheriff Issaka, Md Rizwan Parvez, Hamid Palangi, Kai-Wei Chang, Yejin Choi, Saadia Gabriel
Abstract: Multi-turn interactions with language models (LMs) pose critical safety risks, as harmful intent can be strategically spread across exchanges. Yet, the vast majority of prior work has focused on single-turn safety, while adaptability and diversity remain among the key challenges of multi-turn red-teaming. To address these challenges, we present X-Teaming, a scalable framework that systematically explores how seemingly harmless interactions escalate into harmful outcomes and generates corresponding attack scenarios. X-Teaming employs collaborative agents for planning, attack optimization, and verification, achieving state-of-the-art multi-turn jailbreak effectiveness and diversity with success rates up to 98.1% across representative leading open-weight and closed-source models. In particular, X-Teaming achieves a 96.2% attack success rate against the latest Claude 3.7 Sonnet model, which has been considered nearly immune to single-turn attacks. Building on X-Teaming, we introduce XGuard-Train, an open-source multi-turn safety training dataset that is 20x larger than the previous best resource, comprising 30K interactive jailbreaks, designed to enable robust multi-turn safety alignment for LMs. Our work offers essential tools and insights for mitigating sophisticated conversational attacks, advancing the multi-turn safety of LMs.
Authors: Anjiang Wei, Huanmi Tan, Tarun Suresh, Daniel Mendoza, Thiago S. F. X. Teixeira, Ke Wang, Caroline Trippel, Alex Aiken
Abstract: Recent advances in Large Language Models (LLMs) have sparked growing interest in applying them to Electronic Design Automation (EDA) tasks, particularly Register Transfer Level (RTL) code generation. While several RTL datasets have been introduced, most focus on syntactic validity rather than functional validation with tests, leading to training examples that compile but may not implement the intended behavior. We present VERICODER, a model for RTL code generation fine-tuned on a dataset validated for functional correctness. This fine-tuning dataset is constructed using a novel methodology that combines unit test generation with feedback-directed refinement. Given a natural language specification and an initial RTL design, we prompt a teacher model (GPT-4o-mini) to generate unit tests and iteratively revise the RTL design based on its simulation results using the generated tests. If necessary, the teacher model also updates the tests to ensure they comply with the natural language specification. As a result of this process, every example in our dataset is functionally validated, consisting of a natural language description, an RTL implementation, and passing tests. Fine-tuned on this dataset of 125,777 examples, VERICODER achieves state-of-the-art metrics in functional correctness on VerilogEval and RTLLM, with relative gains of up to 71.7% and 27.4%, respectively. An ablation study further shows that models trained on our functionally validated dataset outperform those trained on functionally non-validated datasets, underscoring the importance of high-quality datasets in RTL code generation. Our code, data, and models are publicly available at https://github.com/Anjiang-Wei/VeriCoder
Authors: Raza Ul Mustafa, Sesha Dassanayake, Noman Ashraf
Abstract: The Quality of Experience (QoE) is the users satisfaction while streaming a video session over an over-the-top (OTT) platform like YouTube. QoE of YouTube reflects the smooth streaming session without any buffering and quality shift events. One of the most important factors nowadays affecting QoE of YouTube is frequent shifts from higher to lower resolutions and vice versa. These shifts ensure a smooth streaming session; however, it might get a lower mean opinion score. For instance, dropping from 1080p to 480p during a video can preserve continuity but might reduce the viewers enjoyment. Over time, OTT platforms are looking for alternative ways to boost user experience instead of relying on traditional Quality of Service (QoS) metrics such as bandwidth, latency, and throughput. As a result, we look into the relationship between quality shifting in YouTube streaming sessions and the channel metrics RSRP, RSRQ, and SNR. Our findings state that these channel metrics positively correlate with shifts. Thus, in real-time, OTT can only rely on them to predict video streaming sessions into lower- and higher-resolution categories, thus providing more resources to improve user experience. Using traditional Machine Learning (ML) classifiers, we achieved an accuracy of 77-percent, while using only RSRP, RSRQ, and SNR. In the era of 5G and beyond, where ultra-reliable, low-latency networks promise enhanced streaming capabilities, the proposed methodology can be used to improve OTT services.
Authors: Mete Erdogan, Sebnem Demirtas
Abstract: Accurate and early diagnosis of pneumonia through X-ray imaging is essential for effective treatment and improved patient outcomes. Recent advancements in machine learning have enabled automated diagnostic tools that assist radiologists in making more reliable and efficient decisions. In this work, we propose a Singular Value Decomposition-based Least Squares (SVD-LS) framework for multi-class pneumonia classification, leveraging powerful feature representations from state-of-the-art self-supervised and transfer learning models. Rather than relying on computationally expensive gradient-based fine-tuning, we employ a closed-form, non-iterative classification approach that ensures efficiency without compromising accuracy. Experimental results demonstrate that SVD-LS achieves competitive performance while offering significantly reduced computational costs, making it a viable alternative for real-time medical imaging applications. The implementation is available at: github.com/meterdogan07/SVD-LS.
Authors: Ibne Farabi Shihab, Sanjeda Akter, Anuj Sharma
Abstract: Quantum machine learning for spin and molecular systems faces critical challenges of scarce labeled data and computationally expensive simulations. To address these limitations, we introduce Hamiltonian-Masked Autoencoding (HMAE), a novel self-supervised framework that pre-trains transformers on unlabeled quantum Hamiltonians, enabling efficient few-shot transfer learning. Unlike random masking approaches, HMAE employs a physics-informed strategy based on quantum information theory to selectively mask Hamiltonian terms based on their physical significance. Experiments on 12,500 quantum Hamiltonians (60% real-world, 40% synthetic) demonstrate that HMAE achieves 85.3% $\pm$ 1.5% accuracy in phase classification and 0.15 $\pm$ 0.02 eV MAE in ground state energy prediction with merely 10 labeled examples - a statistically significant improvement (p < 0.01) over classical graph neural networks (78.1% $\pm$ 2.1%) and quantum neural networks (76.8% $\pm$ 2.3%). Our method's primary advantage is exceptional sample efficiency - reducing required labeled examples by 3-5x compared to baseline methods - though we emphasize that ground truth values for fine-tuning and evaluation still require exact diagonalization or tensor networks. We explicitly acknowledge that our current approach is limited to small quantum systems (specifically limited to 12 qubits during training, with limited extension to 16-20 qubits in testing) and that, while promising within this regime, this size restriction prevents immediate application to larger systems of practical interest in materials science and quantum chemistry.
Authors: Owen Kwon, Abraham George, Alison Bartsch, Amir Barati Farimani
Abstract: Real robots are expected to repeat the same behavior in new environments with very little new data, yet modern controllers either incur heavy per-step inference or require deployment-time fine-tuning. We propose RT-Cache, a training-free retrieval-as-control pipeline that caches diverse image action trajectories in a unified vector memory and, at test time, embeds the current frame to retrieve and replay multi-step snippets, replacing per-step model calls. A hierarchical search keeps lookups sub-second at million scale, shifting cost from compute to storage and enabling real-time control on modest GPUs. Across real-robot tasks and large open logs, RT-Cache achieves higher success and lower completion time than strong retrieval baselines (approximately x2 higher success and ~30% faster in our settings), and a single-episode anchoring study shows immediate adaptation to a more complex, contact-rich task without fine-tuning. RT-Cache turns experience into an append-only memory, offering a simple, scalable path to few-shot deployment today and a foundation for multimodal keys and optional integration with high-level policies. Project page: https://rt-cache.github.io/.
Authors: Hao Wang, Pinzhi Huang, Jihan Yang, Saining Xie, Daisuke Kawahara
Abstract: The rapid evolution of multimodal large language models (MLLMs) has significantly enhanced their real-world applications. However, achieving consistent performance across languages, especially when integrating cultural knowledge, remains a significant challenge. To better assess this issue, we introduce two new benchmarks: KnowRecall and VisRecall, which evaluate cross-lingual consistency in MLLMs. KnowRecall is a visual question answering benchmark designed to measure factual knowledge consistency in 15 languages, focusing on cultural and historical questions about global landmarks. VisRecall assesses visual memory consistency by asking models to describe landmark appearances in 9 languages without access to images. Experimental results reveal that state-of-the-art MLLMs, including proprietary ones, still struggle to achieve cross-lingual consistency. This underscores the need for more robust approaches that produce truly multilingual and culturally aware models.
Authors: Fangqiao Tian, An Luo, Jin Du, Xun Xian, Robert Specht, Ganghua Wang, Xuan Bi, Jiawei Zhou, Ashish Kundu, Jayanth Srinivasa, Charles Fleming, Rui Zhang, Zirui Liu, Mingyi Hong, Jie Ding
Abstract: A multi-agent AI system (MAS) is composed of multiple autonomous agents that interact, exchange information, and make decisions based on internal generative models. Recent advances in large language models and tool-using agents have made MAS increasingly practical in areas like scientific discovery and collaborative automation. However, key questions remain: When are MAS more effective than single-agent systems? What new safety risks arise from agent interactions? And how should we evaluate their reliability and structure? This paper outlines a formal framework for analyzing MAS, focusing on two core aspects: effectiveness and safety. We explore whether MAS truly improve robustness, adaptability, and performance, or merely repackage known techniques like ensemble learning. We also study how inter-agent dynamics may amplify or suppress system vulnerabilities. While MAS are relatively new to the signal processing community, we envision them as a powerful abstraction that extends classical tools like distributed estimation and sensor fusion to higher-level, policy-driven inference. Through experiments on data science automation, we highlight the potential of MAS to reshape how signal processing systems are designed and trusted.
Authors: Aleksandr Tsymbalov, Mikhail Khovrichev
Abstract: Machine learning models for text classification are trained to predict a class for a given text. To do this, training and validation samples must be prepared: a set of texts is collected, and each text is assigned a class. These classes are usually assigned by human annotators with different expertise levels, depending on the specific classification task. Collecting such samples from scratch is labor-intensive because it requires finding specialists and compensating them for their work; moreover, the number of available specialists is limited, and their productivity is constrained by human factors. While it may not be too resource-intensive to collect samples once, the ongoing need to retrain models (especially in incremental learning pipelines) to address data drift (also called model drift) makes the data collection process crucial and costly over the model's entire lifecycle. This paper proposes several approaches to replace human annotators with Large Language Models (LLMs) to test classifier predictions for correctness, helping ensure model quality and support high-quality incremental learning.
Authors: Tin Trung Nguyen, Jiannan Xu, Zora Che, Phuong-Anh Nguyen-Le, Rushil Dandamudi, Donald Braman, Furong Huang, Hal Daum\'e III, Zubin Jelveh
Abstract: Although popularized AI fairness metrics, e.g., demographic parity, have uncovered bias in AI-assisted decision-making outcomes, they do not consider how much effort one has spent to get to where one is today in the input feature space. However, the notion of effort is important in how Philosophy and humans understand fairness. We propose a philosophy-informed approach to conceptualize and evaluate Effort-aware Fairness (EaF), grounded in the concept of Force, which represents the temporal trajectory of predictive features coupled with inertia. Besides theoretical formulation, our empirical contributions include: (1) a pre-registered human subjects experiment, which shows that for both stages of the (individual) fairness evaluation process, people consider the temporal trajectory of a predictive feature more than its aggregate value; (2) pipelines to compute Effort-aware Individual/Group Fairness in the criminal justice and personal finance contexts. Our work may enable AI model auditors to uncover and potentially correct unfair decisions against individuals who have spent significant efforts to improve but are still stuck with systemic disadvantages outside their control.
Authors: Agnideep Aich
Abstract: Classical estimators, the cornerstones of statistical inference, face insurmountable challenges when applied to important emerging classes of Archimedean copulas. These models exhibit pathological properties, including numerically unstable densities, non-monotonic parameter-to-dependence mappings, and vanishingly small likelihood gradients, rendering methods like Maximum Likelihood (MLE) and Method of Moments (MoM) inconsistent or computationally infeasible. We introduce IGNIS, a unified neural estimation framework that sidesteps these barriers by learning a direct, robust mapping from data-driven dependency measures to the underlying copula parameter theta. IGNIS utilizes a multi-input architecture and a theory-guided output layer (softplus(z) + 1) to automatically enforce the domain constraint theta_hat >= 1. Trained and validated on four families (Gumbel, Joe, and the numerically challenging A1/A2), IGNIS delivers accurate and stable estimates for real-world financial and health datasets, demonstrating its necessity for reliable inference in modern, complex dependence models where traditional methods fail.
Authors: Andy Bonnetto, Haozhe Qi, Franklin Leong, Matea Tashkovska, Mahdi Rad, Solaiman Shokur, Friedhelm Hummel, Silvestro Micera, Marc Pollefeys, Alexander Mathis
Abstract: Understanding behavior requires datasets that capture humans while carrying out complex tasks. The kitchen is an excellent environment for assessing human motor and cognitive function, as many complex actions are naturally exhibited in kitchens from chopping to cleaning. Here, we introduce the EPFL-Smart-Kitchen-30 dataset, collected in a noninvasive motion capture platform inside a kitchen environment. Nine static RGB-D cameras, inertial measurement units (IMUs) and one head-mounted HoloLens~2 headset were used to capture 3D hand, body, and eye movements. The EPFL-Smart-Kitchen-30 dataset is a multi-view action dataset with synchronized exocentric, egocentric, depth, IMUs, eye gaze, body and hand kinematics spanning 29.7 hours of 16 subjects cooking four different recipes. Action sequences were densely annotated with 33.78 action segments per minute. Leveraging this multi-modal dataset, we propose four benchmarks to advance behavior understanding and modeling through 1) a vision-language benchmark, 2) a semantic text-to-motion generation benchmark, 3) a multi-modal action recognition benchmark, 4) a pose-based action segmentation benchmark. We expect the EPFL-Smart-Kitchen-30 dataset to pave the way for better methods as well as insights to understand the nature of ecologically-valid human behavior. Code and data are available at https://github.com/amathislab/EPFL-Smart-Kitchen
Authors: Junjie Xing, Yeye He, Mengyu Zhou, Haoyu Dong, Shi Han, Lingjiao Chen, Dongmei Zhang, Surajit Chaudhuri, H. V. Jagadish
Abstract: Tables and table-based use cases play a crucial role in many important real-world applications, such as spreadsheets, databases, and computational notebooks, which traditionally require expert-level users like data engineers, data analysts, and database administrators to operate. Although LLMs have shown remarkable progress in working with tables (e.g., in spreadsheet and database copilot scenarios), comprehensive benchmarking of such capabilities remains limited. In contrast to an extensive and growing list of NLP benchmarks, evaluations of table-related tasks are scarce, and narrowly focus on tasks like NL-to-SQL and Table-QA, overlooking the broader spectrum of real-world tasks that professional users face. This gap limits our understanding and model progress in this important area. In this work, we introduce MMTU, a large-scale benchmark with over 30K questions across 25 real-world table tasks, designed to comprehensively evaluate models ability to understand, reason, and manipulate real tables at the expert-level. These tasks are drawn from decades' worth of computer science research on tabular data, with a focus on complex table tasks faced by professional users. We show that MMTU require a combination of skills -- including table understanding, reasoning, and coding -- that remain challenging for today's frontier models, where even frontier reasoning models like OpenAI o4-mini and DeepSeek R1 score only around 60%, suggesting significant room for improvement. We highlight key findings in our evaluation using MMTU and hope that this benchmark drives further advances in understanding and developing foundation models for structured data processing and analysis. Our code and data are available at https://github.com/MMTU-Benchmark/MMTU and https://huggingface.co/datasets/MMTU-benchmark/MMTU.
URLs: https://github.com/MMTU-Benchmark/MMTU, https://huggingface.co/datasets/MMTU-benchmark/MMTU.
Authors: Benjamin Smith, Jianhui Gao, Jessica Gronsbell
Abstract: Fairness is a growing area of machine learning (ML) that focuses on ensuring models do not produce systematically biased outcomes for specific groups, particularly those defined by protected attributes such as race, gender, or age. Evaluating fairness is a critical aspect of ML model development, as biased models can perpetuate structural inequalities. The {fairmetrics} R package offers a user-friendly framework for rigorously evaluating numerous group-based fairness criteria, including metrics based on independence (e.g., statistical parity), separation (e.g., equalized odds), and sufficiency (e.g., predictive parity). Group-based fairness criteria assess whether a model is equally accurate or well-calibrated across a set of predefined groups so that appropriate bias mitigation strategies can be implemented. {fairmetrics} provides both point and interval estimates for multiple metrics through a convenient wrapper function and includes an example dataset derived from the Medical Information Mart for Intensive Care, version II (MIMIC-II) database (Goldberger et al., 2000; Raffa, 2016).
Authors: Hongyu Yao, Zijin Hong, Hao Chen, Yuanchen Bei, Zhiqing Li, Qijie Shen, Zuobin Ying, Huan Gong, Feiran Huang
Abstract: Graph-based multi-task learning at billion-scale presents a significant challenge, as different tasks correspond to distinct billion-scale graphs. Traditional multi-task learning methods often neglect these graph structures, relying solely on individual user and item embeddings. However, disregarding graph structures overlooks substantial potential for improving performance. In this paper, we introduce the Macro Graph of Expert (MGOE) framework, the first approach capable of leveraging macro graph embeddings to capture task-specific macro features while modeling the correlations between task-specific experts. Specifically, we propose the concept of a Macro Graph Bottom, which, for the first time, enables multi-task learning models to incorporate graph information effectively. We design the Macro Prediction Tower to dynamically integrate macro knowledge across tasks. MGOE has been deployed at scale, powering multi-task learning for the homepage of a leading billion-scale recommender system. Extensive offline experiments conducted on three public benchmark datasets demonstrate its superiority over state-of-the-art multi-task learning methods, establishing MGOE as a breakthrough in multi-task graph-based recommendation. Furthermore, online A/B tests confirm the superiority of MGOE in billion-scale recommender systems.
Authors: Michela Lapenna, Caterina De Bacco
Abstract: Graphs are a powerful data structure for representing relational data and are widely used to describe complex real-world systems. Probabilistic Graphical Models (PGMs) and Graph Neural Networks (GNNs) can both leverage graph-structured data, but their inherent functioning is different. The question is how do they compare in capturing the information contained in networked datasets? We address this objective by solving a link prediction task and we conduct three main experiments, on both synthetic and real networks: one focuses on how PGMs and GNNs handle input features, while the other two investigate their robustness to noisy features and increasing heterophily of the graph. PGMs do not necessarily require features on nodes, while GNNs cannot exploit the network edges alone, and the choice of input features matters. We find that GNNs are outperformed by PGMs when input features are low-dimensional or noisy, mimicking many real scenarios where node attributes might be scalar or noisy. Then, we find that PGMs are more robust than GNNs when the heterophily of the graph is increased. Finally, to assess performance beyond prediction tasks, we also compare the two frameworks in terms of their computational complexity and interpretability.
Authors: Haotian Yin, Aleksander Plocharski, Michal Jan Wlodarczyk, Mikolaj Kida, Przemyslaw Musialski
Abstract: Neural signed-distance fields (SDFs) are a versatile backbone for neural geometry representation, but enforcing CAD-style developability usually requires Gaussian-curvature penalties with full Hessian evaluation and second-order differentiation, which are costly in memory and time. We introduce an off-diagonal Weingarten loss that regularizes only the mixed shape operator term that represents the gap between principal curvatures and flattens the surface. We present two variants: a finite-difference version using six SDF evaluations plus one gradient, and an auto-diff version using a single Hessian-vector product. Both converge to the exact mixed term and preserve the intended geometric properties without assembling the full Hessian. On the ABC benchmarks the losses match or exceed Hessian-based baselines while cutting GPU memory and training time by roughly a factor of two. The method is drop-in and framework-agnostic, enabling scalable curvature-aware SDF learning for engineering-grade shape reconstruction. Our code is available at https://flatcad.github.io/.
Authors: Grzegorz Gruszczynski, Jakub Meixner, Michal Jan Wlodarczyk, Przemyslaw Musialski
Abstract: We propose a novel PDE-driven corruption process for generative image synthesis based on advection-diffusion processes which generalizes existing PDE-based approaches. Our forward pass formulates image corruption via a physically motivated PDE that couples directional advection with isotropic diffusion and Gaussian noise, controlled by dimensionless numbers (Peclet, Fourier). We implement this PDE numerically through a GPU-accelerated custom Lattice Boltzmann solver for fast evaluation. To induce realistic turbulence, we generate stochastic velocity fields that introduce coherent motion and capture multi-scale mixing. In the generative process, a neural network learns to reverse the advection-diffusion operator thus constituting a novel generative model. We discuss how previous methods emerge as specific cases of our operator, demonstrating that our framework generalizes prior PDE-based corruption techniques. We illustrate how advection improves the diversity and quality of the generated images while keeping the overall color palette unaffected. This work bridges fluid dynamics, dimensionless PDE theory, and deep generative modeling, offering a fresh perspective on physically informed image corruption processes for diffusion-based synthesis.
Authors: Gilad Lerman, Kang Li, Tyler Maunu, Teng Zhang
Abstract: Robust subspace estimation is fundamental to many machine learning and data analysis tasks. Iteratively Reweighted Least Squares (IRLS) is an elegant and empirically effective approach to this problem, yet its theoretical properties remain poorly understood. This paper establishes that, under deterministic conditions, a variant of IRLS with dynamic smoothing regularization converges linearly to the underlying subspace from any initialization. We extend these guarantees to affine subspace estimation, a setting that lacks prior recovery theory. Additionally, we illustrate the practical benefits of IRLS through an application to low-dimensional neural network training. Our results provide the first global convergence guarantees for IRLS in robust subspace recovery and, more broadly, for nonconvex IRLS on a Riemannian manifold.
Authors: David Poole
Abstract: Artificial intelligence seems to be taking over the world with systems that model pixels, words, and phonemes. The world is arguably made up, not of pixels, words, and phonemes but of entities (objects, things, including events) with properties and relations among them. Surely we should model these, not the perception or description of them. You might suspect that concentrating on modeling words and pixels is because all of the (valuable) data in the world is in terms of text and images. If you look into almost any company you will find their most valuable data is in spreadsheets, databases and other relational formats. These are not the form that are studied in introductory machine learning, but are full of product numbers, student numbers, transaction numbers and other identifiers that can't be interpreted naively as numbers. The field that studies this sort of data has various names including relational learning, statistical relational AI, and many others. This paper explains why relational learning is not taking over the world -- except in a few cases with restricted relations -- and what needs to be done to bring it to it's rightful prominence.
Authors: Sahar Ghoflsaz Ghinani, Elaheh Sadredini
Abstract: Federated Learning (FL) enables collaborative model training without centralizing client data, making it attractive for privacy-sensitive domains. While existing approaches employ cryptographic techniques such as homomorphic encryption, differential privacy, or secure multiparty computation to mitigate inference attacks-including model inversion, membership inference, and gradient leakage-they often suffer from high computational, communication, or memory overheads. Moreover, many methods overlook the confidentiality of the global model itself, which may be proprietary and sensitive. These challenges limit the practicality of secure FL, especially in cross-silo deployments involving large datasets and strict compliance requirements. We present FuSeFL, a fully secure and scalable FL scheme designed for cross-silo settings. FuSeFL decentralizes training across client pairs using lightweight secure multiparty computation (MPC), while confining the server's role to secure aggregation. This design eliminates server bottlenecks, avoids data offloading, and preserves full confidentiality of data, model, and updates throughout training. FuSeFL defends against inference threats, achieves up to 95% lower communication latency and 50% lower server memory usage, and improves accuracy over prior secure FL solutions, demonstrating strong security and efficiency at scale.
Authors: William Ravenscroft, George Close, Kit Bower-Morris, Jamie Stacey, Dmitry Sityaev, Kris Y. Hong
Abstract: Large-scale in-the-wild speech datasets have become more prevalent in recent years due to increased interest in models that can learn useful features from unlabelled data for tasks such as speech recognition or synthesis. These datasets often contain undesirable features, such as multiple speakers, non-target languages, and music, which may impact model learning. The Whilter model is proposed as a multitask solution to identify these undesirable samples. Whilter uses a Whisper encoder with an attention-based classifier to solve five diverse classification problems at once. In addition, an annotated dataset is published for a subset of two popular in-the-wild corpora. Whilter achieves F1 scores above 85% and equal error rates of 6.5% to 7.8% for three of five subtasks, outperforming a state-of-the-art BEATs classifier on speech-specific classes, with a notable decrease in processing time compared to a combination of single-task alternatives.
Authors: Stefanos Gkikas, Ioannis Kyprakis, Manolis Tsiknakis
Abstract: Pain is a complex condition that affects a large portion of the population. Accurate and consistent evaluation is essential for individuals experiencing pain and supports the development of effective and advanced management strategies. Automatic pain assessment systems provide continuous monitoring, aid clinical decision-making, and aim to reduce distress while preventing functional decline. This study has been submitted to the Second Multimodal Sensing Grand Challenge for Next-Gen Pain Assessment (AI4PAIN). The proposed method introduces a pipeline that employs respiration as the input signal and integrates a highly efficient cross-attention transformer with a multi-windowing strategy. Extensive experiments demonstrate that respiration serves as a valuable physiological modality for pain assessment. Furthermore, results show that compact and efficient models, when properly optimized, can deliver strong performance, often surpassing larger counterparts. The proposed multi-window strategy effectively captures short-term and long-term features, along with global characteristics, enhancing the model's representational capacity.
Authors: Sebastian Wind, Jeta Sopa, Daniel Truhn, Mahshad Lotfinia, Tri-Thien Nguyen, Keno Bressem, Lisa Adams, Mirabela Rusu, Harald K\"ostler, Gerhard Wellein, Andreas Maier, Soroosh Tayebi Arasteh
Abstract: Clinical decision-making in radiology increasingly benefits from artificial intelligence (AI), particularly through large language models (LLMs). However, traditional retrieval-augmented generation (RAG) systems for radiology question answering (QA) typically rely on single-step retrieval, limiting their ability to handle complex clinical reasoning tasks. Here we propose an agentic RAG framework enabling LLMs to autonomously decompose radiology questions, iteratively retrieve targeted clinical evidence from Radiopaedia.org, and dynamically synthesize evidence-based responses. We evaluated 25 LLMs spanning diverse architectures, parameter scales (0.5B to >670B), and training paradigms (general-purpose, reasoning-optimized, clinically fine-tuned), using 104 expert-curated radiology questions from previously established RSNA-RadioQA and ExtendedQA datasets. To assess generalizability, we additionally tested on an unseen internal dataset of 65 real-world radiology board examination questions. Agentic retrieval significantly improved mean diagnostic accuracy over zero-shot prompting and conventional online RAG. The greatest gains occurred in small-scale models, while very large models (>200B parameters) demonstrated minimal changes (<2% improvement). Additionally, agentic retrieval reduced hallucinations (mean 9.4%) and retrieved clinically relevant context in 46% of cases, substantially aiding factual grounding. Even clinically fine-tuned models showed gains from agentic retrieval (e.g., MedGemma-27B), indicating that retrieval remains beneficial despite embedded domain knowledge. These results highlight the potential of agentic frameworks to enhance factuality and diagnostic accuracy in radiology QA, warranting future studies to validate their clinical utility. All datasets, code, and the full agentic framework are publicly available to support open research and clinical translation.
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: Bowen Zhang, Genan Dai, Hu Huang, Long Lan
Abstract: Graph Neural Networks (GNNs) have demonstrated impressive capabilities in modeling graph-structured data, while Spiking Neural Networks (SNNs) offer high energy efficiency through sparse, event-driven computation. However, existing spiking GNNs predominantly operate in Euclidean space and rely on fixed geometric assumptions, limiting their capacity to model complex graph structures such as hierarchies and cycles. To overcome these limitations, we propose \method{}, a novel Geometry-Aware Spiking Graph Neural Network that unifies spike-based neural dynamics with adaptive representation learning on Riemannian manifolds. \method{} features three key components: a Riemannian Embedding Layer that projects node features into a pool of constant-curvature manifolds, capturing non-Euclidean structures; a Manifold Spiking Layer that models membrane potential evolution and spiking behavior in curved spaces via geometry-consistent neighbor aggregation and curvature-based attention; and a Manifold Learning Objective that enables instance-wise geometry adaptation through jointly optimized classification and link prediction losses defined over geodesic distances. All modules are trained using Riemannian SGD, eliminating the need for backpropagation through time. Extensive experiments on multiple benchmarks show that GSG achieves superior accuracy, robustness, and energy efficiency compared to both Euclidean SNNs and manifold-based GNNs, establishing a new paradigm for curvature-aware, energy-efficient graph learning.
Authors: Lam Ngo, Huong Ha, Jeffrey Chan, Hongyu Zhang
Abstract: High-dimensional Bayesian Optimization (BO) has attracted significant attention in recent research. However, existing methods have mainly focused on optimizing in continuous domains, while combinatorial (ordinal and categorical) and mixed domains still remain challenging. In this paper, we first propose MOCA-HESP, a novel high-dimensional BO method for combinatorial and mixed variables. The key idea is to leverage the hyper-ellipsoid space partitioning (HESP) technique with different categorical encoders to work with high-dimensional, combinatorial and mixed spaces, while adaptively selecting the optimal encoders for HESP using a multi-armed bandit technique. Our method, MOCA-HESP, is designed as a \textit{meta-algorithm} such that it can incorporate other combinatorial and mixed BO optimizers to further enhance the optimizers' performance. Finally, we develop three practical BO methods by integrating MOCA-HESP with state-of-the-art BO optimizers for combinatorial and mixed variables: standard BO, CASMOPOLITAN, and Bounce. Our experimental results on various synthetic and real-world benchmarks show that our methods outperform existing baselines. Our code implementation can be found at https://github.com/LamNgo1/moca-hesp
Authors: Ke Ma, Jun Long, Hongxiao Fei, Liujie Hua, Yiran Qian, Zhen Dai, Yueyi Luo
Abstract: Pre-trained Vision-Language Models (VLMs) face a significant adaptation gap when applied to Zero-Shot Anomaly Detection (ZSAD), stemming from their lack of local inductive biases for dense prediction and their reliance on inflexible feature fusion paradigms. We address these limitations through an Architectural Co-Design framework that jointly refines feature representation and cross-modal fusion. Our method proposes a parameter-efficient Convolutional Low-Rank Adaptation (Conv-LoRA) adapter to inject local inductive biases for fine-grained representation, and introduces a Dynamic Fusion Gateway (DFG) that leverages visual context to adaptively modulate text prompts, enabling a powerful bidirectional fusion. Extensive experiments on diverse industrial and medical benchmarks demonstrate superior accuracy and robustness, validating that this synergistic co-design is critical for robustly adapting foundation models to dense perception tasks.
Authors: Vasileios Saketos, Sebastian Kaltenbach, Sergey Litvinov, Petros Koumoutsakos
Abstract: Algorithmic discovery has traditionally relied on human ingenuity and extensive experimentation. Here we investigate whether a prominent scientific computing algorithm, the Kalman Filter, can be discovered through an automated, data-driven, evolutionary process that relies on Cartesian Genetic Programming (CGP) and Large Language Models (LLM). We evaluate the contributions of both modalities (CGP and LLM) in discovering the Kalman filter under varying conditions. Our results demonstrate that our framework of CGP and LLM-assisted evolution converges to near-optimal solutions when Kalman optimality assumptions hold. When these assumptions are violated, our framework evolves interpretable alternatives that outperform the Kalman filter. These results demonstrate that combining evolutionary algorithms and generative models for interpretable, data-driven synthesis of simple computational modules is a potent approach for algorithmic discovery in scientific computing.
Authors: Azam Nouri
Abstract: This study investigates whether second-order geometric cues - planar curvature magnitude, curvature sign, and gradient orientation - are sufficient on their own to drive a multilayer perceptron (MLP) classifier for handwritten character recognition (HCR), offering an alternative to convolutional neural networks (CNNs). Using these three handcrafted feature maps as inputs, our curvature-orientation MLP achieves 97 percent accuracy on MNIST digits and 89 percent on EMNIST letters. These results underscore the discriminative power of curvature-based representations for handwritten character images and demonstrate that the advantages of deep learning can be realized even with interpretable, hand-engineered features.
Authors: Andrei-Valentin T\u{a}nase, Elena Pelican
Abstract: Tokenization remains a fundamental yet underexplored bottleneck in natural language processing, with strategies largely static despite remarkable progress in model architectures. We present SupraTok, a novel tokenization architecture that reimagines subword segmentation through three innovations: cross-boundary pattern learning that discovers multi-word semantic units, entropy-driven data curation that optimizes training corpus quality, and multi-phase curriculum learning for stable convergence. Our approach extends Byte-Pair Encoding by learning "superword" tokens, coherent multi-word expressions that preserve semantic unity while maximizing compression efficiency. SupraTok achieves 31% improvement in English tokenization efficiency (5.91 versus 4.51 characters per token) compared to OpenAI's o200k tokenizer and 30% improvement over Google's Gemma 3 tokenizer (256k vocabulary), while maintaining competitive performance across 38 languages. When integrated with a GPT-2 scale model (124M parameters) trained on 10 billion tokens from the FineWeb-Edu dataset, SupraTok yields 8.4% improvement on HellaSWAG and 9.5% on MMLU benchmarks without architectural modifications. While these results are promising at this scale, further validation at larger model scales is needed. These findings suggest that efficient tokenization can complement architectural innovations as a path to improved language model performance.
Authors: Azam Nouri
Abstract: We revisit the classical Sobel operator to ask a simple question: Are first-order edge maps sufficient to drive an all-dense multilayer perceptron (MLP) for handwritten character recognition (HCR), as an alternative to convolutional neural networks (CNNs)? Using only horizontal and vertical Sobel derivatives as input, we train an MLP on MNIST and EMNIST Letters. Despite its extreme simplicity, the resulting network reaches 98% accuracy on MNIST digits and 92% on EMNIST letters -- approaching CNNs while offering a smaller memory footprint and transparent features. Our findings highlight that much of the class-discriminative information in handwritten character images is already captured by first-order gradients, making edge-aware MLPs a compelling option for HCR.
Authors: Diaa Addeen Abuhani, Marco Seccaroni, Martina Mazzarello, Imran Zualkernan, Fabio Duarte, Carlo Ratti
Abstract: Urban tree biodiversity is critical for climate resilience, ecological stability, and livability in cities, yet most municipalities lack detailed knowledge of their canopies. Field-based inventories provide reliable estimates of Shannon and Simpson diversity but are costly and time-consuming, while supervised AI methods require labeled data that often fail to generalize across regions. We introduce an unsupervised clustering framework that integrates visual embeddings from street-level imagery with spatial planting patterns to estimate biodiversity without labels. Applied to eight North American cities, the method recovers genus-level diversity patterns with high fidelity, achieving low Wasserstein distances to ground truth for Shannon and Simpson indices and preserving spatial autocorrelation. This scalable, fine-grained approach enables biodiversity mapping in cities lacking detailed inventories and offers a pathway for continuous, low-cost monitoring to support equitable access to greenery and adaptive management of urban ecosystems.
Authors: Xinhua Chen, Sitao Huang, Cong Guo, Chiyue Wei, Yintao He, Jianyi Zhang, Hai "Helen" Li, Yiran Chen
Abstract: Diffusion-based Large Language Models (dLLMs) parallelize text generation by framing decoding as a denoising process, but suffer from high computational overhead since they predict all future suffix tokens at each step while retaining only a small fraction. We propose Diffusion Scratchpad (DPad), a training-free method that restricts attention to a small set of nearby suffix tokens, preserving fidelity while eliminating redundancy. DPad integrates two strategies: (i) a sliding window, which maintains a fixed-length suffix window, and (ii) distance-decay dropout, which deterministically removes distant suffix tokens before attention computation. This simple design is compatible with existing optimizations such as prefix caching and can be implemented with only a few lines of code. Comprehensive evaluations across multiple benchmarks on LLaDA-1.5 and Dream models demonstrate that DPad delivers up to $\mathbf{61.4\times}$ speedup over vanilla dLLMs while maintaining comparable accuracy, highlighting its potential for efficient and scalable long-sequence inference. Our code is available at https://github.com/Crys-Chen/DPad.
Authors: Mandeep Rathee, V Venktesh, Sean MacAvaney, Avishek Anand
Abstract: Retrieval-Augmented Generation (RAG) has emerged as a standard framework for knowledge-intensive NLP tasks, combining large language models (LLMs) with document retrieval from external corpora. Despite its widespread use, most RAG pipelines continue to treat retrieval and reasoning as isolated components, retrieving documents once and then generating answers without further interaction. This static design often limits performance on complex tasks that require iterative evidence gathering or high-precision retrieval. Recent work in both the information retrieval (IR) and NLP communities has begun to close this gap by introducing adaptive retrieval and ranking methods that incorporate feedback. In this survey, we present a structured overview of advanced retrieval and ranking mechanisms that integrate such feedback. We categorize feedback signals based on their source and role in improving the query, retrieved context, or document pool. By consolidating these developments, we aim to bridge IR and NLP perspectives and highlight retrieval as a dynamic, learnable component of end-to-end RAG systems.
Authors: Ardian Selmonaj, Miroslav Strupl, Oleg Szehr, Alessandro Antonucci
Abstract: To reliably deploy Multi-Agent Reinforcement Learning (MARL) systems, it is crucial to understand individual agent behaviors. While prior work typically evaluates overall team performance based on explicit reward signals, it is unclear how to infer agent contributions in the absence of any value feedback. In this work, we investigate whether meaningful insights into agent behaviors can be extracted solely by analyzing the policy distribution. Inspired by the phenomenon that intelligent agents tend to pursue convergent instrumental values, we introduce Intended Cooperation Values (ICVs), a method based on information-theoretic Shapley values for quantifying each agent's causal influence on their co-players' instrumental empowerment. Specifically, ICVs measure an agent's action effect on its teammates' policies by assessing their decision (un)certainty and preference alignment. By analyzing action effects on policies and value functions across cooperative and competitive MARL tasks, our method identifies which agent behaviors are beneficial to team success, either by fostering deterministic decisions or by preserving flexibility for future action choices, while also revealing the extent to which agents adopt similar or diverse strategies. Our proposed method offers novel insights into cooperation dynamics and enhances explainability in MARL systems.
Authors: Kaan Berke Ugurlar, Joaqu\'in de Navascu\'es, Michael Taynnan Barros
Abstract: Understanding the dynamic organization and homeostasis of living tissues requires high-resolution, time-resolved imaging coupled with methods capable of extracting interpretable, predictive insights from complex datasets. Here, we present the Vision Transformer Digital Twin Surrogate Network (VT-DTSN), a deep learning framework for predictive modeling of 3D+T imaging data from biological tissue. By leveraging Vision Transformers pretrained with DINO (Self-Distillation with NO Labels) and employing a multi-view fusion strategy, VT-DTSN learns to reconstruct high-fidelity, time-resolved dynamics of a Drosophila midgut while preserving morphological and feature-level integrity across imaging depths. The model is trained with a composite loss prioritizing pixel-level accuracy, perceptual structure, and feature-space alignment, ensuring biologically meaningful outputs suitable for in silico experimentation and hypothesis testing. Evaluation across layers and biological replicates demonstrates VT-DTSN's robustness and consistency, achieving low error rates and high structural similarity while maintaining efficient inference through model optimization. This work establishes VT-DTSN as a feasible, high-fidelity surrogate for cross-timepoint reconstruction and for studying tissue dynamics, enabling computational exploration of cellular behaviors and homeostasis to complement time-resolved imaging studies in biological research.
Authors: Huanpeng Chu, Wei Wu, Guanyu Fen, Yutao Zhang
Abstract: Diffusion models have emerged as a powerful paradigm for generative tasks such as image synthesis and video generation, with Transformer architectures further enhancing performance. However, the high computational cost of diffusion Transformers-stemming from a large number of sampling steps and complex per-step computations-presents significant challenges for real-time deployment. In this paper, we introduce OmniCache, a training-free acceleration method that exploits the global redundancy inherent in the denoising process. Unlike existing methods that determine caching strategies based on inter-step similarities and tend to prioritize reusing later sampling steps, our approach originates from the sampling perspective of DIT models. We systematically analyze the model's sampling trajectories and strategically distribute cache reuse across the entire sampling process. This global perspective enables more effective utilization of cached computations throughout the diffusion trajectory, rather than concentrating reuse within limited segments of the sampling procedure. In addition, during cache reuse, we dynamically estimate the corresponding noise and filter it out to reduce its impact on the sampling direction. Extensive experiments demonstrate that our approach accelerates the sampling process while maintaining competitive generative quality, offering a promising and practical solution for efficient deployment of diffusion-based generative models.
Authors: Ana-Cristina Rogoz, Radu Tudor Ionescu, Alexandra-Valentina Anghel, Ionut-Lucian Antone-Iordache, Simona Coniac, Andreea Iuliana Ionescu
Abstract: Question answering (QA) is an actively studied topic, being a core natural language processing (NLP) task that needs to be addressed before achieving Artificial General Intelligence (AGI). However, the lack of QA datasets in specific domains and languages hinders the development of robust AI models able to generalize across various domains and languages. To this end, we introduce MedQARo, the first large-scale medical QA benchmark in Romanian, alongside a comprehensive evaluation of state-of-the-art large language models (LLMs). We construct a high-quality and large-scale dataset comprising 102,646 QA pairs related to cancer patients. The questions regard medical case summaries of 1,011 patients, requiring either keyword extraction or reasoning to be answered correctly. MedQARo is the result of a time-consuming manual annotation process carried out by seven physicians specialized in oncology or radiotherapy, who spent a total of about 2,100 work hours to generate the QA pairs. We experiment with four LLMs from distinct families of models on MedQARo. Each model is employed in two scenarios, namely one based on zero-shot prompting and one based on supervised fine-tuning. Our results show that fine-tuned models significantly outperform their zero-shot counterparts, clearly indicating that pretrained models fail to generalize on MedQARo. Our findings demonstrate the importance of both domain-specific and language-specific fine-tuning for reliable clinical QA in Romanian. We publicly release our dataset and code at https://github.com/ana-rogoz/MedQARo.
Authors: Anilkumar Swamy, Vincent Leroy, Philippe Weinzaepfel, Jean-S\'ebastien Franco, Gr\'egory Rogez
Abstract: Hand-object 3D reconstruction has become increasingly important for applications in human-robot interaction and immersive AR/VR experiences. A common approach for object-agnostic hand-object reconstruction from RGB sequences involves a two-stage pipeline: hand-object 3D tracking followed by multi-view 3D reconstruction. However, existing methods rely on keypoint detection techniques, such as Structure from Motion (SfM) and hand-keypoint optimization, which struggle with diverse object geometries, weak textures, and mutual hand-object occlusions, limiting scalability and generalization. As a key enabler to generic and seamless, non-intrusive applicability, we propose in this work a robust, keypoint detector-free approach to estimating hand-object 3D transformations from monocular motion video/images. We further integrate this with a multi-view reconstruction pipeline to accurately recover hand-object 3D shape. Our method, named HOSt3R, is unconstrained, does not rely on pre-scanned object templates or camera intrinsics, and reaches state-of-the-art performance for the tasks of object-agnostic hand-object 3D transformation and shape estimation on the SHOWMe benchmark. We also experiment on sequences from the HO3D dataset, demonstrating generalization to unseen object categories.