Authors: Yifei He, Pranit Chawla, Yaser Souri, Subhojit Som, Xia Song
Abstract: Computer use agents (CUAs) can operate real-world digital interfaces but remain difficult to train due to the high cost of graphical user interface (GUI) interaction and the scarcity of high-quality trajectory data. Existing datasets rely on human demonstrations, limiting scalability. A natural alternative is to synthesize data from strong CUAs, yet their rollouts are highly noisy, with incorrect or suboptimal actions consisting a large proportion of the steps, making naive imitation ineffective. To tackle this challenge, we introduce a scalable data synthesis pipeline that transforms noisy rollouts into reliable supervision without human annotation. The core idea is step-level filtering, which evaluates actions individually to retain only correct steps, complemented by reasoning augmentation for improved planning. Using this pipeline, we construct WebSTAR, a dataset of 13.3K trajectories and 100K graded, reasoning-rich steps synthesized from OpenAI's computer-use-preview model. We train Qwen-2.5-VL-Instruct models (7B and 32B) on WebSTAR. On WebVoyager, our 7B model surpasses SoTA open-source CUA model UI-TARS-1.5-7B by more than 15% with only supervised finetuning. Building on step-level grading, we further create WebSCORE, a dataset of graded step-level actions, and train StepRM, a 7B multimodal reward model distilled from o4-mini, which matches its grading quality while being far more efficient to deploy at scale. Our results establish step-level filtering as a key principle for scalable CUA training and construct two new datasets (WebSTAR, WebSCORE) and a lightweight reward model (StepRM) as practical tools to advance robust and efficient CUAs.
Authors: Farica Zhuang, Dinara Aliyeva, Shu Yang, Zixuan Wen, Duy Duong-Tran, Christos Davatzikos, Tianlong Chen, Song Wang, Li Shen
Abstract: Accurate and early diagnosis of Alzheimer's disease (AD) can benefit from integrating complementary information from multiple modalities, mirroring clinical practice. However, conventional fusion approaches often rely on simple concatenation of features, which cannot adaptively balance the contributions of biomarkers such as amyloid PET and MRI across brain regions. In this work, we propose MREF-AD, a Multimodal Regional Expert Fusion model for AD diagnosis. It is a Mixture-of-Experts (MoE) framework that models meso-scale brain regions in each modality as an independent expert and employs two-level gating networks to learn subject-specific fusion weights. Beyond improving diagnostic performance, MREF-AD provides modality- and region-level insight into how structural and molecular imaging jointly contribute to disease diagnosis. Using data from the Alzheimer's Disease Neuroimaging Initiative (ADNI), MREF-AD achieves state-of-the-art performance over baselines while providing enhanced interpretability of brain region-specific biomarker relevance, underscoring its utility as a general framework for adaptive and interpretable multimodal fusion in neuroimaging.
Authors: Dev Vyas
Abstract: Mixture of Experts (MoE) architectures have demonstrated remarkable success in scaling neural networks, yet their application to continual learning remains fundamentally limited by a critical vulnerability: the learned gating network itself suffers from catastrophic forgetting. We introduce Mixture of Bidders (MoB), a novel framework that reconceptualizes expert routing as a decentralized economic mechanism. MoB replaces learned gating networks with Vickrey-Clarke-Groves (VCG) auctions, where experts compete for each data batch by bidding their true cost -- a principled combination of execution cost (predicted loss) and forgetting cost (Elastic Weight Consolidation penalty). This game-theoretic approach provides three key advantages: (1) {stateless routing that is immune to catastrophic forgetting, (2) \textbf{truthful bidding} guaranteed by dominant-strategy incentive compatibility, and (3) emergent specialization without explicit task boundaries. On Split-MNIST benchmarks, MoB achieves 88.77% average accuracy compared to 19.54% for Gated MoE and 27.96% for Monolithic EWC, representing a 4.5 times improvement over the strongest baseline. We further extend MoB with autonomous self-monitoring experts that detect their own knowledge consolidation boundaries, eliminating the need for explicit task demarcation.
Authors: Jiang Liu, Yujie Li, Chan Zhou, Yihao Xie, Qilong Sun, Xin Shu, Peiwei Li, Chunyong Yang, Yiziting Zhu, Jiaqi Zhu, Yuwen Chen, Bo An, Hao Wu, Bin Yi
Abstract: Objective: Sepsis is a life-threatening condition caused by severe infection leading to acute organ dysfunction. This study proposes a data-driven metric and a continuous reward function to optimize personalized heparin therapy in surgical sepsis patients. Methods: Data from the MIMIC-IV v1.0 and eICU v2.0 databases were used for model development and evaluation. The training cohort consisted of abdominal surgery patients receiving unfractionated heparin (UFH) after postoperative sepsis onset. We introduce a new RL-based framework: converting the discrete SOFA score to a continuous cxSOFA for more nuanced state and reward functions; Second, defining "good" or "bad" strategies based on cxSOFA by a stepwise manner; Third, proposing a Treatment Effect Comparison Matrix (TECM), analogous to a confusion matrix for classification tasks, to evaluate the treatment strategies. We applied different RL algorithms, Q-Learning, DQN, DDQN, BCQ and CQL to optimize the treatment and comprehensively evaluated the framework. Results: Among the AI-derived strategies, the cxSOFA-CQL model achieved the best performance, reducing mortality from 1.83% to 0.74% with the average hospital stay from 11.11 to 9.42 days. TECM demonstrated consistent outcomes across models, highlighting robustness. Conclusion: The proposed RL framework enables interpretable and robust optimization of heparin therapy in surgical sepsis. Continuous cxSOFA scoring and TECM-based evaluation provide nuanced treatment assessment, showing promise for improving clinical outcomes and decision-support reliability.
Authors: Matvey Nepomnyaschiy, Oleg Pereziabov, Anvar Tliamov, Stanislav Mikhailov, Ilya Afanasyev
Abstract: Effective human-agent interaction (HAI) relies on accurate and adaptive perception of human emotional states. While multimodal deep learning models - leveraging facial expressions, speech, and textual cues - offer high accuracy in emotion recognition, their training and maintenance are often computationally intensive and inflexible to modality changes. In this work, we propose a novel multi-agent framework for training multimodal emotion recognition systems, where each modality encoder and the fusion classifier operate as autonomous agents coordinated by a central supervisor. This architecture enables modular integration of new modalities (e.g., audio features via emotion2vec), seamless replacement of outdated components, and reduced computational overhead during training. We demonstrate the feasibility of our approach through a proof-of-concept implementation supporting vision, audio, and text modalities, with the classifier serving as a shared decision-making agent. Our framework not only improves training efficiency but also contributes to the design of more flexible, scalable, and maintainable perception modules for embodied and virtual agents in HAI scenarios.
Authors: Zhanpeng Chen, Weihao Gao, Shunyu Wang, Yanan Zhu, Hong Meng, Yuexian Zou
Abstract: Generating precise 3D molecular geometries is crucial for drug discovery and material science. While prior efforts leverage 1D representations like SELFIES to ensure molecular validity, they fail to fully exploit the rich chemical knowledge entangled within 1D models, leading to a disconnect between 1D syntactic generation and 3D geometric realization. To bridge this gap, we propose MolSculpt, a novel framework that "sculpts" 3D molecular geometries from chemical syntax. MolSculpt is built upon a frozen 1D molecular foundation model and a 3D molecular diffusion model. We introduce a set of learnable queries to extract inherent chemical knowledge from the foundation model, and a trainable projector then injects this cross-modal information into the conditioning space of the diffusion model to guide the 3D geometry generation. In this way, our model deeply integrates 1D latent chemical knowledge into the 3D generation process through end-to-end optimization. Experiments demonstrate that MolSculpt achieves state-of-the-art (SOTA) performance in \textit{de novo} 3D molecule generation and conditional 3D molecule generation, showing superior 3D fidelity and stability on both the GEOM-DRUGS and QM9 datasets. Code is available at https://github.com/SakuraTroyChen/MolSculpt.
Authors: Roy van Zuijlen, Duarte Antunes
Abstract: Memoryless and finite-memory policies offer a practical alternative for solving partially observable Markov decision processes (POMDPs), as they operate directly in the output space rather than in the high-dimensional belief space. However, extending classical methods such as policy iteration to this setting remains difficult; the output process is non-Markovian, making policy-improvement steps interdependent across stages. We introduce a new family of monotonically improving policy-iteration algorithms that alternate between single-stage output-based policy improvements and policy evaluations according to a prescribed periodic pattern. We show that this family admits optimal patterns that maximize a natural computational-efficiency index, and we identify the simplest pattern with minimal period. Building on this structure, we further develop a model-free variant that estimates values from data and learns memoryless policies directly. Across several POMDPs examples, our method achieves significant computational speedups over policy-gradient baselines and recent specialized algorithms in both model-based and model-free settings.
Authors: Duo Zhou, Jorge Chavez, Hesun Chen, Grani A. Hanasusanto, Huan Zhang
Abstract: State-of-the-art neural network (NN) verifiers demonstrate that applying the branch-and-bound (BaB) procedure with fast bounding techniques plays a key role in tackling many challenging verification properties. In this work, we introduce the linear constraint-driven clipping framework, a class of scalable and efficient methods designed to enhance the efficacy of NN verifiers. Under this framework, we develop two novel algorithms that efficiently utilize linear constraints to 1) reduce portions of the input space that are either verified or irrelevant to a subproblem in the context of branch-and-bound, and 2) directly improve intermediate bounds throughout the network. The process novelly leverages linear constraints that often arise from bound propagation methods and is general enough to also incorporate constraints from other sources. It efficiently handles linear constraints using a specialized GPU procedure that can scale to large neural networks without the use of expensive external solvers. Our verification procedure, Clip-and-Verify, consistently tightens bounds across multiple benchmarks and can significantly reduce the number of subproblems handled during BaB. We show that our clipping algorithms can be integrated with BaB-based verifiers such as $\alpha,\beta$-CROWN, utilizing either the split constraints in activation-space BaB or the output constraints that denote the unverified input space. We demonstrate the effectiveness of our procedure on a broad range of benchmarks where, in some instances, we witness a 96% reduction in the number of subproblems during branch-and-bound, and also achieve state-of-the-art verified accuracy across multiple benchmarks. Clip-and-Verify is part of the $\alpha,\beta$-CROWN verifier (http://abcrown.org), the VNN-COMP 2025 winner. Code available at https://github.com/Verified-Intelligence/Clip_and_Verify.
URLs: http://abcrown.org),, https://github.com/Verified-Intelligence/Clip_and_Verify.
Authors: Sana Rahmani, Javad Hashemi, Ali Etemad
Abstract: Label ambiguity is an inherent problem in real-world electrocardiogram (ECG) diagnosis, arising from overlapping conditions and diagnostic disagreement. However, current ECG models are trained under the assumption of clean and non-ambiguous annotations, which limits both the development and the meaningful evaluation of models under real-world conditions. Although Partial Label Learning (PLL) frameworks are designed to learn from ambiguous labels, their effectiveness in medical time-series domains, ECG in particular, remains largely unexplored. In this work, we present the first systematic study of PLL methods for ECG diagnosis. We adapt nine PLL algorithms to multi-label ECG diagnosis and evaluate them using a diverse set of clinically motivated ambiguity generation strategies, capturing both unstructured (e.g., random) and structured ambiguities (e.g., cardiologist-derived similarities, treatment relationships, and diagnostic taxonomies). Our experiments on the PTB-XL and Chapman datasets demonstrate that PLL methods vary substantially in their robustness to different types and degrees of ambiguity. Through extensive analysis, we identify key limitations of current PLL approaches in clinical settings and outline future directions for developing robust and clinically aligned ambiguity-aware learning frameworks for ECG diagnosis.
Authors: Mohammadjavad Ahmadpour, Amirmahdi Meighani, Payam Taebi, Omid Ghahroodi, Amirmohammad Izadi, Mahdieh Soleymani Baghshah
Abstract: Test-time scaling (TTS) has emerged as a powerful paradigm for improving the reasoning ability of Large Language Models (LLMs) by allocating additional computation at inference, yet its application to multimodal systems such as Vision-Language Models (VLMs) remains underexplored. In this work, we present a systematic empirical study of inference time reasoning methods applied across both open-source and closed-source VLMs on different benchmarks. Our results reveal that while closed-source models consistently benefit from structured reasoning and iterative Self-Refinement, open-source VLMs show inconsistent behavior: external verification provides the most reliable gains, whereas iterative refinement often degrades performance. We further find that the effectiveness of TTS is dataset-dependent, yielding clear improvements on multi-step reasoning tasks but offering only limited gains on perception-focused benchmarks. These findings demonstrate that TTS is not a universal solution and must be tailored to both model capabilities and task characteristics, motivating future work on adaptive TTS strategies and multimodal reward models.
Authors: Xinyu Zhang, Conor Hassan, Julien Martinelli, Daolang Huang, Samuel Kaski
Abstract: Balancing competing objectives is omnipresent across disciplines, from drug design to autonomous systems. Multi-objective Bayesian optimization is a promising solution for such expensive, black-box problems: it fits probabilistic surrogates and selects new designs via an acquisition function that balances exploration and exploitation. In practice, it requires tailored choices of surrogate and acquisition that rarely transfer to the next problem, is myopic when multi-step planning is often required, and adds refitting overhead, particularly in parallel or time-sensitive loops. We present TAMO, a fully amortized, universal policy for multi-objective black-box optimization. TAMO uses a transformer architecture that operates across varying input and objective dimensions, enabling pretraining on diverse corpora and transfer to new problems without retraining: at test time, the pretrained model proposes the next design with a single forward pass. We pretrain the policy with reinforcement learning to maximize cumulative hypervolume improvement over full trajectories, conditioning on the entire query history to approximate the Pareto frontier. Across synthetic benchmarks and real tasks, TAMO produces fast proposals, reducing proposal time by 50-1000x versus alternatives while matching or improving Pareto quality under tight evaluation budgets. These results show that transformers can perform multi-objective optimization entirely in-context, eliminating per-task surrogate fitting and acquisition engineering, and open a path to foundation-style, plug-and-play optimizers for scientific discovery workflows.
Authors: Kshitiz Khanal
Abstract: The DC Optimal Power Flow (DC-OPF) problem is fundamental to power system operations, requiring rapid solutions for real-time grid management. While traditional optimization solvers provide optimal solutions, their computational cost becomes prohibitive for large-scale systems requiring frequent recalculations. Machine learning approaches offer promise for acceleration but often struggle with constraint satisfaction and cost optimality. We present a novel two-stage learning framework that combines physics-informed Graph Neural Networks (GNNs) with Continuous Flow Matching (CFM) for solving DC-OPF problems. Our approach embeds fundamental physical principles--including economic dispatch optimality conditions, Kirchhoff's laws, and Karush-Kuhn-Tucker (KKT) complementarity conditions--directly into the training objectives. The first stage trains a GNN to produce feasible initial solutions by learning from physics-informed losses that encode power system constraints. The second stage employs CFM, a simulation-free continuous normalizing flow technique, to refine these solutions toward optimality through learned vector field regression. Evaluated on the IEEE 30-bus system across five load scenarios ranging from 70\% to 130\% nominal load, our method achieves near-optimal solutions with cost gaps below 0.1\% for nominal loads and below 3\% for extreme conditions, while maintaining 100\% feasibility. Our framework bridges the gap between fast but approximate neural network predictions and optimal but slow numerical solvers, offering a practical solution for modern power systems with high renewable penetration requiring frequent dispatch updates.
Authors: Pengfei Li, Yuelin Han, Adam Wierman, Shaolei Ren
Abstract: Fairness and action smoothness are two crucial considerations in many online optimization problems, but they have yet to be addressed simultaneously. In this paper, we study a new and challenging setting of fairness-regularized smoothed online convex optimization with switching costs. First, to highlight the fundamental challenges introduced by the long-term fairness regularizer evaluated based on the entire sequence of actions, we prove that even without switching costs, no online algorithms can possibly achieve a sublinear regret or finite competitive ratio compared to the offline optimal algorithm as the problem episode length $T$ increases. Then, we propose FairOBD (Fairness-regularized Online Balanced Descent), which reconciles the tension between minimizing the hitting cost, switching cost, and fairness cost. Concretely, FairOBD decomposes the long-term fairness cost into a sequence of online costs by introducing an auxiliary variable and then leverages the auxiliary variable to regularize the online actions for fair outcomes. Based on a new approach to account for switching costs, we prove that FairOBD offers a worst-case asymptotic competitive ratio against a novel benchmark -- the optimal offline algorithm with parameterized constraints -- by considering $T\to\infty$. Finally, we run trace-driven experiments of dynamic computing resource provisioning for socially responsible AI inference to empirically evaluate FairOBD, showing that FairOBD can effectively reduce the total fairness-regularized cost and better promote fair outcomes compared to existing baseline solutions.
Authors: Vladimer Khasia
Abstract: Implicit Neural Representations (INRs) have emerged as a powerful paradigm for parameterizing physical fields, yet they often suffer from spectral bias and the computational expense of non-convex optimization. We introduce the Vekua Layer (VL), a differentiable spectral method grounded in the classical theory of Generalized Analytic Functions. By restricting the hypothesis space to the kernel of the governing differential operator -- specifically utilizing Harmonic and Fourier-Bessel bases -- the VL transforms the learning task from iterative gradient descent to a strictly convex least-squares problem solved via linear projection. We evaluate the VL against Sinusoidal Representation Networks (SIRENs) on homogeneous elliptic Partial Differential Equations (PDEs). Our results demonstrate that the VL achieves machine precision ($\text{MSE} \approx 10^{-33}$) on exact reconstruction tasks and exhibits superior stability in the presence of incoherent sensor noise ($\text{MSE} \approx 0.03$), effectively acting as a physics-informed spectral filter. Furthermore, we show that the VL enables "holographic" extrapolation of global fields from partial boundary data via analytic continuation, a capability absent in standard coordinate-based approximations.
Authors: Lennard Manuel, Hamid Gadirov, Steffen Frey
Abstract: Analyzing and visualizing scientific ensemble datasets with high dimensionality and complexity poses significant challenges. Dimensionality reduction techniques and autoencoders are powerful tools for extracting features, but they often struggle with such high-dimensional data. This paper presents an enhanced autoencoder framework that incorporates a clustering loss, based on the soft silhouette score, alongside a contrastive loss to improve the visualization and interpretability of ensemble datasets. First, EfficientNetV2 is used to generate pseudo-labels for the unlabeled portions of the scientific ensemble datasets. By jointly optimizing the reconstruction, clustering, and contrastive objectives, our method encourages similar data points to group together while separating distinct clusters in the latent space. UMAP is subsequently applied to this latent representation to produce 2D projections, which are evaluated using the silhouette score. Multiple types of autoencoders are evaluated and compared based on their ability to extract meaningful features. Experiments on two scientific ensemble datasets - channel structures in soil derived from Markov chain Monte Carlo, and droplet-on-film impact dynamics - show that models incorporating clustering or contrastive loss marginally outperform the baseline approaches.
Authors: Takuya Kurihana, Xiaojian Zhang, Wing Yee Au, Hon Yung Wong
Abstract: Modern cities are increasingly reliant on data-driven insights to support decision making in areas such as transportation, public safety and environmental impact. However, city-level data often exists in heterogeneous formats, collected independently by local agencies with diverse objectives and standards. Despite their numerous, wide-ranging, and uniformly consumable nature, national-level datasets exhibit significant heterogeneity and multi-modality. This research proposes a heterogeneous data pipeline that performs cross-domain data fusion over time-varying, spatial-varying and spatial-varying time-series datasets. We aim to address complex urban problems across multiple domains and localities by harnessing the rich information over 50 data sources. Specifically, our data-learning module integrates homophily from spatial-varying dataset into graph-learning, embedding information of various localities into models. We demonstrate the generalizability and flexibility of the framework through five real-world observations using a variety of publicly accessible datasets (e.g., ride-share, traffic crash, and crime reports) collected from multiple cities. The results show that our proposed framework demonstrates strong predictive performance while requiring minimal reconfiguration when transferred to new localities or domains. This research advances the goal of building data-informed urban systems in a scalable way, addressing one of the most pressing challenges in smart city analytics.
Authors: Wei Duan, Jie Lu, En Yu, Junyu Xuan
Abstract: Graph-based multi-agent reinforcement learning (MARL) enables coordinated behavior under partial observability by modeling agents as nodes and communication links as edges. While recent methods excel at learning sparse coordination graphs-determining who communicates with whom-they do not address what information should be transmitted under hard bandwidth constraints. We study this bandwidth-limited regime and show that naive dimensionality reduction consistently degrades coordination performance. Hard bandwidth constraints force selective encoding, but deterministic projections lack mechanisms to control how compression occurs. We introduce Bandwidth-constrained Variational Message Encoding (BVME), a lightweight module that treats messages as samples from learned Gaussian posteriors regularized via KL divergence to an uninformative prior. BVME's variational framework provides principled, tunable control over compression strength through interpretable hyperparameters, directly constraining the representations used for decision-making. Across SMACv1, SMACv2, and MPE benchmarks, BVME achieves comparable or superior performance while using 67--83% fewer message dimensions, with gains most pronounced on sparse graphs where message quality critically impacts coordination. Ablations reveal U-shaped sensitivity to bandwidth, with BVME excelling at extreme ratios while adding minimal overhead.
Authors: Alwin Jin, Sean M. Hendryx, Vaskar Nath
Abstract: Current benchmarks that test LLMs on static, already-solved problems (e.g., math word problems) effectively demonstrated basic capability acquisition. The natural progression has been toward larger, more comprehensive and challenging collections of static problems, an approach that inadvertently constrains the kinds of advances we can measure and incentivize. To address this limitation, we argue for progress-oriented benchmarks, problem environments whose objectives are themselves the core targets of scientific progress, so that achieving state of the art on the benchmark advances the field. As a introductory step, we instantiate an environment based on the NanoGPT speedrun. The environment standardizes a dataset slice, a reference model and training harness, and rich telemetry, with run-time verification and anti-gaming checks. Evaluation centers on the scientific delta achieved: best-attained loss and the efficiency frontier. Using this environment, we achieve a new state-of-the-art training time, improving upon the previous record by 3 seconds, and qualitatively observe the emergence of novel algorithmic ideas. Moreover, comparisons between models and agents remain possible, but they are a means, not the end; the benchmark's purpose is to catalyze reusable improvements to the language modeling stack. With this release, the overarching goal is to seed a community shift from static problem leaderboards to test-time research on open-ended yet measurable scientific problems. In this new paradigm, progress on the benchmark is progress on the science, thus reframing "benchmarking" as a vehicle for scientific advancement.
Authors: Conor Rowan
Abstract: Physics-informed machine learning uses governing ordinary and/or partial differential equations to train neural networks to represent the solution field. Like any machine learning problem, the choice of activation function influences the characteristics and performance of the solution obtained from physics-informed training. Several studies have compared common activation functions on benchmark differential equations, and have unanimously found that the rectified linear unit (ReLU) is outperformed by competitors such as the sigmoid, hyperbolic tangent, and swish activation functions. In this work, we diagnose the poor performance of ReLU on physics-informed machine learning problems. While it is well-known that the piecewise linear form of ReLU prevents it from being used on second-order differential equations, we show that ReLU fails even on variational problems involving only first derivatives. We identify the cause of this failure as second derivatives of the activation, which are taken not in the formulation of the loss, but in the process of training. Namely, we show that automatic differentiation in PyTorch fails to characterize derivatives of discontinuous fields, which causes the gradient of the physics-informed loss to be mis-specified, thus explaining the poor performance of ReLU.
Authors: Divya Kothandaraman, Jaclyn Pytlarz
Abstract: Memorization in large-scale text-to-image diffusion models poses significant security and intellectual property risks, enabling adversarial attribute extraction and the unauthorized reproduction of sensitive or proprietary features. While conventional dememorization techniques, such as regularization and data filtering, limit overfitting to specific training examples, they fail to systematically prevent the internalization of prohibited concept-level features. Simply discarding all images containing a sensitive feature wastes invaluable training data, necessitating a method for selective unlearning at the concept level. To address this, we introduce a Gradient Projection Framework designed to enforce a stringent requirement of concept-level feature exclusion. Our defense operates during backpropagation by systematically identifying and excising training signals aligned with embeddings of prohibited attributes. Specifically, we project each gradient update onto the orthogonal complement of the sensitive feature's embedding space, thereby zeroing out its influence on the model's weights. Our method integrates seamlessly into standard diffusion model training pipelines and complements existing defenses. We analyze our method against an adversary aiming for feature extraction. In extensive experiments, we demonstrate that our framework drastically reduces memorization while rigorously preserving generation quality and semantic fidelity. By reframing memorization control as selective learning, our approach establishes a new paradigm for IP-safe and privacy-preserving generative AI.
Authors: Ryoma Sato, Shinji Ito
Abstract: We point out that EXP3 can be implemented in constant time per round, propose more practical algorithms, and analyze the trade-offs between the regret bounds and time complexities of these algorithms.
Authors: Haoyue Dai, Yiwen Qiu, Ignavier Ng, Xinshuai Dong, Peter Spirtes, Kun Zhang
Abstract: Addressing selection bias in latent variable causal discovery is important yet underexplored, largely due to a lack of suitable statistical tools: While various tools beyond basic conditional independencies have been developed to handle latent variables, none have been adapted for selection bias. We make an attempt by studying rank constraints, which, as a generalization to conditional independence constraints, exploits the ranks of covariance submatrices in linear Gaussian models. We show that although selection can significantly complicate the joint distribution, interestingly, the ranks in the biased covariance matrices still preserve meaningful information about both causal structures and selection mechanisms. We provide a graph-theoretic characterization of such rank constraints. Using this tool, we demonstrate that the one-factor model, a classical latent variable model, can be identified under selection bias. Simulations and real-world experiments confirm the effectiveness of using our rank constraints.
Authors: Adilet Metinov, Gulida M. Kudakeeva, Bolotbek uulu Nursultan, Gulnara D. Kabaeva
Abstract: We present Adaptive Soft Rolling KV Freeze with Entropy-Guided Recovery (ASR-KF-EGR), a training-free inference-time framework for efficient large language model generation. Our method introduces a reversible soft-freeze mechanism that temporarily suspends key-value (KV) updates for low-importance tokens identified within a sliding attention window. Unlike eviction-based approaches that permanently discard context, ASR-KF-EGR preserves all tokens in off-GPU storage and restores them on demand. We extend the framework with sublinear freeze scheduling, where freeze duration grows sublinearly with repeated low-importance detections, preventing over-aggressive compression. Preliminary experiments on LLaMA-3 8B demonstrate 55-67% reduction in active KV cache size while maintaining generation quality and passing needle-in-haystack retrieval tests. The method is architecture-agnostic, requires no fine-tuning, and provides a practical solution for memory-constrained deployment of long-context LLMs.
Authors: Jianyu Wang, Jacob Nean-Hua Sheikh, Cat P. Le, Hoda Bidkhori
Abstract: Lifelong deep learning (LDL) trains neural networks to learn sequentially across tasks while preserving prior knowledge. We propose Task-Aware Multi-Expert (TAME), a continual learning algorithm that leverages task similarity to guide expert selection and knowledge transfer. TAME maintains a pool of pretrained neural networks and activates the most relevant expert for each new task. A shared dense layer integrates features from the chosen expert to generate predictions. To reduce catastrophic forgetting, TAME uses a replay buffer that stores representative samples and embeddings from previous tasks and reuses them during training. An attention mechanism further prioritizes the most relevant stored information for each prediction. Together, these components allow TAME to adapt flexibly while retaining important knowledge across evolving task sequences. Experiments on binary classification tasks derived from CIFAR-100 show that TAME improves accuracy on new tasks while sustaining performance on earlier ones, highlighting its effectiveness in balancing adaptation and retention in lifelong learning settings.
Authors: Yunkai Zhang, Yawen Zhang, Ming Zheng, Kezhen Chen, Chongyang Gao, Ruian Ge, Siyuan Teng, Amine Jelloul, Jinmeng Rao, Xiaoyuan Guo, Chiang-Wei Fang, Zeyu Zheng, Jie Yang
Abstract: Time-series data is critical across many scientific and industrial domains, including environmental analysis, agriculture, transportation, and finance. However, mining insights from this data typically requires deep domain expertise, a process that is both time-consuming and labor-intensive. In this paper, we propose \textbf{Insight Miner}, a large-scale multimodal model (LMM) designed to generate high-quality, comprehensive time-series descriptions enriched with domain-specific knowledge. To facilitate this, we introduce \textbf{TS-Insights}\footnote{Available at \href{https://huggingface.co/datasets/zhykoties/time-series-language-alignment}{https://huggingface.co/datasets/zhykoties/time-series-language-alignment}.}, the first general-domain dataset for time series and language alignment. TS-Insights contains 100k time-series windows sampled from 20 forecasting datasets. We construct this dataset using a novel \textbf{agentic workflow}, where we use statistical tools to extract features from raw time series before synthesizing them into coherent trend descriptions with GPT-4. Following instruction tuning on TS-Insights, Insight Miner outperforms state-of-the-art multimodal models, such as LLaVA \citep{liu2023llava} and GPT-4, in generating time-series descriptions and insights. Our findings suggest a promising direction for leveraging LMMs in time series analysis, and serve as a foundational step toward enabling LLMs to interpret time series as a native input modality.
URLs: https://huggingface.co/datasets/zhykoties/time-series-language-alignment, https://huggingface.co/datasets/zhykoties/time-series-language-alignment
Authors: Francesco Innocenti, El Mehdi Achour
Abstract: In-context learning (ICL) refers to the ability of a model to learn new tasks from examples in its input without any parameter updates. In contrast to previous theories of ICL relying on toy models and data settings, recently it has been shown that an abstraction of a transformer block can be seen as implicitly updating the weights of its feedforward network according to the context (Dherin et al., 2025). Here, we provide a simple generalisation of this result for (i) all sequence positions beyond the last, (ii) any transformer block beyond the first, and (iii) more realistic residual blocks including layer normalisation. We empirically verify our theory on simple in-context linear regression tasks and investigate the relationship between the implicit updates related to different tokens within and between blocks. These results help to bring the theory of Dherin et al. (2025) even closer to practice, with potential for validation on large-scale models.
Authors: Albert Miao, Chenliang Zhou, Jiawei Zhou, Cengiz Oztireli
Abstract: Sparse Autoencoders (SAEs) are a powerful dictionary learning technique for decomposing neural network activations, translating the hidden state into human ideas with high semantic value despite no external intervention or guidance. However, this technique has rarely been applied outside of the textual domain, limiting theoretical explorations of feature decomposition. We present the \textbf{first application of SAEs to the 3D domain}, analyzing the features used by a state-of-the-art 3D reconstruction VAE applied to 53k 3D models from the Objaverse dataset. We observe that the network encodes discrete rather than continuous features, leading to our key finding: \textbf{such models approximate a discrete state space, driven by phase-like transitions from feature activations}. Through this state transition framework, we address three otherwise unintuitive behaviors -- the inclination of the reconstruction model towards positional encoding representations, the sigmoidal behavior of reconstruction loss from feature ablation, and the bimodality in the distribution of phase transition points. This final observation suggests the model \textbf{redistributes the interference caused by superposition to prioritize the saliency of different features}. Our work not only compiles and explains unexpected phenomena regarding feature decomposition, but also provides a framework to explain the model's feature learning dynamics. The code and dataset of encoded 3D objects will be available on release.
Authors: Hao Zhou (Beijing University of Posts,Telecommunications), Suman Sourav (Aalborg University), Binbin Chen (Singapore University of Technology,Design), Ke Yu (Beijing University of Posts,Telecommunications)
Abstract: Programmable Logic Controllers (PLCs) are critical components in Industrial Control Systems (ICSs). Their potential exposure to external world makes them susceptible to cyber-attacks. Existing detection methods against controller logic attacks use either specification-based or learnt models. However, specification-based models require experts' manual efforts or access to PLC's source code, while machine learning-based models often fall short of providing explanation for their decisions. We design SRLR -- a it Symbolic Regression based Logic Recovery} solution to identify the logic of a PLC based only on its inputs and outputs. The recovered logic is used to generate explainable rules for detecting controller logic attacks. SRLR enhances the latest deep symbolic regression methods using the following ICS-specific properties: (1) some important ICS control logic is best represented in frequency domain rather than time domain; (2) an ICS controller can operate in multiple modes, each using different logic, where mode switches usually do not happen frequently; (3) a robust controller usually filters out outlier inputs as ICS sensor data can be noisy; and (4) with the above factors captured, the degree of complexity of the formulas is reduced, making effective search possible. Thanks to these enhancements, SRLR consistently outperforms all existing methods in a variety of ICS settings that we evaluate. In terms of the recovery accuracy, SRLR's gain can be as high as 39% in some challenging environment. We also evaluate SRLR on a distribution grid containing hundreds of voltage regulators, demonstrating its stability in handling large-scale, complex systems with varied configurations.
Authors: Hideo Mukai, Hoshitaro Ohnishi
Abstract: Quantum computers have the possibility of a much reduced calculation load compared with classical computers in specific problems. Quantum error correction (QEC) is vital for handling qubits, which are vulnerable to external noise. In QEC, actual errors are predicted from the results of syndrome measurements by stabilizer generators, in place of making direct measurements of the data qubits. Here, we propose Quantum Golay code Error Correction (QGEC), a QEC method using Golay code, which is an efficient coding method in classical information theory. We investigated our method's ability in decoding calculations with the Transformer. We evaluated the accuracy of the decoder in a code space defined by the generative polynomials with three different weights sets and three noise models with different correlations of bit-flip error and phase-flip error. Furthermore, under a noise model following a discrete uniform distribution, we compared the decoding performance of Transformer decoders with identical architectures trained respectively on Golay and toric codes. The results showed that the noise model with the smaller correlation gave better accuracy, while the weights of the generative polynomials had little effect on the accuracy of the decoder. In addition, they showed that Golay code requiring 23 data qubits and having a code distance of 7 achieved higher decoding accuracy than toric code which requiring 50 data qubits and having a code distance of 5. This suggests that implementing quantum error correction using a Transformer may enable the Golay code to realize fault-tolerant quantum computation more efficiently.
Authors: Pranav Guruprasad, Sudipta Chowdhury, Harsh Sikka, Mridul Sharma, Helen Lu, Sean Rivera, Aryan Khurana, Hangliang Ren, Yangyue Wang
Abstract: Generalist multimodal agents are expected to unify perception, language, and control - operating robustly across diverse real world domains. However, current evaluation practices remain fragmented across isolated benchmarks, making it difficult to assess whether today's foundation models truly generalize beyond their training distributions. We introduce MultiNet v1.0, a unified benchmark for measuring the cross domain generality of vision language models (VLMs) and vision language action models (VLAs) across six foundational capability regimes. Visual grounding, spatial reasoning, tool use, physical commonsense, multi agent coordination, and continuous robot control. Evaluating GPT 5, Pi0, and Magma, we find that no model demonstrates consistent generality. All exhibit substantial degradation on unseen domains, unfamiliar modalities, or cross domain task shifts despite strong performance within their training distributions.These failures manifest as modality misalignment, output format instability, and catastrophic knowledge degradation under domain transfer.Our findings reveal a persistent gap between the aspiration of generalist intelligence and the actual capabilities of current foundation models.MultiNet v1.0 provides a standardized evaluation substrate for diagnosing these gaps and guiding the development of future generalist agents.Code, data, and leaderboards are publicly available.
Authors: Tingxu Yan, Ye Yuan
Abstract: Dynamic graphs are prevalent in real-world scenarios, where continuous structural changes induce catastrophic forgetting in graph neural networks (GNNs). While continual learning has been extended to dynamic graphs, existing methods overlook the effects of topological changes on existing nodes. To address it, we propose a novel framework for continual learning on dynamic graphs, named Condensation-Concatenation-based Continual Learning (CCC). Specifically, CCC first condenses historical graph snapshots into compact semantic representations while aiming to preserve the original label distribution and topological properties. Then it concatenates these historical embeddings with current graph representations selectively. Moreover, we refine the forgetting measure (FM) to better adapt to dynamic graph scenarios by quantifying the predictive performance degradation of existing nodes caused by structural updates. CCC demonstrates superior performance over state-of-the-art baselines across four real-world datasets in extensive experiments.
Authors: Sara Sameer, Wei Zhang, Kannan Dhivya Dharshini, Xin Lou, Yulin Gao, Terence Goh, Qingyu Yan
Abstract: Batteries are critical components in modern energy systems such as electric vehicles and power grid energy storage. Effective battery health management is essential for battery system safety, cost-efficiency, and sustainability. In this paper, we propose Pace, a physics-aware attentive temporal convolutional network for battery health estimation. Pace integrates raw sensor measurements with battery physics features derived from the equivalent circuit model. We develop three battery-specific modules, including dilated temporal blocks for efficient temporal encoding, chunked attention blocks for context modeling, and a dual-head output block for fusing short- and long-term battery degradation patterns. Together, the modules enable Pace to predict battery health accurately and efficiently in various battery usage conditions. In a large public dataset, Pace performs much better than existing models, achieving an average performance improvement of 6.5 and 2.0x compared to two best-performing baseline models. We further demonstrate its practical viability with a real-time edge deployment on a Raspberry Pi. These results establish Pace as a practical and high-performance solution for battery health analytics.
Authors: Cong Yao, Chunye Gong, Jin Zhang
Abstract: Accurate core loss modeling is critical for the design of high-efficiency power electronic systems. Traditional core loss modeling methods have limitations in prediction accuracy. To advance this field, the IEEE Power Electronics Society launched the MagNet Challenge in 2023, the first international competition focused on data-driven power electronics design methods, aiming to uncover complex loss patterns in magnetic components through a data-driven paradigm. Although purely data-driven models demonstrate strong fitting performance, their interpretability and cross-distribution generalization capabilities remain limited. To address these issues, this paper proposes a hybrid model, SEPI-TFPNet, which integrates empirical models with deep learning. The physical-prior submodule employs a spectral entropy discrimination mechanism to select the most suitable empirical model under different excitation waveforms. The data-driven submodule incorporates convolutional neural networks, multi-head attention mechanisms, and bidirectional long short-term memory networks to extract flux-density time-series features. An adaptive feature fusion module is introduced to improve multimodal feature interaction and integration. Using the MagNet dataset containing various magnetic materials, this paper evaluates the proposed method and compares it with 21 representative models from the 2023 challenge and three advanced methods from 2024-2025. The results show that the proposed method achieves improved modeling accuracy and robustness.
Authors: Jinming Ge, Linfeng Du, Likith Anaparty, Shangkun Li, Tingyuan Liang, Afzal Ahmad, Vivek Chaturvedi, Sharad Sinha, Zhiyao Xie, Jiang Xu, Wei Zhang
Abstract: High-Level Synthesis (HLS) tools are widely adopted in FPGA-based domain-specific accelerator design. However, existing tools rely on fixed optimization strategies inherited from software compilations, limiting their effectiveness. Tailoring optimization strategies to specific designs requires deep semantic understanding, accurate hardware metric estimation, and advanced search algorithms -- capabilities that current approaches lack. We propose DAPO, a design structure-aware pass ordering framework that extracts program semantics from control and data flow graphs, employs contrastive learning to generate rich embeddings, and leverages an analytical model for accurate hardware metric estimation. These components jointly guide a reinforcement learning agent to discover design-specific optimization strategies. Evaluations on classic HLS designs demonstrate that our end-to-end flow delivers a 2.36 speedup over Vitis HLS on average.
Authors: Minwoo Park, Junwoo Chang, Jongeun Choi, Roberto Horowitz
Abstract: Equivariant diffusion policies (EDPs) combine the generative expressivity of diffusion models with the strong generalization and sample efficiency afforded by geometric symmetries. While steering these policies with reinforcement learning (RL) offers a promising mechanism for fine-tuning beyond demonstration data, directly applying standard (non-equivariant) RL can be sample-inefficient and unstable, as it ignores the symmetries that EDPs are designed to exploit. In this paper, we theoretically establish that the diffusion process of an EDP is equivariant, which in turn induces a group-invariant latent-noise MDP that is well-suited for equivariant diffusion steering. Building on this theory, we introduce a principled symmetry-aware steering framework and compare standard, equivariant, and approximately equivariant RL strategies through comprehensive experiments across tasks with varying degrees of symmetry. While we identify the practical boundaries of strict equivariance under symmetry breaking, we show that exploiting symmetry during the steering process yields substantial benefits-enhancing sample efficiency, preventing value divergence, and achieving strong policy improvements even when EDPs are trained from extremely limited demonstrations.
Authors: Jie Wang, Zheng Yan, Jiahe Lan, Xuyan Li, Elisa Bertino
Abstract: Trust prediction provides valuable support for decision-making, risk mitigation, and system security enhancement. Recently, Graph Neural Networks (GNNs) have emerged as a promising approach for trust prediction, owing to their ability to learn expressive node representations that capture intricate trust relationships within a network. However, current GNN-based trust prediction models face several limitations: (i) Most of them fail to capture trust dynamicity, leading to questionable inferences. (ii) They rarely consider the heterogeneous nature of real-world networks, resulting in a loss of rich semantics. (iii) None of them support context-awareness, a basic property of trust, making prediction results coarse-grained. To this end, we propose CAT, the first Context-Aware GNN-based Trust prediction model that supports trust dynamicity and accurately represents real-world heterogeneity. CAT consists of a graph construction layer, an embedding layer, a heterogeneous attention layer, and a prediction layer. It handles dynamic graphs using continuous-time representations and captures temporal information through a time encoding function. To model graph heterogeneity and leverage semantic information, CAT employs a dual attention mechanism that identifies the importance of different node types and nodes within each type. For context-awareness, we introduce a new notion of meta-paths to extract contextual features. By constructing context embeddings and integrating a context-aware aggregator, CAT can predict both context-aware trust and overall trust. Extensive experiments on three real-world datasets demonstrate that CAT outperforms five groups of baselines in trust prediction, while exhibiting strong scalability to large-scale graphs and robustness against both trust-oriented and GNN-oriented attacks.
Authors: Yifan Niu, Aochuan Chen, Tingyang Xu, Jia Li
Abstract: It has been demonstrated that adversarial graphs, i.e., graphs with imperceptible perturbations, can cause deep graph models to fail on classification tasks. In this work, we extend the concept of adversarial graphs to the community detection problem, which is more challenging. We propose novel attack and defense techniques for community detection problem, with the objective of hiding targeted individuals from detection models and enhancing the robustness of community detection models, respectively. These techniques have many applications in real-world scenarios, for example, protecting personal privacy in social networks and understanding camouflage patterns in transaction networks. To simulate interactive attack and defense behaviors, we further propose a game-theoretic framework, called CD-GAME. One player is a graph attacker, while the other player is a Rayleigh Quotient defender. The CD-GAME models the mutual influence and feedback mechanisms between the attacker and the defender, revealing the dynamic evolutionary process of the game. Both players dynamically update their strategies until they reach the Nash equilibrium. Extensive experiments demonstrate the effectiveness of our proposed attack and defense methods, and both outperform existing baselines by a significant margin. Furthermore, CD-GAME provides valuable insights for understanding interactive attack and defense scenarios in community detection problems. We found that in traditional single-step attack or defense, attacker tends to employ strategies that are most effective, but are easily detected and countered by defender. When the interactive game reaches a Nash equilibrium, attacker adopts more imperceptible strategies that can still achieve satisfactory attack effectiveness even after defense.
Authors: Yifan Niu, Han Xiao, Dongyi Liu, Nuo Chen, Jia Li
Abstract: As Large Language Models (LLMs) are increasingly deployed in real-world applications, it is important to ensure their behaviors align with human values, societal norms, and ethical principles. However, safety alignment under Reinforcement Learning (RL) often suffers from forgetting learned general abilities, which is also known as the alignment tax. To address this issue, we introduce Null-Space constrained Policy Optimization (NSPO), a novel RL framework for LLM safety alignment while preserving their core abilities. The safety policy gradients are geometrically projected into the null space of general tasks, thereby mitigating the safety alignment tax. In addition, we theoretically prove that NSPO preserves the model's original core capabilities, while still guaranteeing a descent direction for effective safety alignment. Extensive experiments demonstrate that NSPO outperforms existing methods by a large margin, achieving state-of-the-art safety performance without sacrificing accuracy on general tasks, including math, code, and instruction-following tasks. Notably, NSPO is data-efficient and only requires 40% of public human-annotated safety data from PKU-SafeRLHF to achieve promising safety performance, without a large amount of mixed general tasks data in existing alignment methods.
Authors: S Sairam, Prateek P Kulkarni
Abstract: We present a novel approach to neural representation learning that incorporates algebraic constraints inspired by Bhargava cubes from number theory. Traditional deep learning methods learn representations in unstructured latent spaces lacking interpretability and mathematical consistency. Our framework maps input data to constrained 3-dimensional latent spaces where embeddings are regularized to satisfy learned quadratic relationships derived from Bhargava's combinatorial structures. The architecture employs a differentiable auxiliary loss function operating independently of classification objectives, guiding models toward mathematically structured representations. We evaluate on MNIST, achieving 99.46% accuracy while producing interpretable 3D embeddings that naturally cluster by digit class and satisfy learned quadratic constraints. Unlike existing manifold learning approaches requiring explicit geometric supervision, our method imposes weak algebraic priors through differentiable constraints, ensuring compatibility with standard optimization. This represents the first application of number-theoretic constructs to neural representation learning, establishing a foundation for incorporating structured mathematical priors in neural networks.
Authors: Siwan Boufad\`ene (LIGM), Fran\c{c}ois-Xavier Vialard (LIGM)
Abstract: We introduce sliced ReLU attention, a new attention mechanism that departs structurally from both softmax and ReLU-based alternatives. Instead of applying a nonlinearity to pairwise dot products, we operate on one-dimensional projections of key--query differences and leverage sorting to obtain quasi-linear complexity. This construction yields a differentiable, non-symmetric kernel that can be computed in O(n log(n)) through a sorting procedure, making it suitable for very long contexts. Beyond computational benefits, the model retains strong theoretical expressive power: we establish two in-context expressivity results, previously known for softmax attention, showing that sliced ReLU attention preserves the ability to perform nontrivial sequence-to-sequence disentangling tasks and satisfies a contextual universal approximation property. Finally, we illustrate the potential practical interest of this kernel in small-scale experiments.
Authors: Arghya Pratihar, Arnab Seal, Swagatam Das, Inesh Chattopadhyay
Abstract: Clustering is a fundamental unsupervised learning task for uncovering patterns in data. While Gaussian Blurring Mean Shift (GBMS) has proven effective for identifying arbitrarily shaped clusters in Euclidean space, it struggles with datasets exhibiting hierarchical or tree-like structures. In this work, we introduce HypeGBMS, a novel extension of GBMS to hyperbolic space. Our method replaces Euclidean computations with hyperbolic distances and employs M\"obius-weighted means to ensure that all updates remain consistent with the geometry of the space. HypeGBMS effectively captures latent hierarchies while retaining the density-seeking behavior of GBMS. We provide theoretical insights into convergence and computational complexity, along with empirical results that demonstrate improved clustering quality in hierarchical datasets. This work bridges classical mean-shift clustering and hyperbolic representation learning, offering a principled approach to density-based clustering in curved spaces. Extensive experimental evaluations on $11$ real-world datasets demonstrate that HypeGBMS significantly outperforms conventional mean-shift clustering methods in non-Euclidean settings, underscoring its robustness and effectiveness.
Authors: Bowen Ding, Yuhan Chen, Jiayang Lv, Jiyao Yuan, Qi Zhu, Shuangshuang Tian, Dantong Zhu, Futing Wang, Heyuan Deng, Fei Mi, Lifeng Shang, Tao Lin
Abstract: While effective post-training integrates Supervised Fine-Tuning (SFT) and Reinforcement Learning (RL), the optimal mechanism for utilizing expert trajectories remains unresolved. We propose the Plasticity-Ceiling Framework to theoretically ground this landscape, decomposing performance into foundational SFT performance and the subsequent RL plasticity. Through extensive benchmarking, we establish the Sequential SFT-then-RL pipeline as the superior standard, overcoming the stability deficits of synchronized approaches. Furthermore, we derive precise scaling guidelines: (1) Transitioning to RL at the SFT Stable or Mild Overfitting Sub-phase maximizes the final ceiling by securing foundational SFT performance without compromising RL plasticity; (2) Refuting ``Less is More'' in the context of SFT-then-RL scaling, we demonstrate that Data Scale determines the primary post-training potential, while Trajectory Difficulty acts as a performance multiplier; and (3) Identifying that the Minimum SFT Validation Loss serves as a robust indicator for selecting the expert trajectories that maximize the final performance ceiling. Our findings provide actionable guidelines for maximizing the value extracted from expert trajectories.
Authors: Hao Wu, Yuan Gao, Fan Xu, Fan Zhang, Guangliang Liu, Yuxuan Liang, Xiaomeng Huang
Abstract: High-precision scientific simulation faces a long-standing trade-off between computational efficiency and physical fidelity. To address this challenge, we propose NeuralOGCM, an ocean modeling framework that fuses differentiable programming with deep learning. At the core of NeuralOGCM is a fully differentiable dynamical solver, which leverages physics knowledge as its core inductive bias. The learnable physics integration captures large-scale, deterministic physical evolution, and transforms key physical parameters (e.g., diffusion coefficients) into learnable parameters, enabling the model to autonomously optimize its physical core via end-to-end training. Concurrently, a deep neural network learns to correct for subgrid-scale processes and discretization errors not captured by the physics model. Both components work in synergy, with their outputs integrated by a unified ODE solver. Experiments demonstrate that NeuralOGCM maintains long-term stability and physical consistency, significantly outperforming traditional numerical models in speed and pure AI baselines in accuracy. Our work paves a new path for building fast, stable, and physically-plausible models for scientific computing.
Authors: Joel Ekstrand, Zahra Taghiyarrenani, Slawomir Nowaczyk
Abstract: Time series forecasting predicts future values from past data. In real-world settings, some anomalous events have lasting effects and influence the forecast, while others are short-lived and should be ignored. Standard forecasting models fail to make this distinction, often either overreacting to noise or missing persistent shifts. We propose Co-TSFA (Contrastive Time Series Forecasting with Anomalies), a regularization framework that learns when to ignore anomalies and when to respond. Co-TSFA generates input-only and input-output augmentations to model forecast-irrelevant and forecast-relevant anomalies, and introduces a latent-output alignment loss that ties representation changes to forecast changes. This encourages invariance to irrelevant perturbations while preserving sensitivity to meaningful distributional shifts. Experiments on the Traffic and Electricity benchmarks, as well as on a real-world cash-demand dataset, demonstrate that Co-TSFA improves performance under anomalous conditions while maintaining accuracy on normal data. An anonymized GitHub repository with the implementation of Co-TSFA is provided and will be made public upon acceptance.
Authors: Qingxiao Sun, Tongxuan Liu, Shen Zhang, Siyu Wu, Peijun Yang, Haotian Liang, Menxin Li, Xiaolong Ma, Zhiwei Liang, Ziyi Ren, Minchao Zhang, Xinyu Liu, Ke Zhang, Depei Qian, Hailong Yang
Abstract: Recommendation system delivers substantial economic benefits by providing personalized predictions. Generative recommendation (GR) integrates LLMs to enhance the understanding of long user-item sequences. Despite employing attention-based architectures, GR's workload differs markedly from that of LLM serving. GR typically processes long prompt while producing short, fixed-length outputs, yet the computational cost of each decode phase is especially high due to the large beam width. In addition, since the beam search involves a vast item space, the sorting overhead becomes particularly time-consuming. We propose xGR, a GR-oriented serving system that meets strict low-latency requirements under highconcurrency scenarios. First, xGR unifies the processing of prefill and decode phases through staged computation and separated KV cache. Second, xGR enables early sorting termination and mask-based item filtering with data structure reuse. Third, xGR reconstructs the overall pipeline to exploit multilevel overlap and multi-stream parallelism. Our experiments with real-world recommendation service datasets demonstrate that xGR achieves at least 3.49x throughput compared to the state-of-the-art baseline under strict latency constraints.
Authors: \'Alvaro Leitao, Jonatan R\'afales
Abstract: In this work, we introduce a machine/deep learning methodology to solve parametric integrals. Besides classical machine learning approaches, we consider a differential learning framework that incorporates derivative information during training, emphasizing its advantageous properties. Our study covers three representative problem classes: statistical functionals (including moments and cumulative distribution functions), approximation of functions via Chebyshev expansions, and integrals arising directly from differential equations. These examples range from smooth closed-form benchmarks to challenging numerical integrals. Across all cases, the differential machine learning-based approach consistently outperforms standard architectures, achieving lower mean squared error, enhanced scalability, and improved sample efficiency.
Authors: Emmanuel K. Katalay, David O. Dimandja, Jordan F. Masakuna
Abstract: The performance of machine learning (ML) models often deteriorates when the underlying data distribution changes over time, a phenomenon known as data distribution drift. When this happens, ML models need to be retrained and redeployed. ML Operations (MLOps) is often manual, i.e., humans trigger the process of model retraining and redeployment. In this work, we present an automated MLOps pipeline designed to address neural network classifier retraining in response to significant data distribution changes. Our MLOps pipeline employs multi-criteria statistical techniques to detect distribution shifts and triggers model updates only when necessary, ensuring computational efficiency and resource optimization. We demonstrate the effectiveness of our framework through experiments on several benchmark anomaly detection data sets, showing significant improvements in model accuracy and robustness compared to traditional retraining strategies. Our work provides a foundation for deploying more reliable and adaptive ML systems in dynamic real-world settings, where data distribution changes are common.
Authors: Federico Pennino, Maurizio Gabbrielli
Abstract: The standard paradigm for training deep learning models on sensor data assumes that more data is always better. However, raw sensor streams are often imbalanced and contain significant redundancy, meaning that not all data points contribute equally to model generalization. In this paper, we show that, in some cases, "less is more" when considering datasets. We do this by reframing the data selection problem: rather than tuning model hyperparameters, we fix the model and optimize the composition of the training data itself. We introduce a framework for discovering the optimal "training diet" from a large, unlabeled time series corpus. Our framework first uses a large-scale encoder and k-means clustering to partition the dataset into distinct, behaviorally consistent clusters. These clusters represent the fundamental 'ingredients' available for training. We then employ the Optuna optimization framework to search the high-dimensional space of possible data mixtures. For each trial, Optuna proposes a specific sampling ratio for each cluster, and a new training set is constructed based on this recipe. A smaller target model is then trained and evaluated. Our experiments reveal that this data-centric search consistently discovers data mixtures that yield models with significantly higher performance compared to baselines trained on the entire dataset. Specifically - evaluated on PMSM dataset - our method improved performance from a baseline MSE of 1.70 to 1.37, a 19.41% improvement.
Authors: Janaina Mour\~ao-Miranda, Zakria Hussain, Konstantinos Tsirlis, Christophe Phillips, John Shawe-Taylor
Abstract: Multiple Kernel Learning (MKL) models combine several kernels in supervised and unsupervised settings to integrate multiple data representations or sources, each represented by a different kernel. MKL seeks an optimal linear combination of base kernels that maximizes a generalized performance measure under a regularization constraint. Various norms have been used to regularize the kernel weights, including $l1$, $l2$ and $lp$, as well as the "elastic-net" penalty, which combines $l1$- and $l2$-norm to promote both sparsity and the selection of correlated kernels. This property makes elastic-net regularized MKL (ENMKL) especially valuable when model interpretability is critical and kernels capture correlated information, such as in neuroimaging. Previous ENMKL methods have followed a two-stage procedure: fix kernel weights, train a support vector machine (SVM) with the weighted kernel, and then update the weights via gradient descent, cutting-plane methods, or surrogate functions. Here, we introduce an alternative ENMKL formulation that yields a simple analytical update for the kernel weights. We derive explicit algorithms for both SVM and kernel ridge regression (KRR) under this framework, and implement them in the open-source Pattern Recognition for Neuroimaging Toolbox (PRoNTo). We evaluate these ENMKL algorithms against $l1$-norm MKL and against SVM (or KRR) trained on the unweighted sum of kernels across three neuroimaging applications. Our results show that ENMKL matches or outperforms $l1$-norm MKL in all tasks and only underperforms standard SVM in one scenario. Crucially, ENMKL produces sparser, more interpretable models by selectively weighting correlated kernels.
Authors: Dooho Lee, Myeong Kong, Minho Jeong, Jaemin Yoo
Abstract: Generalizing a pretrained model to unseen datasets without retraining is an essential step toward a foundation model. However, achieving such cross-dataset, fully inductive inference is difficult in graph-structured data where feature spaces vary widely in both dimensionality and semantics. Any transformation in the feature space can easily violate the inductive applicability to unseen datasets, strictly limiting the design space of a graph model. In this work, we introduce the view space, a novel representational axis in which arbitrary graphs can be naturally encoded in a unified manner. We then propose Graph View Transformation (GVT), a node- and feature-permutation-equivariant mapping in the view space. GVT serves as the building block for Recurrent GVT, a fully inductive model for node representation learning. Pretrained on OGBN-Arxiv and evaluated on 27 node-classification benchmarks, Recurrent GVT outperforms GraphAny, the prior fully inductive graph model, by +8.93% and surpasses 12 individually tuned GNNs by at least +3.30%. These results establish the view space as a principled and effective ground for fully inductive node representation learning.
Authors: Sam Gijsen, Marc-Andre Schulz, Kerstin Ritter
Abstract: The development of foundation models for functional magnetic resonance imaging (fMRI) time series holds significant promise for predicting phenotypes related to disease and cognition. Current models, however, are often trained using a mask-and-reconstruct objective on small brain regions. This focus on low-level information leads to representations that are sensitive to noise and temporal fluctuations, necessitating extensive fine-tuning for downstream tasks. We introduce Brain-Semantoks, a self-supervised framework designed specifically to learn abstract representations of brain dynamics. Its architecture is built on two core innovations: a semantic tokenizer that aggregates noisy regional signals into robust tokens representing functional networks, and a self-distillation objective that enforces representational stability across time. We show that this objective is stabilized through a novel training curriculum, ensuring the model robustly learns meaningful features from low signal-to-noise time series. We demonstrate that learned representations enable strong performance on a variety of downstream tasks even when only using a linear probe. Furthermore, we provide comprehensive scaling analyses indicating more unlabeled data reliably results in out-of-distribution performance gains without domain adaptation.
Authors: Stefan Tabakov, Asen Popov, Dimitar Dimitrov, S. Ensiye Kiyamousavi, Vladimir Hristov, Boris Kraychev
Abstract: Current vision-language-action (VLA) models generalize poorly, particularly when tasks require new compositions of skills or objects. We introduce Atomic Action Slicing (AAS), a planner-aligned approach that decomposes long-horizon demonstrations into short, typed atomic actions that are easier for planners to use and policies to learn. Using LIBERO demonstrations, AAS produces a validated dataset of 2,124 atomic segments labeled with action type, temporal span, and confidence. A stronger segmenter (Gemini 2.5 Pro) closely matches planner-defined plans and remains robust under keyframe jitter, while smaller models perform worse on multi-object tasks. Fine-tuning CLIP-RT+ on our atomic dataset improves task success from 94.2% to 95.3% on LIBERO-Goal and 83.8% to 88.8% on LIBERO-Long. We publicly release the GATE-VLAP dataset on HuggingFace(https://huggingface.co/datasets/gate-institute/GATE-VLAP-datasets)
URLs: https://huggingface.co/datasets/gate-institute/GATE-VLAP-datasets)
Authors: Alexander Tyurin
Abstract: Even for the gradient descent (GD) method applied to neural network training, understanding its optimization dynamics, including convergence rate, iterate trajectories, function value oscillations, and especially its implicit acceleration, remains a challenging problem. We analyze nonlinear models with the logistic loss and show that the steps of GD reduce to those of generalized perceptron algorithms (Rosenblatt, 1958), providing a new perspective on the dynamics. This reduction yields significantly simpler algorithmic steps, which we analyze using classical linear algebra tools. Using these tools, we demonstrate on a minimalistic example that the nonlinearity in a two-layer model can provably yield a faster iteration complexity $\tilde{O}(\sqrt{d})$ compared to $\Omega(d)$ achieved by linear models, where $d$ is the number of features. This helps explain the optimization dynamics and the implicit acceleration phenomenon observed in neural networks. The theoretical results are supported by extensive numerical experiments. We believe that this alternative view will further advance research on the optimization of neural networks.
Authors: Javier Fumanal-Idocin, Raquel Fernandez-Peralta, Javier Andreu-Perez
Abstract: Fuzzy rule-based systems have been mostly used in interpretable decision-making because of their interpretable linguistic rules. However, interpretability requires both sensible linguistic partitions and small rule-base sizes, which are not guaranteed by many existing fuzzy rule-mining algorithms. Evolutionary approaches can produce high-quality models but suffer from prohibitive computational costs, while neural-based methods like ANFIS have problems retaining linguistic interpretations. In this work, we propose an adaptation of classical tree-based splitting algorithms from crisp rules to fuzzy trees, combining the computational efficiency of greedy algoritms with the interpretability advantages of fuzzy logic. This approach achieves interpretable linguistic partitions and substantially improves running time compared to evolutionary-based approaches while maintaining competitive predictive performance. Our experiments on tabular classification benchmarks proof that our method achieves comparable accuracy to state-of-the-art fuzzy classifiers with significantly lower computational cost and produces more interpretable rule bases with constrained complexity. Code is available in: https://github.com/Fuminides/fuzzy_greedy_tree_public
Authors: Afonso Louren\c{c}o, Jo\~ao Gama, Eric P. Xing, Goreti Marreiros
Abstract: In streaming scenarios, models must learn continuously, adapting to concept drifts without erasing previously acquired knowledge. However, existing research communities address these challenges in isolation. Continual Learning (CL) focuses on long-term retention and mitigating catastrophic forgetting, often without strict real-time constraints. Stream Learning (SL) emphasizes rapid, efficient adaptation to high-frequency data streams, but typically neglects forgetting. Recent efforts have tried to combine these paradigms, yet no clear algorithmic overlap exists. We argue that large in-context tabular models (LTMs) provide a natural bridge for Streaming Continual Learning (SCL). In our view, unbounded streams should be summarized on-the-fly into compact sketches that can be consumed by LTMs. This recovers the classical SL motivation of compressing massive streams with fixed-size guarantees, while simultaneously aligning with the experience-replay desiderata of CL. To clarify this bridge, we show how the SL and CL communities implicitly adopt a divide-to-conquer strategy to manage the tension between plasticity (performing well on the current distribution) and stability (retaining past knowledge), while also imposing a minimal complexity constraint that motivates diversification (avoiding redundancy in what is stored) and retrieval (re-prioritizing past information when needed). Within this perspective, we propose structuring SCL with LTMs around two core principles of data selection for in-context learning: (1) distribution matching, which balances plasticity and stability, and (2) distribution compression, which controls memory size through diversification and retrieval mechanisms.
Authors: Sebastian Hirt, Valentinus Suwanto, Hendrik Alsmeier, Maik Pfefferkorn, Rolf Findeisen
Abstract: Learning controller parameters from closed-loop data has been shown to improve closed-loop performance. Bayesian optimization, a widely used black-box and sample-efficient learning method, constructs a probabilistic surrogate of the closed-loop performance from few experiments and uses it to select informative controller parameters. However, it typically struggles with dense high-dimensional controller parameterizations, as they may appear, for example, in tuning model predictive controllers, because standard surrogate models fail to capture the structure of such spaces. This work suggests that the use of Bayesian neural networks as surrogate models may help to mitigate this limitation. Through a comparison between Gaussian processes with Matern kernels, finite-width Bayesian neural networks, and infinite-width Bayesian neural networks on a cart-pole task, we find that Bayesian neural network surrogate models achieve faster and more reliable convergence of the closed-loop cost and enable successful optimization of parameterizations with hundreds of dimensions. Infinite-width Bayesian neural networks also maintain performance in settings with more than one thousand parameters, whereas Matern-kernel Gaussian processes rapidly lose effectiveness. These results indicate that Bayesian neural network surrogate models may be suitable for learning dense high-dimensional controller parameterizations and offer practical guidance for selecting surrogate models in learning-based controller design.
Authors: Aditya Tripathi, Karan Sharma, Rahul Mishra, Tapas Kumar Maiti
Abstract: Federated Learning (FL) distributes model training across clients who retain their data locally, but this architecture exposes a fundamental vulnerability: Byzantine clients can inject arbitrarily corrupted updates that degrade or subvert the global model. While robust aggregation methods (including Krum, Bulyan, and coordinate-wise defenses) offer theoretical guarantees under idealized assumptions, their effectiveness erodes substantially when client data distributions are heterogeneous (non-IID) and adversaries can observe or approximate the defense mechanism. This paper introduces SpectralKrum, a defense that fuses spectral subspace estimation with geometric neighbor-based selection. The core insight is that benign optimization trajectories, despite per-client heterogeneity, concentrate near a low-dimensional manifold that can be estimated from historical aggregates. SpectralKrum projects incoming updates into this learned subspace, applies Krum selection in compressed coordinates, and filters candidates whose orthogonal residual energy exceeds a data-driven threshold. The method requires no auxiliary data, operates entirely on model updates, and preserves FL privacy properties. We evaluate SpectralKrum against eight robust baselines across seven attack scenarios on CIFAR-10 with Dirichlet-distributed non-IID partitions (alpha = 0.1). Experiments spanning over 56,000 training rounds show that SpectralKrum is competitive against directional and subspace-aware attacks (adaptive-steer, buffer-drift), but offers limited advantage under label-flip and min-max attacks where malicious updates remain spectrally indistinguishable from benign ones.
Authors: Vladimer Khasia
Abstract: Coordinate-based neural networks have emerged as a powerful tool for representing continuous physical fields, yet they face two fundamental pathologies: spectral bias, which hinders the learning of high-frequency dynamics, and the curse of dimensionality, which causes parameter explosion in discrete feature grids. We propose the Adaptive Vekua Cascade (AVC), a hybrid architecture that bridges deep learning and classical approximation theory. AVC decouples manifold learning from function approximation by using a deep network to learn a diffeomorphic warping of the physical domain, projecting complex spatiotemporal dynamics onto a latent manifold where the solution is represented by a basis of generalized analytic functions. Crucially, we replace the standard gradient-descent output layer with a differentiable linear solver, allowing the network to optimally resolve spectral coefficients in a closed form during the forward pass. We evaluate AVC on a suite of five rigorous physics benchmarks, including high-frequency Helmholtz wave propagation, sparse medical reconstruction, and unsteady 3D Navier-Stokes turbulence. Our results demonstrate that AVC achieves state-of-the-art accuracy while reducing parameter counts by orders of magnitude (e.g., 840 parameters vs. 4.2 million for 3D grids) and converging 2-3x faster than implicit neural representations. This work establishes a new paradigm for memory-efficient, spectrally accurate scientific machine learning. The code is available at https://github.com/VladimerKhasia/vecua.
Authors: Etienne Boursier, Claire Boyer
Abstract: Softmax attention is a central component of transformer architectures, yet its nonlinear structure poses significant challenges for theoretical analysis. We develop a unified, measure-based framework for studying single-layer softmax attention under both finite and infinite prompts. For i.i.d. Gaussian inputs, we lean on the fact that the softmax operator converges in the infinite-prompt limit to a linear operator acting on the underlying input-token measure. Building on this insight, we establish non-asymptotic concentration bounds for the output and gradient of softmax attention, quantifying how rapidly the finite-prompt model approaches its infinite-prompt counterpart, and prove that this concentration remains stable along the entire training trajectory in general in-context learning settings with sub-Gaussian tokens. In the case of in-context linear regression, we use the tractable infinite-prompt dynamics to analyze training at finite prompt length. Our results allow optimization analyses developed for linear attention to transfer directly to softmax attention when prompts are sufficiently long, showing that large-prompt softmax attention inherits the analytical structure of its linear counterpart. This, in turn, provides a principled and broadly applicable toolkit for studying the training dynamics and statistical behavior of softmax attention layers in large prompt regimes.
Authors: Ahmad Shamail, Claire McWhite
Abstract: Many systems exhibit complex interactions between their components: some features or actions amplify each other's effects, others provide redundant information, and some contribute independently. We present a simple geometric method for discovering interactions and redundancies: when elements are added in random sequential orders and their contributions plotted over many trials, characteristic L-shaped patterns emerge that directly reflect interaction structure. The approach quantifies how the contribution of each element depends on those added before it, revealing patterns that distinguish interaction, independence, and redundancy on a unified scale. When pairwise contributions are visualized as two--dimensional point clouds, redundant pairs form L--shaped patterns where only the first-added element contributes, while synergistic pairs form L--shaped patterns where only elements contribute together. Independent elements show order--invariant distributions. We formalize this with the L--score, a continuous measure ranging from $-1$ (perfect synergy, e.g. $Y=X_1X_2$) to $0$ (independence) to $+1$ (perfect redundancy, $X_1 \approx X_2$). The relative scaling of the L--shaped arms reveals feature dominance in which element consistently provides more information. Although computed only from pairwise measurements, higher--order interactions among three or more elements emerge naturally through consistent cross--pair relationships (e.g. AB, AC, BC). The method is metric--agnostic and broadly applicable to any domain where performance can be evaluated incrementally over non-repeating element sequences, providing a unified geometric approach to uncovering interaction structure.
Authors: Zheqi Hu, Xuanjing Chen, Jinlin Hu
Abstract: With the rapid growth of AI-generated content (AIGC) across domains such as music, video, and literature, the demand for emotionally aware recommendation systems has become increasingly important. Traditional recommender systems primarily rely on user behavioral data such as clicks, views, or ratings, while neglecting users' real-time emotional and intentional states during content interaction. To address this limitation, this study proposes a Multi-Modal Emotion and Intent Recognition Model (MMEI) based on a BERT-based Cross-Modal Transformer with Attention-Based Fusion, integrated into a cloud-native personalized AIGC recommendation framework. The proposed system jointly processes visual (facial expression), auditory (speech tone), and textual (comments or utterances) modalities through pretrained encoders ViT, Wav2Vec2, and BERT, followed by an attention-based fusion module to learn emotion-intent representations. These embeddings are then used to drive personalized content recommendations through a contextual matching layer. Experiments conducted on benchmark emotion datasets (AIGC-INT, MELD, and CMU-MOSEI) and an AIGC interaction dataset demonstrate that the proposed MMEI model achieves a 4.3% improvement in F1-score and a 12.3% reduction in cross-entropy loss compared to the best fusion-based transformer baseline. Furthermore, user-level online evaluations reveal that emotion-driven recommendations increase engagement time by 15.2% and enhance satisfaction scores by 11.8%, confirming the model's effectiveness in aligning AI-generated content with users' affective and intentional states. This work highlights the potential of cross-modal emotional intelligence for next-generation AIGC ecosystems, enabling adaptive, empathetic, and context-aware recommendation experiences.
Authors: Qiming Zhang, Xiucheng Wang, Nan Cheng, Zhisheng Yin, Xiang Li
Abstract: Radio maps (RMs) provide a spatially continuous description of wireless propagation, enabling cross-layer optimization and unifying communication and sensing for integrated sensing and communications (ISAC). However, constructing high-fidelity RMs at operational scales is difficult, since physics-based solvers are time-consuming and require precise scene models, while learning methods degrade under incomplete priors and sparse measurements, often smoothing away critical discontinuities. We present RMSup, a physics-informed super-resolution framework that functions with uniform sparse sampling and imperfect environment priors. RMSup extracts Helmholtz equation-informed boundary and singularity prompts from the measurements, fuses them with base-station side information and coarse scene descriptors as conditional inputs, and employs a boundary-aware dual-head network to reconstruct a high-fidelity RM and recover environmental contours jointly. Experimental results show the proposed RMsup achieves state-of-the-art performance both in RM construction and ISAC-related environment sensing.
Authors: Arif D\"onmez
Abstract: We propose Symmetry-Loss, a brain-inspired algorithmic principle that enforces invariance and equivariance through a differentiable constraint derived from environmental symmetries. The framework models learning as the iterative refinement of an effective symmetry group, paralleling developmental processes in which cortical representations align with the world's structure. By minimizing structural surprise, i.e. deviations from symmetry consistency, Symmetry-Loss operationalizes a Free-Energy--like objective for representation learning. This formulation bridges predictive-coding and group-theoretic perspectives, showing how efficient, stable, and compositional representations can emerge from symmetry-based self-organization. The result is a general computational mechanism linking developmental learning in the brain with principled representation learning in artificial systems.
Authors: Igor Pivovarov, Sergey Shumsky
Abstract: The existence of 'what' and 'where' pathways of information processing in the brain was proposed almost 30 years ago, but there is still a lack of a clear mathematical model that could show how these pathways work together. We propose a biologically inspired mathematical model that uses this idea to identify and separate the self from the environment and then build and use a self-model for better predictions. This is a model of neocortical columns governed by the basal ganglia to make predictions and choose the next action, where some columns act as 'what' columns and others act as 'where' columns. Based on this model, we present a reinforcement learning agent that learns purposeful behavior in a virtual environment. We evaluate the agent on the Atari games Pong and Breakout, where it successfully learns to play. We conclude that the ability to separate the self from the environment gives advantages to the agent and therefore such a model could appear in living organisms during evolution. We propose Self-Awareness Principle 1: the ability to separate the self from the world is a necessary but insufficient condition for self-awareness.
Authors: Michal Sanocki, Julija Zavadlav
Abstract: The vastness of chemical space makes generalization a central challenge in the development of machine learning interatomic potentials (MLIPs). While MLIPs could enable large-scale atomistic simulations with near-quantum accuracy, their usefulness is often limited by poor transferability to out-of-distribution samples. Here, we systematically evaluate different MLIP architectures with long-range corrections across diverse chemical spaces and show that such schemes are essential, not only for improving in-distribution performance but, more importantly, for enabling significant gains in transferability to unseen regions of chemical space. To enable a more rigorous benchmarking, we introduce biased train-test splitting strategies, which explicitly test the model performance in significantly different regions of chemical space. Together, our findings highlight the importance of long-range modeling for achieving generalizable MLIPs and provide a framework for diagnosing systematic failures across chemical space. Although we demonstrate our methodology on metal-organic frameworks, it is broadly applicable to other materials, offering insights into the design of more robust and transferable MLIPs.
Authors: Sharvaj Kubal, Naomi Graham, Matthieu Heitz, Andrew Warren, Michael P. Friedlander, Yaniv Plan, Geoffrey Schiebinger
Abstract: We present an approach to denoising spatial transcriptomics images that is particularly effective for uncovering cell identities in the regime of ultra-low sequencing depths, and also allows for interpolation of gene expression. The method -- Spatial Transcriptomics via Adaptive Regularization and Kernels (STARK) -- augments kernel ridge regression with an incrementally adaptive graph Laplacian regularizer. In each iteration, we (1) perform kernel ridge regression with a fixed graph to update the image, and (2) update the graph based on the new image. The kernel ridge regression step involves reducing the infinite dimensional problem on a space of images to finite dimensions via a modified representer theorem. Starting with a purely spatial graph, and updating it as we improve our image makes the graph more robust to noise in low sequencing depth regimes. We show that the aforementioned approach optimizes a block-convex objective through an alternating minimization scheme wherein the sub-problems have closed form expressions that are easily computed. This perspective allows us to prove convergence of the iterates to a stationary point of this non-convex objective. Statistically, such stationary points converge to the ground truth with rate $\mathcal{O}(R^{-1/2})$ where $R$ is the number of reads. In numerical experiments on real spatial transcriptomics data, the denoising performance of STARK, evaluated in terms of label transfer accuracy, shows consistent improvement over the competing methods tested.
Authors: Muhammad Usamah Shahid, Muddassar Farooq
Abstract: Cancer patients may undergo lengthy and painful chemotherapy treatments, comprising several successive regimens or plans. Treatment inefficacy and other adverse events can lead to discontinuation (or failure) of these plans, or prematurely changing them, which results in a significant amount of physical, financial, and emotional toxicity to the patients and their families. In this work, we build treatment failure models based on the Real World Evidence (RWE) gathered from patients' profiles available in our oncology EMR/EHR system. We also describe our feature engineering pipeline, experimental methods, and valuable insights obtained about treatment failures from trained models. We report our findings on five primary cancer types with the most frequent treatment failures (or discontinuations) to build unique and novel feature vectors from the clinical notes, diagnoses, and medications that are available in our oncology EMR. After following a novel three axes - performance, complexity and explainability - design exploration framework, boosted random forests are selected because they provide a baseline accuracy of 80% and an F1 score of 75%, with reduced model complexity, thus making them more interpretable to and usable by oncologists.
Authors: Anoop Krishnan
Abstract: In the quest for fairness in artificial intelligence, novel approaches to enhance it in facial image based gender classification algorithms using text guided methodologies are presented. The core methodology involves leveraging semantic information from image captions during model training to improve generalization capabilities. Two key strategies are presented: Image Text Matching (ITM) guidance and Image Text fusion. ITM guidance trains the model to discern fine grained alignments between images and texts to obtain enhanced multimodal representations. Image text fusion combines both modalities into comprehensive representations for improved fairness. Exensive experiments conducted on benchmark datasets demonstrate these approaches effectively mitigate bias and improve accuracy across gender racial groups compared to existing methods. Additionally, the unique integration of textual guidance underscores an interpretable and intuitive training paradigm for computer vision systems. By scrutinizing the extent to which semantic information reduces disparities, this research offers valuable insights into cultivating more equitable facial analysis algorithms. The proposed methodologies contribute to addressing the pivotal challenge of demographic bias in gender classification from facial images. Furthermore, this technique operates in the absence of demographic labels and is application agnostic.
Authors: Joe Suk, Samory Kpotufe
Abstract: We study outlier (a.k.a., anomaly) detection for single-pass non-stationary streaming data. In the well-studied offline or batch outlier detection problem, traditional methods such as kernel One-Class SVM (OCSVM) are both computationally heavy and prone to large false-negative (Type II) errors under non-stationarity. To remedy this, we introduce SONAR, an efficient SGD-based OCSVM solver with strongly convex regularization. We show novel theoretical guarantees on the Type I/II errors of SONAR, superior to those known for OCSVM, and further prove that SONAR ensures favorable lifelong learning guarantees under benign distribution shifts. In the more challenging problem of adversarial non-stationary data, we show that SONAR can be used within an ensemble method and equipped with changepoint detection to achieve adaptive guarantees, ensuring small Type I/II errors on each phase of data. We validate our theoretical findings on synthetic and real-world datasets.
Authors: Christopher Driggers-Ellis, Detravious Brinkley, Ray Chen, Aashish Dhawan, Daisy Zhe Wang, Christan Grant
Abstract: Multi30k is frequently cited in the multimodal machine translation (MMT) literature, offering parallel text data for training and fine-tuning deep learning models. However, it is limited to four languages: Czech, English, French, and German. This restriction has led many researchers to focus their investigations only on these languages. As a result, MMT research on diverse languages has been stalled because the official Multi30k dataset only represents European languages in Latin scripts. Previous efforts to extend Multi30k exist, but the list of supported languages, represented language families, and scripts is still very short. To address these issues, we propose MultiScript30k, a new Multi30k dataset extension for global languages in various scripts, created by translating the English version of Multi30k (Multi30k-En) using NLLB200-3.3B. The dataset consists of over \(30000\) sentences and provides translations of all sentences in Multi30k-En into Ar, Es, Uk, Zh\_Hans and Zh\_Hant. Similarity analysis shows that Multi30k extension consistently achieves greater than \(0.8\) cosine similarity and symmetric KL divergence less than \(0.000251\) for all languages supported except Zh\_Hant which is comparable to the previous Multi30k extensions ArEnMulti30k and Multi30k-Uk. COMETKiwi scores reveal mixed assessments of MultiScript30k as a translation of Multi30k-En in comparison to the related work. ArEnMulti30k scores nearly equal MultiScript30k-Ar, but Multi30k-Uk scores $6.4\%$ greater than MultiScript30k-Uk per split.
Authors: Kata Vuk, Nicolas Alexander Ihlo, Merle Behr
Abstract: Feature and Interaction Importance (FII) methods are essential in supervised learning for assessing the relevance of input variables and their interactions in complex prediction models. In many domains, such as personalized medicine, local interpretations for individual predictions are often required, rather than global scores summarizing overall feature importance. Random Forests (RFs) are widely used in these settings, and existing interpretability methods typically exploit tree structures and split statistics to provide model-specific insights. However, theoretical understanding of local FII methods for RF remains limited, making it unclear how to interpret high importance scores for individual predictions. We propose a novel, local, model-specific FII method that identifies frequent co-occurrences of features along decision paths, combining global patterns with those observed on paths specific to a given test point. We prove that our method consistently recovers the true local signal features and their interactions under a Locally Spike Sparse (LSS) model and also identifies whether large or small feature values drive a prediction. We illustrate the usefulness of our method and theoretical results through simulation studies and a real-world data example.
Authors: Devansh Arpit
Abstract: We identify test prediction variance (TPV) -- the first-order sensitivity of model outputs to parameter perturbations around a trained solution -- as a unifying quantity that links several classical observations about generalization in deep networks. TPV is a fully label-free object whose trace form separates the geometry of the trained model from the specific perturbation mechanism, allowing a broad family of parameter perturbations like SGD noise, label noise, finite-precision noise, and other post-training perturbations to be analyzed under a single framework. Theoretically, we show that TPV estimated on the training set converges to its test-set value in the overparameterized limit, providing the first result that prediction variance under local parameter perturbations can be inferred from training inputs alone. Empirically, TPV exhibits a striking stability across datasets and architectures -- including extremely narrow networks -- and correlates well with clean test loss. Finally, we demonstrate that modeling pruning as a TPV perturbation yields a simple label-free importance measure that performs competitively with state-of-the-art pruning methods, illustrating the practical utility of TPV. Code available at github.com/devansharpit/TPV.
Authors: Biraj Dahal, Jiahui Cheng, Hao Liu, Rongjie Lai, Wenjing Liao
Abstract: Many problems in science and engineering involve time-dependent, high dimensional datasets arising from complex physical processes, which are costly to simulate. In this work, we propose WeldNet: Windowed Encoders for Learning Dynamics, a data-driven nonlinear model reduction framework to build a low-dimensional surrogate model for complex evolution systems. Given time-dependent training data, we split the time domain into multiple overlapping windows, within which nonlinear dimension reduction is performed by auto-encoders to capture latent codes. Once a low-dimensional representation of the data is learned, a propagator network is trained to capture the evolution of the latent codes in each window, and a transcoder is trained to connect the latent codes between adjacent windows. The proposed windowed decomposition significantly simplifies propagator training by breaking long-horizon dynamics into multiple short, manageable segments, while the transcoders ensure consistency across windows. In addition to the algorithmic framework, we develop a mathematical theory establishing the representation power of WeldNet under the manifold hypothesis, justifying the success of nonlinear model reduction via deep autoencoder-based architectures. Our numerical experiments on various differential equations indicate that WeldNet can capture nonlinear latent structures and their underlying dynamics, outperforming both traditional projection-based approaches and recently developed nonlinear model reduction methods.
Authors: Akhil S Anand, Elias Aarekol, Martin Mziray Dalseg, Magnus Stalhane, Sebastien Gros
Abstract: Combinatorial sequential decision making problems are typically modeled as mixed integer linear programs (MILPs) and solved via branch and bound (B&B) algorithms. The inherent difficulty of modeling MILPs that accurately represent stochastic real world problems leads to suboptimal performance in the real world. Recently, machine learning methods have been applied to build MILP models for decision quality rather than how accurately they model the real world problem. However, these approaches typically rely on supervised learning, assume access to true optimal decisions, and use surrogates for the MILP gradients. In this work, we introduce a proof of concept CORL framework that end to end fine tunes an MILP scheme using reinforcement learning (RL) on real world data to maximize its operational performance. We enable this by casting an MILP solved by B&B as a differentiable stochastic policy compatible with RL. We validate the CORL method in a simple illustrative combinatorial sequential decision making example.
Authors: Tri Le, Khang Nguyen, Baoru Huang, Tung D. Ta, Anh Nguyen
Abstract: Crafting computer-aided design (CAD) models has long been a painstaking and time-intensive task, demanding both precision and expertise from designers. With the emergence of 3D generation, this task has undergone a transformative impact, shifting not only from visual fidelity to functional utility but also enabling editable CAD designs. Prior works have achieved early success in single-part CAD generation, which is not well-suited for real-world applications, as multiple parts need to be assembled under semantic and geometric constraints. In this paper, we propose CADKnitter, a compositional CAD generation framework with a geometry-guided diffusion sampling strategy. CADKnitter is able to generate a complementary CAD part that follows both the geometric constraints of the given CAD model and the semantic constraints of the desired design text prompt. We also curate a dataset, so-called KnitCAD, containing over 310,000 samples of CAD models, along with textual prompts and assembly metadata that provide semantic and geometric constraints. Intensive experiments demonstrate that our proposed method outperforms other state-of-the-art baselines by a clear margin.
Authors: Adilet Metinov, Gulida M. Kudakeeva, Gulnara D. Kabaeva
Abstract: Current AI code generation systems suffer from significant latency bottlenecks due to CPU-GPU data transfers during compilation, execution, and testing phases. We establish theoretical foundations for three complementary approaches to GPU-native compilation that eliminate these transfers: (1) parallel traditional compilation adapted for GPU execution, (2) neural compilation using learned sequence-to-sequence translation with probabilistic verification, and (3) hybrid architectures combining both strategies. We derive latency and energy bounds demonstrating potential speedups of 10-100x for code iteration cycles. Our analysis shows that traditional GPU compilation provides 2-5x improvements through transfer elimination, neural compilation achieves 10-100x speedups via massive parallelism, and hybrid approaches offer practical deployment paths with guaranteed correctness. We formalize the probabilistic verification framework that enables trading compilation accuracy for parallel exploration, and discuss implications for self-improving AI systems and future analog computing substrates.
Authors: John F. Wu, Joshua E. G. Peek, Sophie J. Miller, Jenny Novacescu, Achu J. Usha, Christopher A. Wilkinson
Abstract: Telescope bibliographies record the pulse of astronomy research by capturing publication statistics and citation metrics for telescope facilities. Robust and scalable bibliographies ensure that we can measure the scientific impact of our facilities and archives. However, the growing rate of publications threatens to outpace our ability to manually label astronomical literature. We therefore present the Automated Mission Classifier (amc), a tool that uses large language models (LLMs) to identify and categorize telescope references by processing large quantities of paper text. A modified version of amc performs well on the TRACS Kaggle challenge, achieving a macro $F_1$ score of 0.84 on the held-out test set. amc is valuable for other telescopes beyond TRACS; we developed the initial software for identifying papers that featured scientific results by NASA missions. Additionally, we investigate how amc can also be used to interrogate historical datasets and surface potential label errors. Our work demonstrates that LLM-based applications offer powerful and scalable assistance for library sciences.
Authors: Gabrijel Boduljak, Yushi Lan, Christian Rupprecht, Andrea Vedaldi
Abstract: Forecasting from partial observations is central to world modeling. Many recent methods represent the world through images, and reduce forecasting to stochastic video generation. Although such methods excel at realism and visual fidelity, predicting pixels is computationally intensive and not directly useful in many applications, as it requires translating RGB into signals useful for decision making. An alternative approach uses features from vision foundation models (VFMs) as world representations, performing deterministic regression to predict future world states. These features can be directly translated into actionable signals such as semantic segmentation and depth, while remaining computationally efficient. However, deterministic regression averages over multiple plausible futures, undermining forecast accuracy by failing to capture uncertainty. To address this crucial limitation, we introduce a generative forecaster that performs autoregressive flow matching in VFM feature space. Our key insight is that generative modeling in this space requires encoding VFM features into a compact latent space suitable for diffusion. We show that this latent space preserves information more effectively than previously used PCA-based alternatives, both for forecasting and other applications, such as image generation. Our latent predictions can be easily decoded into multiple useful and interpretable output modalities: semantic segmentation, depth, surface normals, and even RGB. With matched architecture and compute, our method produces sharper and more accurate predictions than regression across all modalities. Our results suggest that stochastic conditional generation of VFM features offers a promising and scalable foundation for future world models.
Authors: Iftekharul Islam, Weizi Li
Abstract: Effective mixed traffic control requires balancing efficiency, fairness, and safety. Existing approaches excel at optimizing efficiency and enforcing safety constraints but lack mechanisms to ensure equitable service, resulting in systematic starvation of vehicles on low-demand approaches. We propose a hierarchical framework combining multi-objective reinforcement learning for local intersection control with strategic routing for network-level coordination. Our approach introduces a Conflict Threat Vector that provides agents with explicit risk signals for proactive conflict avoidance, and a queue parity penalty that ensures equitable service across all traffic streams. Extensive experiments on a real-world network across different robot vehicle (RV) penetration rates demonstrate substantial improvements: up to 53% reductions in average wait time, up to 86% reductions in maximum starvation, and up to 86\% reduction in conflict rate compared to baselines, while maintaining fuel efficiency. Our analysis reveals that strategic routing effectiveness scales with RV penetration, becoming increasingly valuable at higher autonomy levels. The results demonstrate that multi-objective optimization through well-curated reward functions paired with strategic RV routing yields significant benefits in fairness and safety metrics critical for equitable mixed-autonomy deployment.
Authors: Qi Deng, Yuxuan Linghu, Zhiyuan Liu
Abstract: Multi-period portfolio optimization is important for real portfolio management, as it accounts for transaction costs, path-dependent risks, and the intertemporal structure of trading decisions that single-period models cannot capture. Classical methods usually follow a two-stage framework: machine learning algorithms are employed to produce forecasts that closely fit the realized returns, and the predicted values are then used in a downstream portfolio optimization problem to determine the asset weights. This separation leads to a fundamental misalignment between predictions and decision outcomes, while also ignoring the impact of transaction costs. To bridge this gap, recent studies have proposed the idea of end-to-end learning, integrating the two stages into a single pipeline. This paper introduces IPMO (Integrated Prediction and Multi-period Portfolio Optimization), a model for multi-period mean-variance portfolio optimization with turnover penalties. The predictor generates multi-period return forecasts that parameterize a differentiable convex optimization layer, which in turn drives learning via portfolio performance. For scalability, we introduce a mirror-descent fixed-point (MDFP) differentiation scheme that avoids factorizing the Karush-Kuhn-Tucker (KKT) systems, which thus yields stable implicit gradients and nearly scale-insensitive runtime as the decision horizon grows. In experiments with real market data and two representative time-series prediction models, the IPMO method consistently outperforms the two-stage benchmarks in risk-adjusted performance net of transaction costs and achieves more coherent allocation paths. Our results show that integrating machine learning prediction with optimization in the multi-period setting improves financial outcomes and remains computationally tractable.
Authors: Mrinal Rawat, Arkajyoti Chakraborty, Neha Gupta, Roberto Pieraccini
Abstract: Supervised fine-tuning (SFT) has emerged as one of the most effective ways to improve the performance of large language models (LLMs) in downstream tasks. However, SFT can have difficulty generalizing when the underlying data distribution changes, even when the new data does not fall completely outside the training domain. Recent reasoning-focused models such as o1 and R1 have demonstrated consistent gains over their non-reasoning counterparts, highlighting the importance of reasoning for improved generalization and reliability. However, collecting high-quality reasoning traces for SFT remains challenging -- annotations are costly, subjective, and difficult to scale. To address this limitation, we leverage Reinforcement Learning (RL) to enable models to learn reasoning strategies directly from task outcomes. We propose a pipeline in which LLMs generate reasoning steps that guide both the invocation of tools (e.g., function calls) and the final answer generation for conversational agents. Our method employs Group Relative Policy Optimization (GRPO) with rewards designed around tool accuracy and answer correctness, allowing the model to iteratively refine its reasoning and actions. Experimental results demonstrate that our approach improves both the quality of reasoning and the precision of tool invocations, achieving a 1.5% relative improvement over the SFT model (trained without explicit thinking) and a 40% gain compared to the base of the vanilla Qwen3-1.7B model. These findings demonstrate the promise of unifying reasoning and action learning through RL to build more capable and generalizable conversational agents.
Authors: John Tanner, Nicholas Davies, Pascal Elahi, Casey R. Myers, Du Huynh, Wei Liu, Mark Reynolds, Jingbo Wang
Abstract: Illegal, unreported, and unregulated (IUU) fishing causes global economic losses of \$10-25 billion annually and undermines marine sustainability and governance. Synthetic Aperture Radar (SAR) provides reliable maritime surveillance under all weather and lighting conditions, but classifying small maritime objects in SAR imagery remains challenging. We investigate quantum machine learning for this task, focusing on Quantum Kernel Methods (QKMs) applied to real and complex SAR chips extracted from the SARFish dataset. We tackle two binary classification problems, the first for distinguishing vessels from non-vessels, and the second for distinguishing fishing vessels from other types of vessels. We compare QKMs applied to real and complex SAR chips against classical Laplacian, RBF, and linear kernels applied to real SAR chips. Using noiseless numerical simulations of the quantum kernels, we find that QKMs are capable of obtaining equal or better performance than the classical kernel on these tasks in the best case, but do not demonstrate a clear advantage for the complex SAR data. This work presents the first application of QKMs to maritime classification in SAR imagery and offers insight into the potential and current limitations of quantum-enhanced learning for maritime surveillance.
Authors: Arnold Brosch, Abdelrahman Eldesokey, Michael Felsberg, Kira Maag
Abstract: Deep neural networks achieve superior performance in semantic segmentation, but are limited to a predefined set of classes, which leads to failures when they encounter unknown objects in open-world scenarios. Recognizing and segmenting these out-of-distribution (OOD) objects is crucial for safety-critical applications such as automated driving. In this work, we present an evidence segmentation framework using a Wasserstein loss, which captures distributional distances while respecting the probability simplex geometry. Combined with Kullback-Leibler regularization and Dice structural consistency terms, our approach leads to improved OOD segmentation performance compared to uncertainty-based approaches.
Authors: Kwun Sy Lee, Jiawei Chen, Fuk Sheng Ford Chung, Tianyu Zhao, Zhenyuan Chen, Debby D. Wang
Abstract: Reliable in silico molecular toxicity prediction is a cornerstone of modern drug discovery, offering a scalable alternative to experimental screening. However, the black-box nature of state-of-the-art models remains a significant barrier to adoption, as high-stakes safety decisions demand verifiable structural insights alongside predictive performance. To address this, we propose a novel multi-task learning (MTL) framework designed to jointly enhance accuracy and interpretability. Our architecture integrates a shared chemical language model with task-specific attention modules. By imposing an L1 sparsity penalty on these modules, the framework is constrained to focus on a minimal set of salient molecular fragments for each distinct toxicity endpoint. The resulting framework is trained end-to-end and is readily adaptable to various transformer-based backbones. Evaluated on the ClinTox, SIDER, and Tox21 benchmark datasets, our approach consistently outperforms both single-task and standard MTL baselines. Crucially, the sparse attention weights provide chemically intuitive visualizations that reveal the specific fragments influencing predictions, thereby enhancing insight into the model's decision-making process.
Authors: Marco Baiesi, Alberto Rosso
Abstract: We show that nonequilibrium dynamics can play a constructive role in unsupervised machine learning by inducing the spontaneous emergence of latent-state cycles. We introduce a model in which visible and hidden variables interact through two independently parametrized transition matrices, defining a Markov chain whose steady state is intrinsically out of equilibrium. Likelihood maximization drives this system toward nonequilibrium steady states with finite entropy production, reduced self-transition probabilities, and persistent probability currents in the latent space. These cycles are not imposed by the architecture but arise from training, and models that develop them avoid the low-log-likelihood regime associated with nearly reversible dynamics while more faithfully reproducing the empirical distribution of data classes. Compared with equilibrium approaches such as restricted Boltzmann machines, our model breaks the detailed balance between the forward and backward conditional transitions and relies on a log-likelihood gradient that depends explicitly on the last two steps of the Markov chain. Hence, this exploration of the interface between nonequilibrium statistical physics and modern machine learning suggests that introducing irreversibility into latent-variable models can enhance generative performance.
Authors: Han Lin, Xichen Pan, Ziqi Huang, Ji Hou, Jialiang Wang, Weifeng Chen, Zecheng He, Felix Juefei-Xu, Junzhe Sun, Zhipeng Fan, Ali Thabet, Mohit Bansal, Chu Wang
Abstract: Multimodal learning has rapidly advanced visual understanding, largely via multimodal large language models (MLLMs) that use powerful LLMs as cognitive cores. In visual generation, however, these powerful core models are typically reduced to global text encoders for diffusion models, leaving most of their reasoning and planning ability unused. This creates a gap: current multimodal LLMs can parse complex layouts, attributes, and knowledge-intensive scenes, yet struggle to generate images or videos with equally precise and structured control. We propose MetaCanvas, a lightweight framework that lets MLLMs reason and plan directly in spatial and spatiotemporal latent spaces and interface tightly with diffusion generators. We empirically implement MetaCanvas on three different diffusion backbones and evaluate it across six tasks, including text-to-image generation, text/image-to-video generation, image/video editing, and in-context video generation, each requiring precise layouts, robust attribute binding, and reasoning-intensive control. MetaCanvas consistently outperforms global-conditioning baselines, suggesting that treating MLLMs as latent-space planners is a promising direction for narrowing the gap between multimodal understanding and generation.
Authors: Mohamed Abdelsamad, Michael Ulrich, Bin Yang, Miao Zhang, Yakov Miron, Abhinav Valada
Abstract: Recent advances in self-supervised learning (SSL) have shown tremendous potential for learning 3D point cloud representations without human annotations. However, SSL for 3D point clouds still faces critical challenges due to irregular geometry, shortcut-prone reconstruction, and unbalanced semantics distribution. In this work, we propose DOS (Distilling Observable Softmaps), a novel SSL framework that self-distills semantic relevance softmaps only at observable (unmasked) points. This strategy prevents information leakage from masked regions and provides richer supervision than discrete token-to-prototype assignments. To address the challenge of unbalanced semantics in an unsupervised setting, we introduce Zipfian prototypes and incorporate them using a modified Sinkhorn-Knopp algorithm, Zipf-Sinkhorn, which enforces a power-law prior over prototype usage and modulates the sharpness of the target softmap during training. DOS outperforms current state-of-the-art methods on semantic segmentation and 3D object detection across multiple benchmarks, including nuScenes, Waymo, SemanticKITTI, ScanNet, and ScanNet200, without relying on extra data or annotations. Our results demonstrate that observable-point softmaps distillation offers a scalable and effective paradigm for learning robust 3D representations.
Authors: Rafa{\l} Potempa, Micha{\l} Kordasz, Sundas Naqeeb Khan, Krzysztof Werner, Kamil Wereszczy\'nski, Krzysztof Simi\'nski, Krzysztof A. Cyran
Abstract: This study aims to introduce the FRQI Pairs method to a wider audience, a novel approach to image classification using Quantum Recurrent Neural Networks (QRNN) with Flexible Representation for Quantum Images (FRQI). The study highlights an innovative approach to use quantum encoded data for an image classification task, suggesting that such quantum-based approaches could significantly reduce the complexity of quantum algorithms. Comparison of the FRQI Pairs method with contemporary techniques underscores the promise of integrating quantum computing principles with neural network architectures for the development of quantum machine learning.
Authors: Priyam Basu, Yunfeng Zhang, Vipul Raheja
Abstract: AI-generated text detectors have recently gained adoption in educational and professional contexts. Prior research has uncovered isolated cases of bias, particularly against English Language Learners (ELLs) however, there is a lack of systematic evaluation of such systems across broader sociolinguistic factors. In this work, we propose BAID, a comprehensive evaluation framework for AI detectors across various types of biases. As a part of the framework, we introduce over 200k samples spanning 7 major categories: demographics, age, educational grade level, dialect, formality, political leaning, and topic. We also generated synthetic versions of each sample with carefully crafted prompts to preserve the original content while reflecting subgroup-specific writing styles. Using this, we evaluate four open-source state-of-the-art AI text detectors and find consistent disparities in detection performance, particularly low recall rates for texts from underrepresented groups. Our contributions provide a scalable, transparent approach for auditing AI detectors and emphasize the need for bias-aware evaluation before these tools are deployed for public use.
Authors: Ekaterina Kalinicheva, Florian Helen, St\'ephane Mermoz, Florian Mouret, Milena Planells
Abstract: Fine-scale forest monitoring is essential for understanding canopy structure and its dynamics, which are key indicators of carbon stocks, biodiversity, and forest health. Deep learning is particularly effective for this task, as it integrates spectral, temporal, and spatial signals that jointly reflect the canopy structure. To address this need, we introduce THREASURE-Net, a novel end-to-end framework for Tree Height Regression And Super-Resolution. The model is trained on Sentinel-2 time series using reference height metrics derived from LiDAR HD data at multiple spatial resolutions over Metropolitan France to produce annual height maps. We evaluate three model variants, producing tree-height predictions at 2.5 m, 5 m, and 10 m resolution. THREASURE-Net does not rely on any pretrained model nor on reference very high resolution optical imagery to train its super-resolution module; instead, it learns solely from LiDAR-derived height information. Our approach outperforms existing state-of-the-art methods based on Sentinel data and is competitive with methods based on very high resolution imagery. It can be deployed to generate high-precision annual canopy-height maps, achieving mean absolute errors of 2.62 m, 2.72 m, and 2.88 m at 2.5 m, 5 m, and 10 m resolution, respectively. These results highlight the potential of THREASURE-Net for scalable and cost-effective structural monitoring of temperate forests using only freely available satellite data. The source code for THREASURE-Net is available at: https://github.com/Global-Earth-Observation/threasure-net.
URLs: https://github.com/Global-Earth-Observation/threasure-net.
Authors: Paulius Rauba, Qiyao Wei, Mihaela van der Schaar
Abstract: We consider the problem of auditing black-box large language models (LLMs) to ensure they behave reliably when deployed in production settings, particularly in high-stakes domains such as legal, medical, and regulatory compliance. Existing approaches for LLM auditing often focus on isolated aspects of model behavior, such as detecting specific biases or evaluating fairness. We are interested in a more general question -- can we understand how the outputs of black-box LLMs depend on each input token? There is a critical need to have such tools in real-world applications that rely on inaccessible API endpoints to language models. However, this is a highly non-trivial problem, as LLMs are stochastic functions (i.e. two outputs will be different by chance), while computing prompt-level gradients to approximate input sensitivity is infeasible. To address this, we propose Distribution-Based Sensitivity Analysis (DBSA), a lightweight model-agnostic procedure to evaluate the sensitivity of the output of a language model for each input token, without making any distributional assumptions about the LLM. DBSA is developed as a practical tool for practitioners, enabling quick, plug-and-play visual exploration of LLMs reliance on specific input tokens. Through illustrative examples, we demonstrate how DBSA can enable users to inspect LLM inputs and find sensitivities that may be overlooked by existing LLM interpretability methods.
Authors: Fabian Fuchs, Mario Ruben Fernandez, Norman Ettrich, Janis Keuper
Abstract: Seismic processing transforms raw data into subsurface images essential for geophysical applications. Traditional methods face challenges, such as noisy data, and manual parameter tuning, among others. Recently deep learning approaches have proposed alternative solutions to some of these problems. However, important challenges of existing deep learning approaches are spatially inconsistent results across neighboring seismic gathers and lack of user-control. We address these limitations by introducing ContextSeisNet, an in-context learning model, to seismic demultiple processing. Our approach conditions predictions on a support set of spatially related example pairs: neighboring common-depth point gathers from the same seismic line and their corresponding labels. This allows the model to learn task-specific processing behavior at inference time by observing how similar gathers should be processed, without any retraining. This method provides both flexibility through user-defined examples and improved lateral consistency across seismic lines. On synthetic data, ContextSeisNet outperforms a U-Net baseline quantitatively and demonstrates enhanced spatial coherence between neighboring gathers. On field data, our model achieves superior lateral consistency compared to both traditional Radon demultiple and the U-Net baseline. Relative to the U-Net, ContextSeisNet also delivers improved near-offset performance and more complete multiple removal. Notably, ContextSeisNet achieves comparable field data performance despite being trained on 90% less data, demonstrating substantial data efficiency. These results establish ContextSeisNet as a practical approach for spatially consistent seismic demultiple with potential applicability to other seismic processing tasks.
Authors: Abdullah Tokmak, Thomas B. Sch\"on, Dominik Baumann
Abstract: Safe Bayesian optimization (BO) with Gaussian processes is an effective tool for tuning control policies in safety-critical real-world systems, specifically due to its sample efficiency and safety guarantees. However, most safe BO algorithms assume homoscedastic sub-Gaussian measurement noise, an assumption that does not hold in many relevant applications. In this article, we propose a straightforward yet rigorous approach for safe BO across noise models, including homoscedastic sub-Gaussian and heteroscedastic heavy-tailed distributions. We provide a high-probability bound on the measurement noise via the scenario approach, integrate these bounds into high probability confidence intervals, and prove safety and optimality for our proposed safe BO algorithm. We deploy our algorithm in synthetic examples and in tuning a controller for the Franka Emika manipulator in simulation.
Authors: Hyungrok Do, Yuyan Wang, Mengling Liu, Myeonggyun Lee
Abstract: Evaluating the health effects of complex environmental mixtures remains a central challenge in environmental health research. Existing approaches vary in their flexibility, interpretability, scalability, and support for diverse outcome types, often limiting their utility in real-world applications. To address these limitations, we propose a neural network-based partial-linear single-index (NeuralPLSI) modeling framework that bridges semiparametric regression modeling interpretability with the expressive power of deep learning. The NeuralPLSI model constructs an interpretable exposure index via a learnable projection and models its relationship with the outcome through a flexible neural network. The framework accommodates continuous, binary, and time-to-event outcomes, and supports inference through a bootstrap-based procedure that yields confidence intervals for key model parameters. We evaluated NeuralPLSI through simulation studies under a range of scenarios and applied it to data from the National Health and Nutrition Examination Survey (NHANES) to demonstrate its practical utility. Together, our contributions establish NeuralPLSI as a scalable, interpretable, and versatile modeling tool for mixture analysis. To promote adoption and reproducibility, we release a user-friendly open-source software package that implements the proposed methodology and supports downstream visualization and inference (\texttt{https://github.com/hyungrok-do/NeuralPLSI}).
Authors: Bj\"orn Deiseroth, Max Henning H\"oth, Kristian Kersting, Letitia Parcalabescu
Abstract: Retrieval-augmented generation (RAG) models rely on retrieved evidence to guide large language model (LLM) generators, yet current systems treat retrieval as a weak heuristic rather than verifiable evidence. As a result, LLMs answer without support, hallucinate under incomplete or misleading context, and rely on spurious evidence. We introduce a training framework that treats the entire RAG pipeline -- both the retriever and the generator -- as an interactive proof system via an adaptation of the Merlin-Arthur (M/A) protocol. Arthur (the generator LLM) trains on questions of unkown provenance: Merlin provides helpful evidence, while Morgana injects adversarial, misleading context. Both use a linear-time XAI method to identify and modify the evidence most influential to Arthur. Consequently, Arthur learns to (i) answer when the context support the answer, (ii) reject when evidence is insufficient, and (iii) rely on the specific context spans that truly ground the answer. We further introduce a rigorous evaluation framework to disentangle explanation fidelity from baseline predictive errors. This allows us to introduce and measure the Explained Information Fraction (EIF), which normalizes M/A certified mutual-information guarantees relative to model capacity and imperfect benchmarks. Across three RAG datasets and two model families of varying sizes, M/A-trained LLMs show improved groundedness, completeness, soundness, and reject behavior, as well as reduced hallucinations -- without needing manually annotated unanswerable questions. The retriever likewise improves recall and MRR through automatically generated M/A hard positives and negatives. Our results demonstrate that autonomous interactive-proof-style supervision provides a principled and practical path toward reliable RAG systems that treat retrieved documents not as suggestions, but as verifiable evidence.
Authors: Tim Cofala, Christian Kalfar, Jingge Xiao, Johanna Schrader, Michelle Tang, Wolfgang Nejdl
Abstract: Therapeutic decision-making in clinical medicine constitutes a high-stakes domain in which AI guidance interacts with complex interactions among patient characteristics, disease processes, and pharmacological agents. Tasks such as drug recommendation, treatment planning, and adverse-effect prediction demand robust, multi-step reasoning grounded in reliable biomedical knowledge. Agentic AI methods, exemplified by TxAgent, address these challenges through iterative retrieval-augmented generation (RAG). TxAgent employs a fine-tuned Llama-3.1-8B model that dynamically generates and executes function calls to a unified biomedical tool suite (ToolUniverse), integrating FDA Drug API, OpenTargets, and Monarch resources to ensure access to current therapeutic information. In contrast to general-purpose RAG systems, medical applications impose stringent safety constraints, rendering the accuracy of both the reasoning trace and the sequence of tool invocations critical. These considerations motivate evaluation protocols treating token-level reasoning and tool-usage behaviors as explicit supervision signals. This work presents insights derived from our participation in the CURE-Bench NeurIPS 2025 Challenge, which benchmarks therapeutic-reasoning systems using metrics that assess correctness, tool utilization, and reasoning quality. We analyze how retrieval quality for function (tool) calls influences overall model performance and demonstrate performance gains achieved through improved tool-retrieval strategies. Our work was awarded the Excellence Award in Open Science. Complete information can be found at https://curebench.ai/.
URLs: https://curebench.ai/.
Authors: Rui Zhang, Han Wan, Yang Liu, Hao Sun
Abstract: Accurate modeling of spatiotemporal dynamics is crucial to understanding complex phenomena across science and engineering. However, this task faces a fundamental challenge when the governing equations are unknown and observational data are sparse. System stiffness, the coupling of multiple time-scales, further exacerbates this problem and hinders long-term prediction. Existing methods fall short: purely data-driven methods demand massive datasets, whereas physics-aware approaches are constrained by their reliance on known equations and fine-grained time steps. To overcome these limitations, we introduce an equation-free learning framework, namely, the Stable Spectral Neural Operator (SSNO), for modeling stiff partial differential equation (PDE) systems based on limited data. Instead of encoding specific equation terms, SSNO embeds spectrally inspired structures in its architecture, yielding strong inductive biases for learning the underlying physics. It automatically learns local and global spatial interactions in the frequency domain, while handling system stiffness with a robust integrating factor time-stepping scheme. Demonstrated across multiple 2D and 3D benchmarks in Cartesian and spherical geometries, SSNO achieves prediction errors one to two orders of magnitude lower than leading models. Crucially, it shows remarkable data efficiency, requiring only very few (2--5) training trajectories for robust generalization to out-of-distribution conditions. This work offers a robust and generalizable approach to learning stiff spatiotemporal dynamics from limited data without explicit \textit{a priori} knowledge of PDE terms.
Authors: Yuze He, Ferdi Kossmann, Srinivasan Seshan, Peter Steenkiste
Abstract: Recent advances in video analytics address real-time data drift by continuously retraining specialized, lightweight DNN models for individual cameras. However, the current practice of retraining a separate model for each camera suffers from high compute and communication costs, making it unscalable. We present ECCO, a new video analytics framework designed for resource-efficient continuous learning. The key insight is that the data drift, which necessitates model retraining, often shows temporal and spatial correlations across nearby cameras. By identifying cameras that experience similar drift and retraining a shared model for them, ECCO can substantially reduce the associated compute and communication costs. Specifically, ECCO introduces: (i) a lightweight grouping algorithm that dynamically forms and updates camera groups; (ii) a GPU allocator that dynamically assigns GPU resources across different groups to improve retraining accuracy and ensure fairness; and (iii) a transmission controller at each camera that configures frame sampling and coordinates bandwidth sharing with other cameras based on its assigned GPU resources. We conducted extensive evaluations on three distinctive datasets for two vision tasks. Compared to leading baselines, ECCO improves retraining accuracy by 6.7%-18.1% using the same compute and communication resources, or supports 3.3 times more concurrent cameras at the same accuracy.
Authors: Ernesto Casablanca, Oliver Sch\"on, Paolo Zuliani, Sadegh Soudjani
Abstract: Ensuring the safety of AI-enabled systems, particularly in high-stakes domains such as autonomous driving and healthcare, has become increasingly critical. Traditional formal verification tools fall short when faced with systems that embed both opaque, black-box AI components and complex stochastic dynamics. To address these challenges, we introduce LUCID (Learning-enabled Uncertainty-aware Certification of stochastIc Dynamical systems), a verification engine for certifying safety of black-box stochastic dynamical systems from a finite dataset of random state transitions. As such, LUCID is the first known tool capable of establishing quantified safety guarantees for such systems. Thanks to its modular architecture and extensive documentation, LUCID is designed for easy extensibility. LUCID employs a data-driven methodology rooted in control barrier certificates, which are learned directly from system transition data, to ensure formal safety guarantees. We use conditional mean embeddings to embed data into a reproducing kernel Hilbert space (RKHS), where an RKHS ambiguity set is constructed that can be inflated to robustify the result to out-of-distribution behavior. A key innovation within LUCID is its use of a finite Fourier kernel expansion to reformulate a semi-infinite non-convex optimization problem into a tractable linear program. The resulting spectral barrier allows us to leverage the fast Fourier transform to generate the relaxed problem efficiently, offering a scalable yet distributionally robust framework for verifying safety. LUCID thus offers a robust and efficient verification framework, able to handle the complexities of modern black-box systems while providing formal guarantees of safety. These unique capabilities are demonstrated on challenging benchmarks.
Authors: Felix Frohnert, Emiel Koridon, Stefano Polla
Abstract: We introduce an unsupervised machine-learning framework that discovers optimally compressed representations of quantum many-body ground states. Using an autoencoder neural network architecture on data from $L$-site Fermi-Hubbard models, we identify minimal latent spaces with a sharp reconstruction quality threshold at $L-1$ latent dimensions, matching the system's intrinsic degrees of freedom. We demonstrate the use of the trained decoder as a differentiable variational ansatz to minimize energy directly within the latent space. Crucially, this approach circumvents the $N$-representability problem, as the learned manifold implicitly restricts the optimization to physically valid quantum states.
Authors: Sacha Braun, David Holzm\"uller, Michael I. Jordan, Francis Bach
Abstract: Evaluating conditional coverage remains one of the most persistent challenges in assessing the reliability of predictive systems. Although conformal methods can give guarantees on marginal coverage, no method can guarantee to produce sets with correct conditional coverage, leaving practitioners without a clear way to interpret local deviations. To overcome sample-inefficiency and overfitting issues of existing metrics, we cast conditional coverage estimation as a classification problem. Conditional coverage is violated if and only if any classifier can achieve lower risk than the target coverage. Through the choice of a (proper) loss function, the resulting risk difference gives a conservative estimate of natural miscoverage measures such as L1 and L2 distance, and can even separate the effects of over- and under-coverage, and non-constant target coverages. We call the resulting family of metrics excess risk of the target coverage (ERT). We show experimentally that the use of modern classifiers provides much higher statistical power than simple classifiers underlying established metrics like CovGap. Additionally, we use our metric to benchmark different conformal prediction methods. Finally, we release an open-source package for ERT as well as previous conditional coverage metrics. Together, these contributions provide a new lens for understanding, diagnosing, and improving the conditional reliability of predictive systems.
Authors: Sotirios Nikoloutsopoulos, Iordanis Koutsopoulos, Michalis K. Titsias
Abstract: We propose a Stochastic Gradient Descent (SGD)-type algorithm for Personalized Federated Learning which can be particularly attractive for mobile energy-limited regimes due to its low per-client computational cost. The model to be trained includes a set of common weights for all clients, and a set of personalized weights that are specific to each client. At each optimization round, randomly selected clients perform multiple full gradient-descent updates over their client-specific weights towards optimizing the loss function on their own datasets, without updating the common weights. This procedure is energy-efficient since it has low computational cost per client. At the final update of each round, each client computes the joint gradient over both the client-specific and the common weights and returns the gradient of common weights to the server, which allows to perform an exact SGD step over the full set of weights in a distributed manner. For the overall optimization scheme, we rigorously prove convergence, even in non-convex settings such as those encountered when training neural networks, with a rate of $\mathcal{O} \left (\frac{1}{\sqrt{T}} \right )$ with respect to communication rounds $T$. In practice, PFLEGO exhibits substantially lower per-round wall-clock time, used as a proxy for energy. Our theoretical guarantees translate to superior performance in practice against baselines such as FedAvg and FedPer, as evaluated in several multi-class classification datasets, in particular, Omniglot, CIFAR-10, MNIST, Fashion-MNIST, and EMNIST.
Authors: Luis S\'anchez-Fern\'andez, Jes\'us A. Fisteus, Rafael L\'opez-Zaragoza
Abstract: We present a novel approach to the core set/instance selection problem in machine learning. Our approach is based on recent results on (proportional) representation in approval-based multi-winner elections. In our model, instances play a double role as voters and candidates. The approval set of each instance in the training set (acting as a voter) is defined from the concept of local set, which already exists in the literature. We then select the election winners by using a representative voting rule, and such winners are the data instances kept in the reduced training set. We evaluate our approach in two experiments involving neural network classifiers and classic machine learning classifiers (KNN and SVM). Our experiments show that, in several cases, our approach improves the performance of state-of-the-art methods, and the differences are statistically significant.
Authors: Zhihao Li, Zhilu Lai, Xiaobo Zhang, Wei Wang
Abstract: Solving high-dimensional partial differential equations (PDEs) efficiently requires handling multi-scale features across varying resolutions. To address this challenge, we present the Multiwavelet-based Multigrid Neural Operator (M2NO), a deep learning framework that integrates a multigrid structure with predefined multiwavelet spaces. M2NO leverages multi-resolution analysis to selectively transfer low-frequency error components to coarser grids while preserving high-frequency details at finer levels. This design enhances both accuracy and computational efficiency without introducing additional complexity. Moreover, M2NO serves as an effective preconditioner for iterative solvers, further accelerating convergence in large-scale PDE simulations. Through extensive evaluations on diverse PDE benchmarks, including high-resolution, super-resolution tasks, and preconditioning settings, M2NO consistently outperforms existing models. Its ability to efficiently capture fine-scale variations and large-scale structures makes it a robust and versatile solution for complex PDE simulations. Our code and datasets are available on https://github.com/lizhihao2022/M2NO.
Authors: Jiayu Li, Zilong Zhao, Kevin Yee, Uzair Javaid, Biplab Sikdar
Abstract: Synthetic tabular data generation has gained significant attention for its potential in data augmentation and privacy-preserving data sharing. While recent methods like diffusion and auto-regressive models (i.e., transformer) have advanced the field, generative adversarial networks (GANs) remain highly competitive due to their training efficiency and strong data generation capabilities. In this paper, we introduce Tabular Auto-Encoder Generative Adversarial Network (TAEGAN), a novel GAN-based framework that leverages a masked auto-encoder as the generator. TAEGAN is the first to incorporate self-supervised warmup training of generator into tabular GANs. It enhances GAN stability and exposes the generator to richer information beyond the discriminator's feedback. Additionally, we propose a novel sampling method tailored for imbalanced or skewed data and an improved loss function to better capture data distribution and correlations. We evaluate TAEGAN against seven state-of-the-art synthetic tabular data generation algorithms. Results from eight datasets show that TAEGAN outperforms all baselines on five datasets, achieving a 27% overall utility boost over the best-performing baseline while maintaining a model size less than 5% of the best-performing baseline model. Code is available at: https://github.com/BetterdataLabs/taegan.
Authors: Jingyang Qiao, Zhizhong Zhang, Xin Tan, Yanyun Qu, Shouhong Ding, Yuan Xie
Abstract: Continual Instruction Tuning (CIT) is adopted to continually instruct Large Models to follow human intent data by data. It is observed that existing gradient update would heavily destroy the performance on previous datasets during CIT process. Instead, Exponential Moving Average (EMA), owns the ability to trace previous parameters, which can aid in decreasing forgetting. Nonetheless, its stable balance weight fails to deal with the ever-changing datasets, leading to the out-of-balance between plasticity and stability. In this paper, we propose a general continual instruction tuning framework to address the challenge. Starting from the trade-off prerequisite and EMA update, we propose the plasticity and stability ideal condition. Based on Taylor expansion in the loss function, we find the optimal balance weight can be automatically determined by the gradients and learned parameters. Therefore, we propose a stable-plasticity balanced coefficient to avoid knowledge interference. Based on the semantic similarity of the instructions, we can determine whether to retrain or expand the training parameters and allocate the most suitable parameters for the testing instances. Extensive experiments across multiple continual instruction tuning benchmarks demonstrate that our approach not only enhances anti-forgetting capabilities but also significantly improves overall continual tuning performance. Our code is available at https://github.com/JingyangQiao/CoIN.
Authors: Shashank Hegde, Satyajeet Das, Gautam Salhotra, Gaurav S. Sukhatme
Abstract: With the increasing availability of open-source robotic data, imitation learning has become a promising approach for both manipulation and locomotion. Diffusion models are now widely used to train large, generalized policies that predict controls or trajectories, leveraging their ability to model multimodal action distributions. However, this generality comes at the cost of larger model sizes and slower inference, an acute limitation for robotic tasks requiring high control frequencies. Moreover, Diffusion Policy (DP), a popular trajectory-generation approach, suffers from a trade-off between performance and action horizon: fewer diffusion queries lead to larger trajectory chunks, which in turn accumulate tracking errors. To overcome these challenges, we introduce WARPD (World model Assisted Reactive Policy Diffusion), a method that generates closed-loop policies (weights for neural policies) directly, instead of open-loop trajectories. By learning behavioral distributions in parameter space rather than trajectory space, WARPD offers two major advantages: (1) extended action horizons with robustness to perturbations, while maintaining high task performance, and (2) significantly reduced inference costs. Empirically, WARPD outperforms DP in long-horizon and perturbed environments, and achieves multitask performance on par with DP while requiring only ~ 1/45th of the inference-time FLOPs per step.
Authors: Haochen Zhang, Junze Yin, Guanchu Wang, Zirui Liu, Lin F. Yang, Tianyi Zhang, Anshumali Shrivastava, Vladimir Braverman
Abstract: Low-rank optimization has emerged as a promising approach to enabling memory-efficient training of large language models (LLMs). Existing low-rank optimization methods typically project gradients onto a low-rank subspace, reducing the memory cost of storing optimizer states. A key challenge in these methods is selecting suitable subspaces to ensure an effective optimization trajectory. Most existing approaches select the dominant subspace to preserve gradient information, as this intuitively provides the best approximation. However, we find that in practice, the dominant subspace stops changing during pretraining, thereby constraining weight updates to similar subspaces. In this paper, we propose importance sampling for low-rank optimization in LLM pretraining with a provable convergence guarantee, which the dominant subspace approach does not have. Empirically, we demonstrate that our method significantly outperforms previous methods in LLM pretraining tasks.
Authors: Jinouwen Zhang, Junjie Ren, Qianhong Ma, Jianyu Wu, Aobo Yang, Yan Lu, Lu Chen, Hairun Xie, Jing Wang, Miao Zhang, Wanli Ouyang, Shixiang Tang
Abstract: Aircraft manufacturing is the jewel in the crown of industry, in which generating high-fidelity airfoil geometries with controllable and editable representations remains a fundamental challenge. Existing deep learning methods, which typically rely on predefined parametric representations (e.g., B\'ezier) or discrete point sets, face an inherent trade-off between expressive power and resolution adaptability. To tackle this challenge, we introduce FuncGenFoil, a novel function-space generative model that directly reconstructs airfoil geometries as function curves. Our method inherits the advantages of arbitrary-resolution sampling and smoothness from parametric functions, as well as the strong expressiveness of discrete point-based representations. Empirical evaluations demonstrate that FuncGenFoil improves upon state-of-the-art methods in airfoil generation, achieving a relative 74.4% reduction in label error and a 23.2% increase in diversity on the AF-200K dataset. Our results highlight the advantages of function-space modeling for aerodynamic shape optimization, offering a powerful and flexible framework for high-fidelity airfoil design.
Authors: Maxime Peyrard, Kyunghyun Cho
Abstract: Statistical inference, a central tool of science, revolves around the study and the usage of statistical estimators: functions that map finite samples to predictions about unknown distribution parameters. In the frequentist framework, estimators are evaluated based on properties such as bias, variance (for parameter estimation), accuracy, power, and calibration (for hypothesis testing). However, crafting estimators with desirable properties is often analytically challenging, and sometimes impossible, e.g., there exists no universally unbiased estimator for the standard deviation. In this work, we introduce meta-statistical learning, an amortized learning framework that recasts estimator design as an optimization problem via supervised learning. This takes a fully empirical approach to discovering statistical estimators; entire datasets are input to permutation-invariant neural networks, such as Set Transformers, trained to predict the target statistical property. The trained model is the estimator, and can be analyzed through the classical frequentist lens. We demonstrate the approach on two tasks: learning a normality test (classification) and estimating mutual information (regression), achieving strong results even with small models. Looking ahead, this paradigm opens a path to automate the discovery of generalizable and flexible statistical estimators.
Authors: A. Quadir, M. Tanveer
Abstract: Restricted kernel machines (RKMs) have considerably improved generalization in machine learning. Recent advancements explored various techniques within the RKM framework, integrating kernel functions with least squares support vector machines (LSSVM) to mirror the energy function of restricted Boltzmann machines (RBM), leading to enhanced performance. However, RKMs may face challenges in generalization when dealing with unevenly distributed or complexly clustered data. Additionally, as the dataset size increases, the computational burden of managing high-dimensional feature spaces can become substantial, potentially hindering performance in large-scale datasets. To address these challenges, we propose twin restricted kernel machine (TRKM). TRKM combines the benefits of twin models with the robustness of the RKM framework to enhance classification and regression tasks. By leveraging the Fenchel-Young inequality, we introduce a novel conjugate feature duality, allowing the formulation of classification and regression problems in terms of dual variables. This duality provides an upper bound to the objective function of the TRKM problem, resulting in a new methodology under the RKM framework. The model uses an energy function similar to that of RBM, incorporating both visible and hidden variables corresponding to both classes. Additionally, the kernel trick is employed to map data into a high-dimensional feature space, where the model identifies an optimal separating hyperplane using a regularized least squares approach. Experiments on UCI and KEEL datasets confirm TRKM's superiority over baselines, showcasing its robustness and efficiency in handling complex data. Furthermore, We implemented the TRKM model on the brain age dataset, demonstrating its efficacy in predicting brain age.
Authors: Qibang Liu, Weiheng Zhong, Hadi Meidani, Diab Abueidda, Seid Koric, Philippe Geubelle
Abstract: Machine-learning-based surrogate models offer significant computational efficiency and faster simulations compared to traditional numerical methods, especially for problems requiring repeated evaluations of partial differential equations. This work introduces the Geometry-Informed Neural Operator Transformer (GINOT), which integrates the transformer architecture with the neural operator framework to enable forward predictions on arbitrary geometries. GINOT employs a sampling and grouping strategy together with an attention mechanism to encode surface point clouds that are unordered, exhibit non-uniform point densities, and contain varying numbers of points for different geometries. The geometry information is seamlessly integrated with query points in the solution decoder through the attention mechanism. The performance of GINOT is validated on multiple challenging datasets, showcasing its high accuracy and strong generalization capabilities for complex and arbitrary 2D and 3D geometries.
Authors: Wenjing Xiao, Wenhao Song, Miaojiang Chen, Min Chen
Abstract: Intelligent fault-tolerant (FT) computing has recently demonstrated significant advantages in predicting and diagnosing faults proactively, thereby ensuring reliable service delivery. However, due to the heterogeneity of fault knowledge, dynamic workloads, and limited data support, existing deep learning-based FT algorithms face challenges in fault detection quality and training efficiency. This is primarily because their homogenization of fault knowledge perception difficuties to fully capture diverse and complex fault patterns. To address these challenges, we propose FT-MoE, a sustainable-learning fault-tolerant computing framework based on a dual-path architecture for high-accuracy fault detection and classification. This model employs a mixture-of-experts (MoE) architecture, enabling different parameters to learn distinct fault knowledge. Additionally, we adopt a two-stage learning scheme that combines comprehensive offline training with continual online tuning, allowing the model to adaptively optimize its parameters in response to evolving real-time workloads. To facilitate realistic evaluation, we construct a new fault detection and classification dataset for edge networks, comprising 10,000 intervals with fine-grained resource features, surpassing existing datasets in both scale and granularity. Finally, we conduct extensive experiments on the FT benchmark to verify the effectiveness of FT-MoE. Results demonstrate that our model outperforms state-of-the-art methods.
Authors: Huanyu Liu, Ge Li, Jia Li, Hao Zhu, Kechi Zhang, Yihong Dong
Abstract: How to design reinforcement learning (RL) tasks that effectively unleash the reasoning capability of large language models (LLMs) remains an open question. Existing RL tasks (e.g., math, programming, and constructing reasoning tasks) suffer from three key limitations: (1) Scalability. They rely heavily on human annotation or expensive LLM synthesis to generate sufficient training data. (2) Verifiability. LLMs' outputs are hard to verify automatically and reliably. (3) Controllable Difficulty. Most tasks lack fine-grained difficulty control, making it hard to train LLMs to develop reasoning ability from easy to hard. To address these limitations, we propose Saturn, a SAT-based RL framework that uses Boolean Satisfiability (SAT) problems to train and evaluate LLMs reasoning. Saturn enables scalable task construction, rule-based verification, and precise difficulty control. Saturn designs a curriculum learning pipeline that continuously improves LLMs' reasoning capability by constructing SAT tasks of increasing difficulty and training LLMs from easy to hard. To ensure stable training, we design a principled mechanism to control difficulty transitions. We introduce Saturn-2.6k, a dataset of 2,660 SAT problems with varying difficulty. It supports the evaluation of how LLM reasoning changes with problem difficulty. We apply Saturn to DeepSeek-R1-Distill-Qwen and obtain Saturn-1.5B and Saturn-7B. We achieve several notable results: (1) On SAT problems, Saturn-1.5B and Saturn-7B achieve average pass@3 improvements of +14.0 and +28.1, respectively. (2) On math and programming tasks, Saturn-1.5B and Saturn-7B improve average scores by +4.9 and +1.8 on benchmarks (e.g., AIME, LiveCodeBench). (3) Compared to the state-of-the-art (SOTA) approach in constructing RL tasks, Saturn achieves further improvements of +8.8%. We release the source code, data, and models to support future research.
Authors: Binghang Lu, Changhong Mou, Guang Lin
Abstract: Physics-informed neural networks (PINNs) have emerged as a powerful tool for solving forward and inverse problems involving partial differential equations (PDEs) by incorporating physical laws into the training process. However, the performance of PINNs is often hindered in real-world scenarios involving noisy observational data and missing physics, particularly in inverse problems. In this work, we propose an iterative multi-objective PINN ensemble Kalman filter (iPINNER) framework that improves the robustness and accuracy of PINNs in both forward and inverse problems by using the \textit{ensemble Kalman filter} and the \textit{non-dominated sorting genetic algorithm} III (NSGA-III). Specifically, NSGA-III is used as a multi-objective optimizer that can generate various ensemble members of PINNs along the optimal Pareto front, while accounting the model uncertainty in the solution space. These ensemble members are then utilized within the EnKF to assimilate noisy observational data. The EnKF's analysis is subsequently used to refine the data loss component for retraining the PINNs, thereby iteratively updating their parameters. The iterative procedure generates improved solutions to the PDEs. The proposed method is tested on two benchmark problems: the one-dimensional viscous Burgers equation and the time-fractional mixed diffusion-wave equation (TFMDWE). The numerical results show it outperforms standard PINNs in handling noisy data and missing physics.
Authors: Annabelle Sujun Tang, Christopher Priebe, Rohan Mahapatra, Lianhui Qin, Hadi Esmaeilzadeh
Abstract: While model serving has unlocked unprecedented capabilities, the high cost of serving large-scale models continues to be a significant barrier to widespread accessibility and rapid innovation. Compiler optimizations have long driven substantial performance improvements, but existing compilers struggle with neural workloads due to the exponentially large and highly interdependent space of possible transformations. Although existing stochastic search techniques can be effective, they are often sample-inefficient and fail to leverage the structural context underlying compilation decisions. We set out to investigate the research question of whether reasoning with large language models (LLMs), without any retraining, can leverage the context-aware decision space of compiler optimizations to significantly improve sample efficiency. To that end, we introduce a novel compilation framework (dubbed Reasoning Compiler) that formulates optimization as a sequential, context-aware decision process guided by a large language model and structured Monte Carlo tree search (MCTS). The LLM acts as a proposal mechanism, suggesting hardware-informed transformations that reflect the current program state and accumulated performance feedback. MCTS incorporates the LLM-generated proposals to balance exploration and exploitation, facilitating structured, context-sensitive traversal of the expansive compiler optimization space. By achieving substantial speedups with markedly fewer samples than leading neural compilers, our approach demonstrates the potential of LLM-guided reasoning to transform the landscape of compiler optimization.
Authors: Alexey Zakharov, Shimon Whiteson
Abstract: Natural language can offer a concise and human-interpretable means of specifying reinforcement learning (RL) tasks. The ability to extract rewards from a language instruction can enable the development of robotic systems that can learn from human guidance; however, it remains a challenging problem, especially in visual environments. Existing approaches that employ large, pretrained language models either rely on non-visual environment representations, require prohibitively large amounts of feedback, or generate noisy, ill-shaped reward functions. In this paper, we propose a novel method, GoalLadder, that leverages vision-language models (VLMs) to train RL agents from a single language instruction in visual environments. GoalLadder works by incrementally discovering states that bring the agent closer to completing a task specified in natural language. To do so, it queries a VLM to identify states that represent an improvement in agent's task progress and to rank them using pairwise comparisons. Unlike prior work, GoalLadder does not trust VLM's feedback completely; instead, it uses it to rank potential goal states using an ELO-based rating system, thus reducing the detrimental effects of noisy VLM feedback. Over the course of training, the agent is tasked with minimising the distance to the top-ranked goal in a learned embedding space, which is trained on unlabelled visual data. This key feature allows us to bypass the need for abundant and accurate feedback typically required to train a well-shaped reward function. We demonstrate that GoalLadder outperforms existing related methods on classic control and robotic manipulation environments with the average final success rate of $\sim$95% compared to only $\sim$45% of the best competitor.
Authors: Jun Ohkubo
Abstract: It is sometimes difficult to achieve a complete observation for a full set of observables, and partial observations are necessary. For deterministic systems, the Mori-Zwanzig formalism provides a theoretical framework for handling partial observations. Recently, data-driven algorithms based on the Koopman operator theory have made significant progress, and there is a discussion to connect the Mori-Zwanzig formalism with the Koopman operator theory. In this work, we discuss the effects of partial observation in stochastic systems using the Koopman operator theory. The discussion clarifies the importance of distinguishing the state space and the function space in stochastic systems. Even in stochastic systems, the delay-embedding technique is beneficial for partial observation, and several numerical experiments show a power-law behavior of error with respect to the amplitude of the additive noise. We also discuss the relation between the exponent of the power-law behavior and the effects of partial observation.
Authors: Jakub Pele\v{s}ka, Gustav \v{S}\'ir
Abstract: Relational databases (RDBs) are widely regarded as the gold standard for storing structured information. Consequently, predictive tasks leveraging this data format hold significant application promise. Recently, Relational Deep Learning (RDL) has emerged as a novel paradigm wherein RDBs are conceptualized as graph structures, enabling the application of various graph neural architectures to effectively address these tasks. However, given its novelty, there is a lack of analysis into the relationships between the performance of various RDL models and the characteristics of the underlying RDBs. In this study, we present REDELEX$-$a comprehensive exploration framework for evaluating RDL models of varying complexity on the most diverse collection of over 70 RDBs, which we make available to the community. Benchmarked alongside key representatives of classic methods, we confirm the generally superior performance of RDL while providing insights into the main factors shaping performance, including model complexity, database sizes and their structural properties.
Authors: Janghoon Ock, Radheesh Sharma Meda, Srivathsan Badrinarayanan, Neha S. Aluru, Achuth Chandrasekhar, Amir Barati Farimani
Abstract: We present a modular framework powered by large language models (LLMs) that automates and streamlines key tasks across the early-stage computational drug discovery pipeline. By combining LLM reasoning with domain-specific tools, the framework performs biomedical data retrieval, literature-grounded question answering via retrieval-augmented generation, molecular generation, multi-property prediction, property-aware molecular refinement, and 3D protein-ligand structure generation. The agent autonomously retrieved relevant biomolecular information, including FASTA sequences, SMILES representations, and literature, and answered mechanistic questions with improved contextual accuracy compared to standard LLMs. It then generated chemically diverse seed molecules and predicted 75 properties, including ADMET-related and general physicochemical descriptors, which guided iterative molecular refinement. Across two refinement rounds, the number of molecules with QED > 0.6 increased from 34 to 55. The number of molecules satisfying empirical drug-likeness filters also rose; for example, compliance with the Ghose filter increased from 32 to 55 within a pool of 100 molecules. The framework also employed Boltz-2 to generate 3D protein-ligand complexes and provide rapid binding affinity estimates for candidate compounds. These results demonstrate that the approach effectively supports molecular screening, prioritization, and structure evaluation. Its modular design enables flexible integration of evolving tools and models, providing a scalable foundation for AI-assisted therapeutic discovery.
Authors: Alexander Xiong, Xuandong Zhao, Aneesh Pappu, Dawn Song
Abstract: Large Language Models (LLMs) have demonstrated remarkable capabilities across a wide range of tasks, yet they also exhibit memorization of their training data. This phenomenon raises critical questions about model behavior, privacy risks, and the boundary between learning and memorization. Addressing these concerns, this paper synthesizes recent studies and investigates the landscape of memorization, the factors influencing it, and methods for its detection and mitigation. We explore key drivers, including training data duplication, training dynamics, and fine-tuning procedures that influence data memorization. In addition, we examine methodologies such as prefix-based extraction, membership inference, and adversarial prompting, assessing their effectiveness in detecting and measuring memorized content. Beyond technical analysis, we also explore the broader implications of memorization, including the legal and ethical implications. Finally, we discuss mitigation strategies, including data cleaning, differential privacy, and post-training unlearning, while highlighting open challenges in balancing the need to minimize harmful memorization with model utility. This paper provides a comprehensive overview of the current state of research on LLM memorization across technical, privacy, and performance dimensions, identifying critical directions for future work.
Authors: Zhuang Qi, Ying-Peng Tang, Lei Meng, Han Yu, Xiaoxiao Li, Xiangxu Meng
Abstract: Federated Class Incremental Learning (FCIL) aims to collaboratively process continuously increasing incoming tasks across multiple clients. Among various approaches, data replay has become a promising solution, which can alleviate forgetting by reintroducing representative samples from previous tasks. However, their performance is typically limited by class imbalance, both within the replay buffer due to limited global awareness and between replayed and newly arrived classes. To address this issue, we propose a class wise balancing data replay method for FCIL (FedCBDR), which employs a global coordination mechanism for class-level memory construction and reweights the learning objective to alleviate the aforementioned imbalances. Specifically, FedCBDR has two key components: 1) the global-perspective data replay module reconstructs global representations of prior task in a privacy-preserving manner, which then guides a class-aware and importance-sensitive sampling strategy to achieve balanced replay; 2) Subsequently, to handle class imbalance across tasks, the task aware temperature scaling module adaptively adjusts the temperature of logits at both class and instance levels based on task dynamics, which reduces the model's overconfidence in majority classes while enhancing its sensitivity to minority classes. Experimental results verified that FedCBDR achieves balanced class-wise sampling under heterogeneous data distributions and improves generalization under task imbalance between earlier and recent tasks, yielding a 2%-15% Top-1 accuracy improvement over six state-of-the-art methods.
Authors: Hyo-Jeong Jang, Hye-Bin Shin, Seong-Whan Lee
Abstract: Electroencephalography (EEG) is a fundamental modality for cognitive state monitoring in brain-computer interfaces (BCIs). However, it is highly susceptible to intrinsic signal errors and human-induced labeling errors, which lead to label noise and ultimately degrade model performance. To enhance EEG learning, multimodal knowledge distillation (KD) has been explored to transfer knowledge from visual models with rich representations to EEG-based models. Nevertheless, KD faces two key challenges: modality gap and soft label misalignment. The former arises from the heterogeneous nature of EEG and visual feature spaces, while the latter stems from label inconsistencies that create discrepancies between ground truth labels and distillation targets. This paper addresses semantic uncertainty caused by ambiguous features and weakly defined labels. We propose a novel cross-modal knowledge distillation framework that mitigates both modality and label inconsistencies. It aligns feature semantics through a prototype-based similarity module and introduces a task-specific distillation head to resolve label-induced inconsistency in supervision. Experimental results demonstrate that our approach improves EEG-based emotion regression and classification performance, outperforming both unimodal and multimodal baselines on a public multimodal dataset. These findings highlight the potential of our framework for BCI applications.
Authors: Rodrigo Tertulino, Ricardo Almeida
Abstract: This study proposes and validates a Federated Learning (FL) framework to proactively identify at-risk students while preserving data privacy. Persistently high dropout rates in distance education remain a pressing institutional challenge. Using the large-scale OULAD dataset, we simulate a privacy-centric scenario where models are trained on early academic performance and digital engagement patterns. Our work investigates the practical trade-offs between model complexity (Logistic Regression vs. a Deep Neural Network) and the impact of local data balancing. The resulting federated model achieves strong predictive power (ROC AUC approximately 85%), demonstrating that FL is a practical and scalable solution for early-warning systems that inherently respects student data sovereignty.
Authors: Hugo Math, Rainer Lienhart
Abstract: Understanding causality in event sequences where outcome labels such as diseases or system failures arise from preceding events like symptoms or error codes is critical. Yet remains an unsolved challenge across domains like healthcare or vehicle diagnostics. We introduce CARGO, a scalable multi-label causal discovery method for sparse, high-dimensional event sequences comprising of thousands of unique event types. Using two pretrained causal Transformers as domain-specific foundation models for event sequences. CARGO infers in parallel, per sequence one-shot causal graphs and aggregates them using an adaptive frequency fusion to reconstruct the global Markov boundaries of labels. This two-stage approach enables efficient probabilistic reasoning at scale while bypassing the intractable cost of full-dataset conditional independence testing. Our results on a challenging real-world automotive fault prediction dataset with over 29,100 unique event types and 474 imbalanced labels demonstrate CARGO's ability to perform structured reasoning.
Authors: Hao Zhang, Zhenjia Li, Runfeng Bao, Yifan Gao, Xi Xiao, Heng Zhang, Shuyang Zhang, Bo Huang, Yuhang Wu, Tianyang Wang, Hao Xu
Abstract: Parameter-Efficient Fine-Tuning (PEFT), especially Low-Rank Adaptation (LoRA), has emerged as a promising approach to fine-tuning large language models(LLMs) while reducing computational and memory overhead. However, LoRA assumes a uniform rank \textit{r} for each incremental matrix, not accounting for the varying significance of weight matrices across different modules and layers. AdaLoRA leverages Singular Value Decomposition (SVD) to parameterize updates and employs pruning of singular values to introduce dynamic rank allocation, thereby enhancing adaptability. However, during the training process, it often encounters issues of slow convergence speed and high computational overhead. To address this issue, we propose HyperAdaLoRA, a novel framework that accelerates the convergence of AdaLoRA by leveraging a hypernetwork. Instead of directly optimizing the components of Singular Value Decomposition $(P, \Lambda, Q)$, HyperAdaLoRA employs a hypernetwork based on attention mechanisms to dynamically generate these parameters. By pruning the outputs of the hypernetwork that generates the singular values, dynamic rank allocation is achieved. Comprehensive experiments on various datasets and models demonstrate that our method achieves faster convergence without sacrificing performance. Additionally, further extension experiments on other LoRA-based approaches validate the broad applicability of our method.
Authors: Haoqiang Kang, Yizhe Zhang, Nikki Lijing Kuang, Nicklas Majamaki, Navdeep Jaitly, Yi-An Ma, Lianhui Qin
Abstract: Large Language Models (LLMs) demonstrate their reasoning ability through chain-of-thought (CoT) generation. However, LLM's autoregressive decoding may limit the ability to revisit and refine earlier tokens in a holistic manner, which can also lead to inefficient exploration for diverse solutions. In this paper, we propose LaDiR (Latent Diffusion Reasoner), a novel reasoning framework that unifies the expressiveness of continuous latent representation with the iterative refinement capabilities of latent diffusion models for an existing LLM. We first construct a structured latent reasoning space using a Variational Autoencoder (VAE) that encodes text reasoning steps into blocks of thought tokens, preserving semantic information and interpretability while offering compact but expressive representations. Subsequently, we utilize a latent diffusion model that learns to denoise a block of latent thought tokens with a blockwise bidirectional attention mask, enabling longer horizon and iterative refinement with adaptive test-time compute. This design allows efficient parallel generation of diverse reasoning trajectories, allowing the model to plan and revise the reasoning process holistically. We conduct evaluations on a suite of mathematical reasoning and planning benchmarks. Empirical results show that LaDiR consistently improves accuracy, diversity, and interpretability over existing autoregressive, diffusion-based, and latent reasoning methods, revealing a new paradigm for text reasoning with latent diffusion.
Authors: Jonas Kirch, Sebastian Becker, Tiago Koketsu Rodrigues, Stefan Harmeling
Abstract: Clustered Federated Learning has emerged as an effective approach for handling heterogeneous data across clients by partitioning them into clusters with similar or identical data distributions. However, most existing methods, including the Iterative Federated Clustering Algorithm (IFCA), rely on a central server to coordinate model updates, which creates a bottleneck and a single point of failure, limiting their applicability in more realistic decentralized learning settings. In this work, we introduce DFCA, a fully decentralized clustered FL algorithm that enables clients to collaboratively train cluster-specific models without central coordination. DFCA uses a sequential running average to aggregate models from neighbors as updates arrive, providing a communication-efficient alternative to batch aggregation while maintaining clustering performance. Our experiments on various datasets demonstrate that DFCA outperforms other decentralized algorithms and performs comparably to centralized IFCA, even under sparse connectivity, highlighting its robustness and practicality for dynamic real-world decentralized networks.
Authors: Qingyao Ai, Yichen Tang, Changyue Wang, Jianming Long, Weihang Su, Yiqun Liu
Abstract: Scaling up data, parameters, and test-time computation has been the mainstream methods to improve LLM systems (LLMsys), but their upper bounds are almost reached due to the gradual depletion of high-quality data and marginal gains obtained from larger computational resource consumption. Inspired by the abilities of human and traditional AI systems in learning from practice, constructing memory and continual learning frameworks for LLMsys has become an important and popular research direction in recent literature. Yet, existing benchmarks for LLM memory often focus on evaluating the system on homogeneous reading comprehension tasks with long-form inputs rather than testing their abilities to learn from accumulated user feedback in service time. Therefore, we propose a user feedback simulation framework and a comprehensive benchmark covering multiple domains, languages, and types of tasks to evaluate the continual learning abilities of LLMsys. Experiments show that the effectiveness and efficiency of state-of-the-art baselines are far from satisfying, and we hope this benchmark could pave the way for future studies on LLM memory and optimization algorithms.
Authors: Wanhao Yu, Zheng Wang, Shuteng Niu, Sen Lin, Li Yang
Abstract: Zeroth-order (ZO) optimization has gained attention as a memory-efficient alternative to first-order (FO) methods, particularly in settings where gradient computation is expensive or even impractical. Beyond its memory efficiency, in this work, we investigate ZO optimization for continual learning (CL) as a novel approach to address the plasticity-stability-efficiency trilemma. Through theoretical analysis and empirical evidence, we show that ZO optimization naturally leads to flatter loss landscapes, which in turn reduce forgetting in CL. However, this stability comes at a cost of plasticity: due to its imprecise gradient estimates and slower convergence, ZO optimization tends to be less effective than FO in acquiring new task-specific knowledge, particularly under constrained training budgets. To better understand this trade-off, we conduct a holistic evaluation of ZO optimization applied to various existing CL methods. Our findings reveal that ZO optimization enhances stability but often undermines plasticity, particularly when used with learnable classifiers. Motivated by this insight, we propose ZO-FC, a simple but effective approach that applies ZO optimization to a single adapter-based PEFT module with FO optimized classifier. This design leverages the stability benefits of ZO while preserving the adaptability of FO updates with negligible memory overhead. Experiments demonstrate that ZO-FC achieves an effective balance between stability and plasticity, offering a practical and memory-efficient solution for on-device CL.
Authors: Huatian Gong, Jiuh-Biing Sheu, Zheng Wang, Xiaoguang Yang, Ran Yan
Abstract: Post-disaster road assessment (PDRA) is essential for emergency response, enabling rapid evaluation of infrastructure conditions and efficient allocation of resources. Although drones provide a flexible and effective tool for PDRA, routing them in large-scale networks remains challenging. Exact and heuristic optimization methods scale poorly and demand domain expertise, while existing deep reinforcement learning (DRL) approaches adopt a single-task paradigm, requiring separate models for each problem variant and lacking adaptability to evolving operational needs. This study proposes a unified model (UM) for drone routing that simultaneously addresses eight PDRA variants. By training a single neural network across multiple problem configurations, UM captures shared structural knowledge while adapting to variant-specific constraints through a modern transformer encoder-decoder architecture. A lightweight adapter mechanism further enables efficient finetuning to unseen attributes without retraining, enhancing deployment flexibility in dynamic disaster scenarios. Extensive experiments demonstrate that the UM reduces training time and parameters by a factor of eight compared with training separate models, while consistently outperforming single-task DRL methods by 6-14%, heuristic algorithms by 22-42%, and commercial solvers by 24-82% in terms of solution quality (total collected information value). The model achieves rapid solutions (1-10 seconds) across networks of up to 1,000 nodes, with robustness confirmed through sensitivity analyses. Moreover, finetuning experiments show that unseen attributes can be effectively incorporated with minimal cost while retaining high solution quality. The source code for UM is publicly available at https://github.com/PJ-HTU/UM_PDRA.
Authors: Bhavya Vasudeva, Puneesh Deora, Yize Zhao, Vatsal Sharan, Christos Thrampoulidis
Abstract: The growing adoption of spectrum-aware matrix-valued optimizers such as Muon and Shampoo in deep learning motivates a systematic study of their generalization properties and, in particular, when they might outperform competitive algorithms. We approach this question by introducing appropriate simplifying abstractions as follows: First, we use imbalanced data as a testbed. Second, we study the canonical form of such optimizers, which is Spectral Gradient Descent (SpecGD) -- each update step is $UV^T$ where $U\Sigma V^T$ is the truncated SVD of the gradient. Third, within this framework we identify a canonical setting for which we precisely quantify when SpecGD outperforms vanilla Euclidean GD. For a Gaussian mixture data model and both linear and bilinear models, we show that unlike GD, which prioritizes learning dominant principal components of the data first, SpecGD learns all principal components of the data at equal rates. We demonstrate how this translates to a growing gap in class balanced loss favoring SpecGD early in training and further show that the gap remains consistent even when the GD counterpart uses adaptive step-sizes via normalization. By extending the analysis to deep linear models, we show that depth amplifies these effects. We empirically verify our theoretical findings on a variety of imbalanced datasets. Our experiments compare practical variants of spectral methods, like Muon and Shampoo, against their Euclidean counterparts and Adam. The results validate our findings that these spectral optimizers achieve superior generalization by promoting a more balanced learning of the data's underlying components.
Authors: Crystal Su, Kuai Yu, Jingrui Zhang, Mingyuan Shao, Daniel Bauer
Abstract: This work presents an ontology-integrated large language model (LLM) framework for chemical engineering that unites structured domain knowledge with generative reasoning. The proposed pipeline aligns model training and inference with the COPE ontology through a sequence of data acquisition, semantic preprocessing, information extraction, and ontology mapping steps, producing templated question-answer pairs that guide fine-tuning. A control-focused decoding stage and citation gate enforce syntactic and factual grounding by constraining outputs to ontology-linked terms, while evaluation metrics quantify both linguistic quality and ontological accuracy. Feedback and future extensions, including semantic retrieval and iterative validation, further enhance the system's interpretability and reliability. This integration of symbolic structure and neural generation provides a transparent, auditable approach for applying LLMs to process control, safety analysis, and other critical engineering contexts.
Authors: Crystal Su, Kuai Yu, Mingyuan Shao, Daniel Bauer
Abstract: Bulk tissue RNA sequencing of heterogeneous samples provides averaged gene expression profiles, obscuring cell type-specific dynamics. To address this, we present a probabilistic hierarchical Bayesian model that deconvolves bulk RNA-seq data into constituent cell-type expression profiles and proportions, leveraging a high-resolution single-cell reference. We apply our model to human endometrial tissue across the menstrual cycle, a context characterized by dramatic hormone-driven cellular composition changes. Our extended framework provides a principled inference of cell type proportions and cell-specific gene expression changes across cycle phases. We demonstrate the model's structure, priors, and inference strategy in detail, and we validate its performance with simulations and comparisons to existing methods. The results reveal dynamic shifts in epithelial, stromal, and immune cell fractions between menstrual phases, and identify cell-type-specific differential gene expression associated with endometrial function (e.g., decidualization markers in stromal cells during the secretory phase). We further conduct robustness tests and show that our Bayesian approach is resilient to reference mismatches and noise. Finally, we discuss the biological significance of our findings, potential clinical implications for fertility and endometrial disorders, and future directions, including integration of spatial transcriptomics.
Authors: Alex Dobra, Jakiw Pidstrigach, Tim Reichelt, Paolo Fraccaro, Anne Jones, Johannes Jakubik, Christian Schroeder de Witt, Philip Torr, Philip Stier
Abstract: Sensitivity analysis is a cornerstone of climate science, essential for understanding phenomena ranging from storm intensity to long-term climate feedbacks. However, computing these sensitivities using traditional physical models is often prohibitively expensive in terms of both computation and development time. While modern AI-based generative models are orders of magnitude faster to evaluate, computing sensitivities with them remains a significant bottleneck. This work addresses this challenge by applying the adjoint state method for calculating gradients in generative flow models. We apply this method to the cBottle generative model, trained on ERA5 and ICON data, to perform sensitivity analysis of any atmospheric variable with respect to sea surface temperatures. We quantitatively validate the computed sensitivities against the model's own outputs. Our results provide initial evidence that this approach can produce reliable gradients, reducing the computational cost of sensitivity analysis from weeks on a supercomputer with a physical model to hours on a GPU, thereby simplifying a critical workflow in climate science. The code can be found at https://github.com/Kwartzl8/cbottle_adjoint_sensitivity.
URLs: https://github.com/Kwartzl8/cbottle_adjoint_sensitivity.
Authors: Runyu Lu, Peng Zhang, Ruochuan Shi, Yuanheng Zhu, Dongbin Zhao, Yang Liu, Dong Wang, Cesare Alippi
Abstract: Equilibrium learning in adversarial games is an important topic widely examined in the fields of game theory and reinforcement learning (RL). Pursuit-evasion game (PEG), as an important class of real-world games from the fields of robotics and security, requires exponential time to be accurately solved. When the underlying graph structure varies, even the state-of-the-art RL methods require recomputation or at least fine-tuning, which can be time-consuming and impair real-time applicability. This paper proposes an Equilibrium Policy Generalization (EPG) framework to effectively learn a generalized policy with robust cross-graph zero-shot performance. In the context of PEGs, our framework is generally applicable to both pursuer and evader sides in both no-exit and multi-exit scenarios. These two generalizability properties, to our knowledge, are the first to appear in this domain. The core idea of the EPG framework is to train an RL policy across different graph structures against the equilibrium policy for each single graph. To construct an equilibrium oracle for single-graph policies, we present a dynamic programming (DP) algorithm that provably generates pure-strategy Nash equilibrium with near-optimal time complexity. To guarantee scalability with respect to pursuer number, we further extend DP and RL by designing a grouping mechanism and a sequence model for joint policy decomposition, respectively. Experimental results show that, using equilibrium guidance and a distance feature proposed for cross-graph PEG training, the EPG framework guarantees desirable zero-shot performance in various unseen real-world graphs. Besides, when trained under an equilibrium heuristic proposed for the graphs with exits, our generalized pursuer policy can even match the performance of the fine-tuned policies from the state-of-the-art PEG methods.
Authors: Kaiyuan Zhai, Jiacheng Cui, Zhehao Zhang, Junyu Xue, Yang Deng, Kui Wu, Guoming Tang
Abstract: Cross-domain HVAC energy prediction is essential for scalable building energy management, particularly because collecting extensive labeled data for every new building is both costly and impractical. Yet, this task remains highly challenging due to the scarcity and heterogeneity of data across different buildings, climate zones, and seasonal patterns. In particular, buildings situated in distinct climatic regions introduce variability that often leads existing methods to overfit to spurious correlations, rely heavily on expert intervention, or compromise on data diversity. To address these limitations, we propose CaberNet, a causal and interpretable deep sequence model that learns invariant (Markov blanket) representations for robust cross-domain prediction. In a purely data-driven fashion and without requiring any prior knowledge, CaberNet integrates i) a global feature gate trained with a self-supervised Bernoulli regularization to distinguish superior causal features from inferior ones, and ii) a domain-wise training scheme that balances domain contributions, minimizes cross-domain loss variance, and promotes latent factor independence. We evaluate CaberNet on real-world datasets collected from three buildings located in three climatically diverse cities, and it consistently outperforms all baselines, achieving a 22.9% reduction in normalized mean squared error (NMSE) compared to the best benchmark. Our code is available at https://github.com/SusCom-Lab/CaberNet-CRL.
Authors: Marco Roccetti
Abstract: The integration of advanced analytical tools, including Machine Learning (ML) and massive data processing, has revolutionized health research, promising unprecedented accuracy in diagnosis and risk prediction. However, the rigor of these complex methods is fundamentally dependent on the quality and integrity of the underlying datasets and the validity of their statistical design. We propose an emblematic case where advanced analysis (ML/Big Data) must necessarily be subsequent to the verification of basic methodological coherence and adherence to established medical protocols, such as the STROBE Statement. This study highlights a crucial cautionary principle: sophisticated analyses amplify, rather than correct, severe methodological flaws rooted in basic design choices, leading to misleading or contradictory findings. By applying simple, standard descriptive statistical methods and established national epidemiological benchmarks to a recently published cohort study on COVID-19 vaccine outcomes and severe adverse events, like cancer, we expose multiple, statistically irreconcilable paradoxes. These paradoxes, specifically the contradictory finding of an increased cancer incidence within an exposure subgroup, concurrent with a suppressed overall Crude Incidence Rate compared to national standards, definitively invalidate the reported risk of increased cancer in the total population. We demonstrate that the observed effects are mathematical artifacts stemming from an uncorrected selection bias in the cohort construction. This analysis serves as a robust reminder that even the most complex health studies must first pass the test of basic epidemiological consistency before any conclusion drawn from subsequent advanced statistical modeling can be considered valid or publishable.
Authors: Alexander W. Goodall, Edwin Hamel-De le Court, Francesco Belardinelli
Abstract: Many reinforcement learning algorithms, particularly those that rely on return estimates for policy improvement, can suffer from poor sample efficiency and training instability due to high-variance return estimates. In this paper we leverage new results from off-policy evaluation; it has recently been shown that well-designed behaviour policies can be used to collect off-policy data for provably lower variance return estimates. This result is surprising as it means collecting data on-policy is not variance optimal. We extend this key insight to the online reinforcement learning setting, where both policy evaluation and improvement are interleaved to learn optimal policies. Off-policy RL has been well studied (e.g., IMPALA), with correct and truncated importance weighted samples for de-biasing and managing variance appropriately. Generally these approaches are concerned with reconciling data collected from multiple workers in parallel, while the policy is updated asynchronously, mismatch between the workers and policy is corrected in a mathematically sound way. Here we consider only one worker - the behaviour policy, which is used to collect data for policy improvement, with provably lower variance return estimates. In our experiments we extend two policy-gradient methods with this regime, demonstrating better sample efficiency and performance over a diverse set of environments.
Authors: S Sairam, Sara Girdhar, Shivam Soni
Abstract: The R-Learner is a powerful, theoretically-grounded framework for estimating heterogeneous treatment effects, prized for its robustness to nuisance model errors. However, its application to network data, where causal heterogeneity is often graph-dependent, presents a critical challenge to its core assumption of a well-specified final-stage model. In this paper, we conduct a large-scale empirical study to systematically dissect the R-Learner framework on graphs. We provide the first rigorous evidence that the primary driver of performance is the inductive bias of the final-stage CATE estimator, an effect that dominates the choice of nuisance models. Our central finding is the quantification of a catastrophic "representation bottleneck": we prove with overwhelming statistical significance (p < 0.001) that R-Learners with a graph-blind final stage fail completely (MSE > 4.0), even when paired with powerful GNN nuisance models. Conversely, our proposed end-to-end Graph R-Learner succeeds and significantly outperforms a strong, non-DML GNN T-Learner baseline. Furthermore, we identify and provide a mechanistic explanation for a subtle, topology-dependent "nuisance bottleneck," linking it to GNN over-squashing via a targeted "Hub-Periphery Trade-off" analysis. Our findings are validated across diverse synthetic and semi-synthetic benchmarks. We release our code as a reproducible benchmark to facilitate future research on this critical "final-stage bottleneck."
Authors: Yuwen Zhang, Viet Tran, Paul Weng
Abstract: In clinical machine learning, the coexistence of multiple models with comparable performance -- a manifestation of the Rashomon Effect -- poses fundamental challenges for trustworthy deployment and evaluation. Small, imbalanced, and noisy datasets, coupled with high-dimensional and weakly identified clinical features, amplify this multiplicity and make conventional validation schemes unreliable. As a result, selecting among equally performing models becomes uncertain, particularly when resource constraints and operational priorities are not considered by conventional metrics like F1 score. To address these issues, we propose two complementary tools for robust model assessment and selection: Intervention Efficiency (IE) and the Perturbation Validation Framework (PVF). IE is a capacity-aware metric that quantifies how efficiently a model identifies actionable true positives when only limited interventions are feasible, thereby linking predictive performance with clinical utility. PVF introduces a structured approach to assess the stability of models under data perturbations, identifying models whose performance remains most invariant across noisy or shifted validation sets. Empirical results on synthetic and real-world healthcare datasets show that using these tools facilitates the selection of models that generalize more robustly and align with capacity constraints, offering a new direction for tackling the Rashomon Effect in clinical settings.
Authors: Lyu Yuhuan
Abstract: Verifying uniform conditions over continuous spaces through random sampling is fundamental in machine learning and control theory, yet classical coverage analyses often yield conservative bounds, particularly at small failure probabilities. We study uniform random sampling on the $d$-dimensional unit hypercube and analyze the number of uncovered subcubes after discretization. By applying a concentration inequality to the uncovered-count statistic, we derive a sample complexity bound with a logarithmic dependence on the failure probability ($\delta$), i.e., $M =O( \tilde{C}\ln(\frac{2\tilde{C}}{\delta}))$, which contrasts sharply with the classical linear $1/\delta$ dependence. Under standard Lipschitz and uniformity assumptions, we present a self-contained derivation and compare our result with classical coupon-collector rates. Numerical studies across dimensions, precision levels, and confidence targets indicate that our bound tracks practical coverage requirements more tightly and scales favorably as $\delta \to 0$. Our findings offer a sharper theoretical tool for algorithms that rely on grid-based coverage guarantees, enabling more efficient sampling, especially in high-confidence regimes.
Authors: Victor Rambaud, Salvador Mascarenhas, Yair Lakretz
Abstract: A cognitive map is an internal model which encodes the abstract relationships among entities in the world, giving humans and animals the flexibility to adapt to new situations, with a strong out-of-distribution (OOD) generalization that current AI systems still do not possess. To bridge this gap, we introduce MapFormers, new architectures based on Transformer models, which can learn cognitive maps from observational data and perform path integration in parallel, in a self-supervised manner. Cognitive maps are learned in the model by disentangling structural relationships in the inputs from their specific content, a property that can be achieved naturally by updating the positional encoding in Transformers with input-dependent matrices. We developed two variants of MapFormers that unify absolute and relative positional encoding to model episodic (EM) and working memory (WM), respectively. We tested MapFormers on several tasks, including a classic 2D navigation task, showing that our models can learn a cognitive map of the underlying space and generalize OOD (e.g., to longer sequences) with near-perfect performance, unlike current architectures. Together, these results demonstrate the superiority of models designed to learn a cognitive map, and the importance of introducing a structural bias for structure-content disentanglement, which can be achieved in Transformers with input-dependent positional encoding. MapFormers have broad applications in both neuroscience and AI, by explaining the neural mechanisms giving rise to cognitive maps, while allowing these relation models to be learned at scale.
Authors: Misha Padidar, Teresa Huang, Andrew Giuliani, Marina Spivak
Abstract: Stellarators are a prospective class of fusion-based power plants that confine a hot plasma with three-dimensional magnetic fields. Typically framed as a PDE-constrained optimization problem, stellarator design is a time-consuming process that can take hours to solve on a computing cluster. Developing fast methods for designing stellarators is crucial for advancing fusion research. Given the recent development of large datasets of optimized stellarators, machine learning approaches have emerged as a potential candidate. Motivated by this, we present an open inverse problem to the machine learning community: to rapidly generate high-quality stellarator designs which have a set of desirable characteristics. As a case study in the problem space, we train a conditional diffusion model on data from the QUASR database to generate quasisymmetric stellarator designs with desirable characteristics (aspect ratio and mean rotational transform). The diffusion model is applied to design stellarators with characteristics not seen during training. We provide evaluation protocols and show that many of the generated stellarators exhibit solid performance: less than 5% deviation from quasisymmetry and the target characteristics. The modest deviation from quasisymmetry highlights an opportunity to reach the sub 1% target. Beyond the case study, we share multiple promising avenues for generative modeling to advance stellarator design.
Authors: Haobin Ke, Sen Zhang, Qingqing Ye, Xun Ran, Haibo Hu
Abstract: Signed graphs with positive and negative edges can model complex relationships in social networks. Leveraging on balance theory that deduces edge signs from multi-hop node pairs, signed graph learning can generate node embeddings that preserve both structural and sign information. However, training on sensitive signed graphs raises significant privacy concerns, as model parameters may leak private link information. Existing protection methods with differential privacy (DP) typically rely on edge or gradient perturbation for unsigned graph protection. Yet, they are not well-suited for signed graphs, mainly because edge perturbation tends to cascading errors in edge sign inference under balance theory, while gradient perturbation increases sensitivity due to node interdependence and gradient polarity change caused by sign flips, resulting in larger noise injection. In this paper, motivated by the robustness of adversarial learning to noisy interactions, we present ASGL, a privacy-preserving adversarial signed graph learning method that preserves high utility while achieving node-level DP. We first decompose signed graphs into positive and negative subgraphs based on edge signs, and then design a gradient-perturbed adversarial module to approximate the true signed connectivity distribution. In particular, the gradient perturbation helps mitigate cascading errors, while the subgraph separation facilitates sensitivity reduction. Further, we devise a constrained breadth-first search tree strategy that fuses with balance theory to identify the edge signs between generated node pairs. This strategy also enables gradient decoupling, thereby effectively lowering gradient sensitivity. Extensive experiments on real-world datasets show that ASGL achieves favorable privacy-utility trade-offs across multiple downstream tasks.
Authors: Andrea Procopio, Marco Esposito, Sara Raggiunto, Andrey Gizdov, Alberto Belli, Paola Pierleoni
Abstract: We study on-device time-series analysis for gait detection in Parkinson's disease (PD) from short windows of triaxial acceleration, targeting resource-constrained wearables and edge nodes. We compare magnitude thresholding to three 1D CNNs for time-series analysis: a literature baseline (separable convolutions) and two ultra-light models - one purely separable and one with residual connections. Using the BioStampRC21 dataset, 2 s windows at 30 Hz, and subject-independent leave-one-subject-out (LOSO) validation on 16 PwPD with chest-worn IMUs, our residual separable model (Model 2, 533 params) attains PR-AUC = 94.5%, F1 = 91.2%, MCC = 89.4%, matching or surpassing the baseline (5,552 params; PR-AUC = 93.7%, F1 = 90.5%, MCC = 88.5%) with approximately 10x fewer parameters. The smallest model (Model 1, 305 params) reaches PR-AUC = 94.0%, F1 = 91.0%, MCC = 89.1%. Thresholding obtains high recall (89.0%) but low precision (76.5%), yielding many false positives and high inter-subject variance. Sensor-position analysis (train-on-all) shows chest and thighs are most reliable; forearms degrade precision/recall due to non-gait arm motion; naive fusion of all sites does not outperform the best single site. Both compact CNNs execute within tight memory/latency budgets on STM32-class MCUs (sub-10 ms on low-power boards), enabling on-sensor gating of transmission/storage. Overall, ultra-light separable CNNs provide a superior accuracy-efficiency-generalization trade-off to fixed thresholds for wearable PD gait detection and underscore the value of tailored time-series models for edge deployment.
Authors: Ningning Chen, Weicai Ye, Ying Jiang
Abstract: We introduce HBLLM, a wavelet-enhanced high-fidelity $1$-bit post-training quantization method for Large Language Models (LLMs). By leveraging Haar wavelet transforms to enhance expressive capacity through frequency decomposition, HBLLM significantly improves quantization fidelity while maintaining minimal overhead. This approach features two innovative structure-aware grouping strategies: (1) frequency-aware multi-parameter intra-row grouping and (2) $\ell_2$-norm-based saliency-driven column selection. For non-salient weights, a shared mean is employed across quantization groups within each frequency band to optimize storage efficiency. Experiments conducted on the OPT and LLaMA models demonstrate that HBLLM achieves state-of-the-art performance in $1$-bit quantization, attaining a perplexity of $6.71$ on LLaMA$2$-$13$B with an average weight storage of only $1.08$ bits. Code available at: https://github.com/Yeyke/HBLLM.
Authors: Songqiao Su, Xiaofei Sun, Xiaoya Li, Albert Wang, Jiwei Li, Chris Shum
Abstract: In this paper, we propose CUDA-L2, a system that combines large language models (LLMs) and reinforcement learning (RL) to automatically optimize Half-precision General Matrix Multiply (HGEMM) CUDA kernels. Using CUDA execution speed as the RL reward, CUDA-L2 automatically optimizes HGEMM kernels across 1,000 configurations. CUDA-L2 systematically outperforms major matmul baselines to date, from the widely-used torch.matmul to state-of-the-art Nvidia's closed-source libraries, i.e., cuBLAS, cuBLASLt. In offline mode, where kernels are executed consecutively without time intervals, CUDA-L2 yields +22.0% over torch.matmul on average; +19.2% over cuBLAS using the optimal layout configuration (normal-normal NN and transposed-normal TN); +16.8% over cuBLASLt-heuristic, which queries cuBLASLt library and selects the algorithm based on the heuristic's suggestion; and +11.4% over the most competitive cuBLASLt-AutoTuning model, which selects the fastest algorithm from up to 100 candidates from cuBLASLt's suggestions. In server mode, where kernels are executed at random intervals simulating real-time inference, the speedups further increase to +28.7%, +26.0%, +22.4%, and +15.9% for torch.matmul, cuBLAS, cuBLASLt-heuristic, and cuBLASLt-AutoTuning respectively. CUDA-L2 shows that even the most performance-critical, heavily-optimized kernels like HGEMM can be improved through LLM-guided RL automation by systematically exploring configuration spaces at scales impractical for humans. Project and code can be found at github.com/deepreinforce-ai/CUDA-L2
Authors: Julian Evan Chrisnanto, Nurfauzi Fadillah, Yulison Herry Chrisnanto
Abstract: Physics-Informed Neural Networks (PINNs) have emerged as a powerful, mesh-free paradigm for solving partial differential equations (PDEs). However, they notoriously struggle with stiff, multi-scale, and nonlinear systems due to the inherent spectral bias of standard multilayer perceptron (MLP) architectures, which prevents them from adequately representing high-frequency components. In this work, we introduce the Adaptive Spectral Physics-Enabled Network (ASPEN), a novel architecture designed to overcome this critical limitation. ASPEN integrates an adaptive spectral layer with learnable Fourier features directly into the network's input stage. This mechanism allows the model to dynamically tune its own spectral basis during training, enabling it to efficiently learn and represent the precise frequency content required by the solution. We demonstrate the efficacy of ASPEN by applying it to the complex Ginzburg-Landau equation (CGLE), a canonical and challenging benchmark for nonlinear, stiff spatio-temporal dynamics. Our results show that a standard PINN architecture catastrophically fails on this problem, diverging into non-physical oscillations. In contrast, ASPEN successfully solves the CGLE with exceptional accuracy. The predicted solution is visually indistinguishable from the high-resolution ground truth, achieving a low median physics residual of 5.10 x 10^-3. Furthermore, we validate that ASPEN's solution is not only pointwise accurate but also physically consistent, correctly capturing emergent physical properties, including the rapid free energy relaxation and the long-term stability of the domain wall front. This work demonstrates that by incorporating an adaptive spectral basis, our framework provides a robust and physically-consistent solver for complex dynamical systems where standard PINNs fail, opening new options for machine learning in challenging physical domains.
Authors: Siqi Wang, Hailong Yang, Junjie Zhu, Xuezhu Wang, Yufan Xu, Depei Qian
Abstract: Reinforcement Learning from Human Feedback (RLHF) is an important fine-tuning technique for large language models (LLMs) and comprises three stages: generation, inference, and training. The generation stage generates samples that are then used to infer learnable experiences for training. We observe that the generation stage is the bottleneck of the entire execution process and consider it a key point for optimization. Specifically, we realize the first attempt to integrate speculative decoding into the RLHF generation stage and propose RLHFSpec, an RLHF system that accelerates generation execution with efficient speculative decoding and sample reallocation. To fully exploit the performance potential provided by speculative decoding, especially dealing with the dynamic workload of the generation stage, RLHFSpec proposes a workload-aware drafting strategy selection mechanism, which selects the near-optimal strategy by jointly considering the verification cost and the number of accepted tokens. Moreover, RLHFSpec also proposes sample reallocation to fully utilize the GPU resources, and optimizes it with an efficient sample migration mechanism. The experimental results show that the RLHFSpec can achieve higher throughput in the generation stage compared to state-of-the-art works. Moreover, due to the effective alleviation of the generation bottleneck, RLHFSpec also shows significant performance speedup in the entire RLHF execution.
Authors: Haiyang Yu, Meng-Chieh Lee, Xiang song, Qi Zhu, Christos Faloutsos
Abstract: We explore the node classification task in the context of graph domain adaptation, which uses both source and target graph structures along with source labels to enhance the generalization capabilities of Graph Neural Networks (GNNs) on target graphs. Structure domain shifts frequently occur, especially when graph data are collected at different times or from varying areas, resulting in poor performance of GNNs on target graphs. Surprisingly, we find that simply incorporating an auxiliary loss function for denoising graph edges on target graphs can be extremely effective in enhancing GNN performance on target graphs. Based on this insight, we propose our framework, GraphDeT, a framework that integrates this auxiliary edge task into GNN training for node classification under domain adaptation. Our theoretical analysis connects this auxiliary edge task to the graph generalization bound with -distance, demonstrating such auxiliary task can imposes a constraint which tightens the bound and thereby improves generalization. The experimental results demonstrate superior performance compared to the existing baselines in handling both time and regional domain graph shifts.
Authors: Vedansh Sharma
Abstract: Classical convergence guarantees for gradient-based learning in games require the pseudo-gradient to be (strongly) monotone in Euclidean geometry as shown by rosen(1965), a condition that often fails even in simple games with strong cross-player couplings. We introduce Small-Gain Nash (SGN), a block small-gain condition in a custom block-weighted geometry. SGN converts local curvature and cross-player Lipschitz coupling bounds into a tractable certificate of contraction. It constructs a weighted block metric in which the pseudo-gradient becomes strongly monotone on any region where these bounds hold, even when it is non-monotone in the Euclidean sense. The continuous flow is exponentially contracting in this designed geometry, and projected Euler and RK4 discretizations converge under explicit step-size bounds derived from the SGN margin and a local Lipschitz constant. Our analysis reveals a certified "timescale band", a non-asymptotic, metric-based certificate that plays a TTUR-like role: rather than forcing asymptotic timescale separation via vanishing, unequal step sizes, SGN identifies a finite band of relative metric weights for which a single-step-size dynamics is provably contractive. We validate the framework on quadratic games where Euclidean monotonicity analysis fails to predict convergence, but SGN successfully certifies it, and extend the construction to mirror/Fisher geometries for entropy-regularized policy gradient in Markov games. The result is an offline certification pipeline that estimates curvature, coupling, and Lipschitz parameters on compact regions, optimizes block weights to enlarge the SGN margin, and returns a structural, computable convergence certificate consisting of a metric, contraction rate, and safe step-sizes for non-monotone games.
Authors: Ci Zhang, Huayu Li, Changdi Yang, Jiangnan Xia, Yanzhi Wang, Xiaolong Ma, Jin Lu, Ao Li, Geng Yuan
Abstract: Recent studies show that using diffusion models for time series signal reconstruction holds great promise. However, such approaches remain largely unexplored in the domain of medical time series. The unique characteristics of the physiological time series signals, such as multivariate, high temporal variability, highly noisy, and artifact-prone, make deep learning-based approaches still challenging for tasks such as imputation. Hence, we propose a novel Mixture of Experts (MoE)-based noise estimator within a score-based diffusion framework. Specifically, the Receptive Field Adaptive MoE (RFAMoE) module is designed to enable each channel to adaptively select desired receptive fields throughout the diffusion process. Moreover, recent literature has found that when generating a physiological signal, performing multiple inferences and averaging the reconstructed signals can effectively reduce reconstruction errors, but at the cost of significant computational and latency overhead. We design a Fusion MoE module and innovatively leverage the nature of MoE module to generate K noise signals in parallel, fuse them using a routing mechanism, and complete signal reconstruction in a single inference step. This design not only improves performance over previous methods but also eliminates the substantial computational cost and latency associated with multiple inference processes. Extensive results demonstrate that our proposed framework consistently outperforms diffusion-based SOTA works on different tasks and datasets.
Authors: Simon Chung, Colby J. Vorland, Donna L. Maney, Andrew W. Brown
Abstract: Datasets may contain observations with multiple labels. If the labels are not mutually exclusive, and if the labels vary greatly in frequency, obtaining a sample that includes sufficient observations with scarcer labels to make inferences about those labels, and which deviates from the population frequencies in a known manner, creates challenges. In this paper, we consider a multivariate Bernoulli distribution as our underlying distribution of a multi-label problem. We present a novel sampling algorithm that takes label dependencies into account. It uses observed label frequencies to estimate multivariate Bernoulli distribution parameters and calculate weights for each label combination. This approach ensures the weighted sampling acquires target distribution characteristics while accounting for label dependencies. We applied this approach to a sample of research articles from Web of Science labeled with 64 biomedical topic categories. We aimed to preserve category frequency order, reduce frequency differences between most and least common categories, and account for category dependencies. This approach produced a more balanced sub-sample, enhancing the representation of minority categories.
Authors: Metod Jazbec, Theo X. Olausson, Louis B\'ethune, Pierre Ablin, Michael Kirchhof, Jo\~ao Monteiro, Victor Turrisi, Jason Ramapuram, Marco Cuturi
Abstract: Diffusion (Large) Language Models (dLLMs) now match the downstream performance of their autoregressive counterparts on many tasks, while holding the promise of being more efficient during inference. One particularly successful variant is masked discrete diffusion, in which a buffer filled with special mask tokens is progressively replaced with tokens sampled from the model's vocabulary. Efficiency can be gained by unmasking several tokens in parallel, but doing too many at once risks degrading the generation quality. Thus, one critical design aspect of dLLMs is the sampling procedure that selects, at each step of the diffusion process, which tokens to replace. Indeed, recent work has found that heuristic strategies such as confidence thresholding lead to both higher quality and token throughput compared to random unmasking. However, such heuristics have downsides: they require manual tuning, and we observe that their performance degrades with larger buffer sizes. In this work, we instead propose to train sampling procedures using reinforcement learning. Specifically, we formalize masked diffusion sampling as a Markov decision process in which the dLLM serves as the environment, and propose a lightweight policy architecture based on a single-layer transformer that maps dLLM token confidences to unmasking decisions. Our experiments show that these trained policies match the performance of state-of-the-art heuristics when combined with semi-autoregressive generation, while outperforming them in the full diffusion setting. We also examine the transferability of these policies, finding that they can generalize to new underlying dLLMs and longer sequence lengths. However, we also observe that their performance degrades when applied to out-of-domain data, and that fine-grained tuning of the accuracy-efficiency trade-off can be challenging with our approach.
Authors: Kesheng Chen, Wenjian Luo, Zhenqian Zhu, Yamin Hu, Yiya Xi
Abstract: Constructing a Pareto set is pivotal for navigating the capability-efficiency trade-offs in Large Language Models (LLMs); however, existing merging techniques remain inadequate for this task. Coarse-grained, model-level methods yield only a sparse set of suboptimal solutions, while fine-grained, layer-wise approaches suffer from the "curse of dimensionality," rendering the search space computationally intractable. To resolve this dichotomy, we propose BAMBO (Bayesian Adaptive Multi-objective Block-wise Optimization), a novel framework that automatically constructs the LLM Pareto set. BAMBO renders the search tractable by introducing a Hybrid Optimal Block Partitioning strategy. Formulated as a 1D clustering problem, this strategy leverages a dynamic programming approach to optimally balance intra-block homogeneity and inter-block information distribution, thereby dramatically reducing dimensionality without sacrificing critical granularity. The entire process is automated within an evolutionary loop driven by the q-Expected Hypervolume Improvement (qEHVI) acquisition function. Experiments demonstrate that BAMBO discovers a superior and more comprehensive Pareto frontier than baselines, enabling agile model selection tailored to diverse operational constraints. Code is available at: https://github.com/xin8coder/BAMBO.
Authors: Jo\~ao Lucas Luz Lima Sarcinelli, Ricardo Marcondes Marcacini
Abstract: Large Language Models (LLMs) have become effective zero-shot classifiers, but their high computational requirements and environmental costs limit their practicality for large-scale annotation in high-performance computing (HPC) environments. To support more sustainable workflows, we present Text2Graph, an open-source Python package that provides a modular implementation of existing text-to-graph classification approaches. The framework enables users to combine LLM-based partial annotation with Graph Neural Network (GNN) label propagation in a flexible manner, making it straightforward to swap components such as feature extractors, edge construction methods, and sampling strategies. We benchmark Text2Graph on a zero-shot setting using five datasets spanning topic classification and sentiment analysis tasks, comparing multiple variants against other zero-shot approaches for text classification. In addition to reporting performance, we provide detailed estimates of energy consumption and carbon emissions, showing that graph-based propagation achieves competitive results at a fraction of the energy and environmental cost.
Authors: Qiyang Li, Seohong Park, Sergey Levine
Abstract: Temporal-difference (TD) methods learn state and action values efficiently by bootstrapping from their own future value predictions, but such a self-bootstrapping mechanism is prone to bootstrapping bias, where the errors in the value targets accumulate across steps and result in biased value estimates. Recent work has proposed to use chunked critics, which estimate the value of short action sequences ("chunks") rather than individual actions, speeding up value backup. However, extracting policies from chunked critics is challenging: policies must output the entire action chunk open-loop, which can be sub-optimal for environments that require policy reactivity and also challenging to model especially when the chunk length grows. Our key insight is to decouple the chunk length of the critic from that of the policy, allowing the policy to operate over shorter action chunks. We propose a novel algorithm that achieves this by optimizing the policy against a distilled critic for partial action chunks, constructed by optimistically backing up from the original chunked critic to approximate the maximum value achievable when a partial action chunk is extended to a complete one. This design retains the benefits of multi-step value propagation while sidestepping both the open-loop sub-optimality and the difficulty of learning action chunking policies for long action chunks. We evaluate our method on challenging, long-horizon offline goal-conditioned tasks and show that it reliably outperforms prior methods. Code: github.com/ColinQiyangLi/dqc.
Authors: Sandeep Polisetty, Juelin Liu, Kobi Falus, Yi Ren Fung, Seung-Hwan Lim, Hui Guan, Marco Serafini
Abstract: Graph neural networks (GNNs), an emerging class of machine learning models for graphs, have gained popularity for their superior performance in various graph analytical tasks. Mini-batch training is commonly used to train GNNs on large graphs, and data parallelism is the standard approach to scale mini-batch training across multiple GPUs. Data parallel approaches contain redundant work as subgraphs sampled by different GPUs contain significant overlap. To address this issue, we introduce a hybrid parallel mini-batch training paradigm called split parallelism. Split parallelism avoids redundant work by splitting the sampling, loading, and training of each mini-batch across multiple GPUs. Split parallelism, however, introduces communication overheads that can be more than the savings from removing redundant work. We further present a lightweight partitioning algorithm that probabilistically minimizes these overheads. We implement split parallelism in GSplit and show that it outperforms state-of-the-art mini-batch training systems like DGL, Quiver, and $P^3$.
Authors: Alan J. X. Guo, Sihan Sun, Xiang Wei, Mengyi Wei, Xin Chen
Abstract: With the emergence of new storage and communication methods, the insertion, deletion, and substitution (IDS) channel has attracted considerable attention. However, many topics on the IDS channel and the associated Levenshtein distance remain open, making the invention of a novel IDS-correcting code a hard task. Furthermore, current studies on single-IDS-correcting code misalign with the requirements of applications which necessitates the correcting of multiple errors. Compromise solutions have involved shortening codewords to reduce the chance of multiple errors. However, the code rates of existing codes are poor at short lengths, diminishing the overall storage density. In this study, a novel method is introduced for designing high-code-rate single-IDS-correcting codewords through deep Levenshtein distance embedding. A deep learning model is utilized to project the sequences into embedding vectors that preserve the Levenshtein distances between the original sequences. This embedding space serves as a proxy for the complex Levenshtein domain, within which algorithms for codeword search and segment correcting is developed. While the concept underpinning this approach is straightforward, it bypasses the mathematical challenges typically encountered in code design. The proposed method results in a code rate that outperforms existing combinatorial solutions, particularly for designing short-length codewords.
Authors: Jianlv Chen, Shitao Xiao, Peitian Zhang, Kun Luo, Defu Lian, Zheng Liu
Abstract: In this paper, we introduce a new embedding model called M3-Embedding, which is distinguished for its versatility in \textit{Multi-Linguality}, \textit{Multi-Functionality}, and \textit{Multi-Granularity}. It provides a uniform support for the semantic retrieval of more than 100 working languages. It can simultaneously accomplish the three common retrieval functionalities: dense retrieval, multi-vector retrieval, and sparse retrieval. Besides, it is also capable of processing inputs of different granularities, spanning from short sentences to long documents of up to 8,192 tokens. The effective training of M3-Embedding presents a series of technical contributions. Notably, we propose a novel self-knowledge distillation approach, where the relevance scores from different retrieval functionalities can be integrated as the teacher signal to enhance the training quality. We also optimize the batching strategy, which enables a large batch size and high training throughput to improve the discriminativeness of embeddings. M3-Embedding exhibits a superior performance in our experiment, leading to new state-of-the-art results on multilingual, cross-lingual, and long-document retrieval benchmarks.
Authors: Yash Sarrof, Yana Veitsman, Michael Hahn
Abstract: Recently, recurrent models based on linear state space models (SSMs) have shown promising performance in language modeling (LM), competititve with transformers. However, there is little understanding of the in-principle abilities of such models, which could provide useful guidance to the search for better LM architectures. We present a comprehensive theoretical study of the capacity of such SSMs as it compares to that of transformers and traditional RNNs. We find that SSMs and transformers have overlapping but distinct strengths. In star-free state tracking, SSMs implement straightforward and exact solutions to problems that transformers struggle to represent exactly. They can also model bounded hierarchical structure with optimal memory even without simulating a stack. On the other hand, we identify a design choice in current SSMs that limits their expressive power. We discuss implications for SSM and LM research, and verify results empirically on a recent SSM, Mamba.
Authors: Kanghee Park, Jiayu Wang, Taylor Berg-Kirkpatrick, Nadia Polikarpova, Loris D'Antoni
Abstract: Large Language Models (LLMs) struggle with reliably generating highly structured outputs, such as program code, mathematical formulas, or well-formed markup. Constrained decoding approaches mitigate this problem by greedily restricting what tokens an LLM can output at each step to guarantee that the output matches a given constraint. Specifically, in grammar-constrained decoding (GCD), the LLM's output must follow a given grammar. In this paper, we demonstrate that GCD techniques (and in general constrained decoding techniques) can distort the LLM's distribution, leading to outputs that are grammatical but appear with likelihoods that are not proportional to the ones given by the LLM, and so ultimately are low-quality. We call the problem of aligning sampling with a grammar constraint, grammar-aligned decoding (GAD), and propose adaptive sampling with approximate expected futures (ASAp), a decoding algorithm that guarantees the output to be grammatical while provably producing outputs that match the conditional probability of the LLM's distribution conditioned on the given grammar constraint. Our algorithm uses prior sample outputs to soundly overapproximate the future grammaticality of different output prefixes. Our evaluation on code generation and structured NLP tasks shows how ASAp often produces outputs with higher likelihood (according to the LLM's distribution) than existing GCD techniques, while still enforcing the desired grammatical constraints.
Authors: Inass Soukarieh, Gerhard Hessler, Herv\'e Minoux, Marcel Mohr, Friedemann Schmidt, Jan Wenzel, Pierre Barbillon, Hugo Gangloff, Pierre Gloaguen
Abstract: Mathematical modeling in systems toxicology enables a comprehensive understanding of the effects of pharmaceutical substances on cardiac health. However, the complexity of these models limits their widespread application in early drug discovery. In this paper, we introduce a novel approach to solving parameterized models of cardiac action potentials by combining meta-learning techniques with Systems Biology-Informed Neural Networks (SBINNs). The proposed method, hyperSBINN, effectively addresses the challenge of predicting the effects of various compounds at different concentrations on cardiac action potentials, outperforming traditional differential equation solvers in speed. Our model efficiently handles scenarios with limited data and complex parameterized differential equations. The hyperSBINN model demonstrates robust performance in predicting APD90 values, indicating its potential as a reliable tool for modeling cardiac electrophysiology and aiding in preclinical drug development. This framework represents an advancement in computational modeling, offering a scalable and efficient solution for simulating and understanding complex biological systems.
Authors: Jin-Hong Du, Kathryn Roeder, Larry Wasserman
Abstract: Data integration methods aim to extract low-dimensional embeddings from high-dimensional outcomes to remove unwanted variations, such as batch effects and unmeasured covariates, across heterogeneous datasets. However, multiple hypothesis testing after integration can be biased due to data-dependent processes. We introduce a robust post-integrated inference method that accounts for latent heterogeneity by utilizing control outcomes. Leveraging causal interpretations, we derive nonparametric identifiability of the direct effects using negative control outcomes. By utilizing surrogate control outcomes as an extension of negative control outcomes, we develop semiparametric inference on projected direct effect estimands, accounting for hidden mediators, confounders, and moderators. These estimands remain statistically meaningful under model misspecifications and with error-prone embeddings. We provide bias quantifications and finite-sample linear expansions with uniform concentration bounds. The proposed doubly robust estimators are consistent and efficient under minimal assumptions and potential misspecification, facilitating data-adaptive estimation with machine learning algorithms. Our proposal is evaluated using random forests through simulations and analysis of single-cell CRISPR perturbed datasets, which may contain potential unmeasured confounders.
Authors: Meenatchi Sundaram Muthu Selva Annamalai, Borja Balle, Jamie Hayes, Emiliano De Cristofaro
Abstract: The Differentially Private Stochastic Gradient Descent (DP-SGD) algorithm supports the training of machine learning (ML) models with formal Differential Privacy (DP) guarantees. Traditionally, DP-SGD processes training data in batches using Poisson subsampling to select each batch at every iteration. More recently, shuffling has become a common alternative due to its better compatibility and lower computational overhead. However, computing tight theoretical DP guarantees under shuffling remains an open problem. As a result, models trained with shuffling are often evaluated as if Poisson subsampling were used, which might result in incorrect privacy guarantees. This raises a compelling research question: can we verify whether there are gaps between the theoretical DP guarantees reported by state-of-the-art models using shuffling and their actual leakage? To do so, we define novel DP-auditing procedures to analyze DP-SGD with shuffling and measure their ability to tightly estimate privacy leakage vis-\`a-vis batch sizes, privacy budgets, and threat models. Overall, we demonstrate that DP models trained using this approach have considerably overestimated their privacy guarantees (by up to 4 times). However, we also find that the gap between the theoretical Poisson DP guarantees and the actual privacy leakage from shuffling is not uniform across all parameter settings and threat models. Finally, we study two common variations of the shuffling procedure that result in even further privacy leakage (up to 10 times). Overall, our work highlights the risk of using shuffling instead of Poisson subsampling in the absence of rigorous analysis methods.
Authors: Nasim Soltani, Michael Loehning, Kaushik Chowdhury
Abstract: Artificial Intelligence (AI)-native receivers prove significant performance improvement in high noise regimes and can potentially reduce communication overhead compared to the traditional receiver. However, their performance highly depends on the representativeness of the training dataset. A major issue is the uncertainty of whether the training dataset covers all test environments and waveform configurations, and thus, whether the trained model is robust in practical deployment conditions. To this end, we propose a joint measurement-recovery framework for AI-native transceivers post deployment, called VERITAS, that continuously looks for distribution shifts in the received signals and triggers finite re-training spurts. VERITAS monitors the wireless channel using 5G pilots fed to an auxiliary neural network that detects out-of-distribution channel profile, transmitter speed, and delay spread. As soon as such a change is detected, a traditional (reference) receiver is activated, which runs for a period of time in parallel to the AI-native receiver. Finally, VERTIAS compares the bit probabilities of the AI-native and the reference receivers for the same received data inputs, and decides whether or not a retraining process needs to be initiated. Our evaluations reveal that VERITAS can detect changes in the channel profile, transmitter speed, and delay spread with 99%, 97%, and 69% accuracies, respectively, followed by timely initiation of retraining for 86%, 93.3%, and 94.8% of inputs in channel profile, transmitter speed, and delay spread test sets, respectively.
Authors: Zhewei Kang, Xuandong Zhao, Dawn Song
Abstract: Best-of-N selection is a key technique for improving the reasoning performance of Large Language Models (LLMs) through increased test-time computation. Current state-of-the-art methods often employ computationally intensive reward models for response evaluation and selection. Reward-free alternatives, like self-consistency and universal self-consistency, are limited in their ability to handle open-ended generation tasks or scale effectively. To address these limitations, we propose self-certainty, a novel and efficient metric that leverages the inherent probability distribution of LLM outputs to estimate response quality without requiring external reward models. We hypothesize that higher distributional self-certainty, aggregated across multiple samples, correlates with improved response accuracy, as it reflects greater confidence in the generated output. Through extensive experiments on various reasoning tasks, we demonstrate that self-certainty (1) scales effectively with increasing sample size N, akin to reward models but without the computational overhead; (2) complements chain-of-thought, improving reasoning performance beyond greedy decoding; and (3) generalizes to open-ended tasks where traditional self-consistency methods fall short. Our findings establish self-certainty as a practical and efficient way for improving LLM reasoning capabilities. The code is available at https://github.com/backprop07/Self-Certainty
Authors: Jake McNaughton, A. H. Abbas, Ivan S. Maksymov
Abstract: The mathematical complexity and high dimensionality of neural networks slow both training and deployment, demanding heavy computational resources. This has driven the search for alternative architectures built from novel components, including new activation functions. Taking a different approach from state-of-the-art neural and neuromorphic computational systems, we employ the current-voltage characteristic of a tunnel diode as a quantum physics-based activation function for deep networks. This tunnel-diode activation function (TDAF) outperforms standard activations in deep architectures, delivering lower loss and higher accuracy in both training and evaluation. We also highlight its promise for implementation in electronic hardware aimed at neuromorphic, ternarised and energy efficient AI systems. Speaking broadly, our work lays a solid foundation for a new bridge between machine learning, semiconductor electronics and quantum physics -- bringing together quantum tunnelling, a phenomenon recognised in six Nobel Prizes (including the 2025 award), and contemporary AI research.
Authors: Yuting Hu, Chenhui Xu, Ruiyang Qin, Dancheng Liu, Amir Nassereldine, Yiyu Shi, Jinjun Xiong
Abstract: Partial perception deficits can compromise autonomous vehicle safety by disrupting environmental understanding. Existing protocols typically default to entirely risk-avoidant actions such as immediate stops, which are detrimental to navigation goals and lack flexibility for rare driving scenarios. Yet, in cases of minor risk, halting the vehicle may be unnecessary, and more adaptive responses are preferable. In this paper, we propose LLM-RCO, a risk-averse framework leveraging large language models (LLMs) to integrate human-like driving commonsense into autonomous systems facing perception deficits. LLM-RCO features four key modules interacting with the dynamic driving environment: hazard inference, short-term motion planner, action condition verifier, and safety constraint generator, enabling proactive and context-aware actions in such challenging conditions. To enhance the driving decision-making of LLMs, we construct DriveLM-Deficit, a dataset of 53,895 video clips featuring deficits of safety-critical objects, annotated for LLM fine-tuning in hazard detection and motion planning. Extensive experiments in adverse driving conditions with the CARLA simulator demonstrate that LLM-RCO promotes proactive maneuvers over purely risk-averse actions in perception deficit scenarios, underscoring its value for boosting autonomous driving resilience against perception loss challenges.
Authors: Sophia Hager, David Mueller, Kevin Duh, Nicholas Andrews
Abstract: As large language models (LLMs) are increasingly used for factual question-answering, it becomes more important for LLMs to have the capability to communicate the likelihood that their answer is correct. For these verbalized expressions of uncertainty to be meaningful, they should reflect the error rates at the expressed level of confidence. However, when prompted to express confidence, the error rates of current LLMs are inconsistent with their communicated confidences, highlighting the need for uncertainty quantification methods. Many prior methods calculate lexical uncertainty, estimating a model's confidence in the specific string it generated. In some cases, however, it may be more useful to estimate semantic uncertainty, or the model's confidence in the answer regardless of how it is verbalized. We propose a simple procedure, uncertainty distillation, to teach an LLM to verbalize calibrated semantic confidences. Using held-out data to map initial uncertainty estimates to meaningful probabilities, we create examples annotated with verbalized probabilities for supervised fine-tuning. We find that our method yields verbalized confidences that correlate well with observed error rates, even when compared to strong baselines, some of which are more than twenty times slower at inference time. Additionally, we demonstrate that our method can be applied to black-box models that allow API-based fine-tuning, resulting in estimates of uncertainty that are both more effective and more efficient than any of our baselines.
Authors: Kilian Tscharke, Maximilian Wendlinger, Sebastian Issel, Pascal Debus
Abstract: Anomaly Detection (AD) is critical in data analysis, particularly within the domain of IT security. In this study, we explore the potential of Quantum Machine Learning for application to AD with special focus on the robustness to noise and adversarial attacks. We build upon previous work on Quantum Support Vector Regression (QSVR) for semisupervised AD by conducting a comprehensive benchmark on IBM quantum hardware using eleven datasets. Our results demonstrate that QSVR achieves strong classification performance and even outperforms the noiseless simulation on two of these datasets. Moreover, we investigate the influence of - in the NISQ-era inevitable - quantum noise on the performance of the QSVR. Our findings reveal that the model exhibits robustness to depolarizing, phase damping, phase flip, and bit flip noise, while amplitude damping and miscalibration noise prove to be more disruptive. Finally, we explore the domain of Quantum Adversarial Machine Learning by demonstrating that QSVR is highly vulnerable to adversarial attacks, with neither quantum noise nor adversarial training improving the model's robustness against such attacks.
Authors: Toshihiro Ota, Yuma Fujimoto
Abstract: The understanding of a dynamical system's properties can be significantly advanced by establishing it as a Hamiltonian system and then systematically exploring its inherent symmetries. By formulating agents' strategies and cumulative payoffs as canonically conjugate variables, we identify the Hamiltonian function that generates the dynamics of poly-matrix zero-sum games. We reveal the symmetries of our Hamiltonian and derive the associated conserved quantities, showing how the conservation of probability and the invariance of the Fenchel coupling are intrinsically encoded within the system. Furthermore, we propose the dissipation FTRL (DFTRL) dynamics by introducing a perturbation that dissipates the Fenchel coupling, proving convergence to the Nash equilibrium and linking DFTRL to last-iterate convergent algorithms. Our results highlight the potential of Hamiltonian dynamics in uncovering the structural properties of learning dynamics in games, and pave the way for broader applications of Hamiltonian dynamics in game theory and machine learning.
Authors: Mikhail Menschikov, Alexander Kharitonov, Maiia Kotyga, Vadim Porvatov, Anna Zhukovskaya, David Kagramanyan, Egor Shvetsov, Evgeny Burnaev
Abstract: Large Language Models (LLMs) exhibit position bias systematically underweighting information based on its location in the context but how this bias varies across languages and models remains unclear. We conduct a multilingual study across five typologically diverse languages (English, Russian, German, Hindi, Vietnamese) and five model architectures, analyzing how position bias interacts with prompting strategies and affects output entropy. Our key findings are: (1) Position bias is primarily model-driven but shows language-specific nuances. Notably, Qwen2.5-7B-Instruct, DeepSeek 7B Chat and Mistral 7B consistently favor late positions challenging the common assumption of universal early-token preference. (2) Explicitly instructing the model, in the presence of irrelevant distractors, that "the most relevant context to the query is marked as 1" unexpectedly reduces accuracy across all languages, questioning standard prompt-engineering practices. (3) Accuracy consistently drops most when relevant information appears in the middle of the context, yet this is not reflected in a corresponding increase in output entropy, suggesting the model remains confident even when it fails to use mid-context cues.
Authors: Luc Brogat-Motte, Alessandro Rudi, Riccardo Bonalli
Abstract: We address the problem of safely learning controlled stochastic dynamics from discrete-time trajectory observations, ensuring system trajectories remain within predefined safe regions during both training and deployment. Safety-critical constraints of this kind are crucial in applications such as autonomous robotics, finance, and biomedicine. We introduce a method that ensures safe exploration and efficient estimation of system dynamics by iteratively expanding an initial known safe control set using kernel-based confidence bounds. After training, the learned model enables predictions of the system's dynamics and permits safety verification of any given control. Our approach requires only mild smoothness assumptions and access to an initial safe control set, enabling broad applicability to complex real-world systems. We provide theoretical guarantees for safety and derive adaptive learning rates that improve with increasing Sobolev regularity of the true dynamics. Experimental evaluations demonstrate the practical effectiveness of our method in terms of safety, estimation accuracy, and computational efficiency.
Authors: Sergio Burdisso, S\'everin Baroudi, Yanis Labrak, David Grunert, Pawel Cyrta, Yiyang Chen, Srikanth Madikeri, Esa\'u Villatoro-Tello, Thomas Schaaf, Ricard Marxer, Petr Motlicek
Abstract: We present SDialog, an MIT-licensed open-source Python toolkit that unifies dialog generation, evaluation and mechanistic interpretability into a single end-to-end framework for building and analyzing LLM-based conversational agents. Built around a standardized \texttt{Dialog} representation, SDialog provides: (1) persona-driven multi-agent simulation with composable orchestration for controlled, synthetic dialog generation, (2) comprehensive evaluation combining linguistic metrics, LLM-as-a-judge and functional correctness validators, (3) mechanistic interpretability tools for activation inspection and steering via feature ablation and induction, and (4) audio generation with full acoustic simulation including 3D room modeling and microphone effects. The toolkit integrates with all major LLM backends, enabling mixed-backend experiments under a unified API. By coupling generation, evaluation, and interpretability in a dialog-centric architecture, SDialog enables researchers to build, benchmark and understand conversational systems more systematically.
Authors: Hiren Madhu, Jo\~ao Felipe Rocha, Tinglin Huang, Siddharth Viswanath, Smita Krishnaswamy, Rex Ying
Abstract: Single-cell transcriptomics and proteomics have become a great source for data-driven insights into biology, enabling the use of advanced deep learning methods to understand cellular heterogeneity and gene expression at the single-cell level. With the advent of spatial-omics data, we have the promise of characterizing cells within their tissue context as it provides both spatial coordinates and intra-cellular transcriptional or protein counts. Proteomics offers a complementary view by directly measuring proteins, which are the primary effectors of cellular function and key therapeutic targets. However, existing models either ignore the spatial information or the complex genetic and proteomic programs within cells. Thus they cannot infer how cell internal regulation adapts to microenvironmental cues. Furthermore, these models often utilize fixed gene vocabularies, hindering their generalizability unseen genes. In this paper, we introduce HEIST, a hierarchical graph transformer foundation model for spatial transcriptomics and proteomics. HEIST models tissues as hierarchical graphs. The higher level graph is a spatial cell graph, and each cell in turn, is represented by its lower level gene co-expression network graph. HEIST achieves this by performing both intra-level and cross-level message passing to utilize the hierarchy in its embeddings and can thus generalize to novel datatypes including spatial proteomics without retraining. HEIST is pretrained on 22.3M cells from 124 tissues across 15 organs using spatially-aware contrastive and masked autoencoding objectives. Unsupervised analysis of HEIST embeddings reveals spatially informed subpopulations missed by prior models. Downstream evaluations demonstrate generalizability to proteomics data and state-of-the-art performance in clinical outcome prediction, cell type annotation, and gene imputation across multiple technologies.
Authors: Fernando Galaz Prieto, Antti Lassila, Maryam Samavaki, Sampsa Pursiainen
Abstract: As DBS technology advances toward directional leads and optimization-based current steering, this study aims to improve the selection of electrode contact configurations using the recently developed L1-norm regularized L1-norm fitting (L1L1) method. The focus is in particular on L1L1's capability to incorporate a priori lead field uncertainty, offering a potential advantage over conventional approaches that do not account for such variability. Our optimization framework incorporates uncertainty by constraining the solution space based on lead field attenuation. This reflects physiological expectations about the VTA and serves to avoid overfitting. By applying this method to 8- and 40-contact electrode configurations, we optimize current distributions within a discretized finite element (FE) model, focusing on the lead field's characteristics. The model accounts for uncertainty through these explicit constraints, enhancing the feasibility, focality, and robustness of the resulting solutions. The L1L1 method was validated through a series of numerical experiments using both noiseless and noisy lead fields, where the noise level was selected to reflect attenuation within VTA. It successfully fits and regularizes the current distribution across target structures, with hyperparameter optimization extracting either bipolar or multipolar electrode configurations. These configurations aim to maximize focused current density or prioritize a high gain field ratio in a discretized FE model. Compared to traditional methods, the L1L1 approach showed competitive performance in concentrating stimulation within the target region while minimizing unintended current spread, particularly under noisy conditions. By incorporating uncertainty directly into the optimization process, we obtain a noise-robust framework for current steering, allowing for variations in lead field models and simulation parameters.
Authors: Meenatchi Sundaram Muthu Selva Annamalai, Borja Balle, Jamie Hayes, Georgios Kaissis, Emiliano De Cristofaro
Abstract: In this paper, we systematize research on auditing Differential Privacy (DP) techniques, aiming to identify key insights and open challenges. First, we introduce a comprehensive framework for reviewing work in the field and establish three cross-contextual desiderata that DP audits should target -- namely, efficiency, end-to-end-ness, and tightness. Then, we systematize the modes of operation of state-of-the-art DP auditing techniques, including threat models, attacks, and evaluation functions. This allows us to highlight key details overlooked by prior work, analyze the limiting factors to achieving the three desiderata, and identify open research problems. Overall, our work provides a reusable and systematic methodology geared to assess progress in the field and identify friction points and future directions for our community to focus on.
Authors: Xinzhuo Li, Adheesh Juvekar, Jiaxun Zhang, Xingyou Liu, Muntasir Wahed, Kiet A. Nguyen, Yifan Shen, Tianjiao Yu, Ismini Lourentzou
Abstract: Segmentation Vision-Language Models (VLMs) have significantly advanced grounded visual understanding, yet they remain prone to pixel-grounding hallucinations, producing masks for incorrect objects or for objects that are entirely absent. Existing evaluations rely almost entirely on text- or label-based perturbations, which check only whether the predicted mask matches the queried label. Such evaluations overlook the spatial footprint and severity of hallucination and therefore fail to reveal vision-driven hallucinations, which are more challenging and more prevalent. To address this gap, we formalize the task of Counterfactual Segmentation Reasoning (CSR), where a model must segment the referenced object in the factual image and abstain in its counterfactual counterpart. To support this task, we curate HalluSegBench, the first large-scale benchmark to diagnose referring and reasoning expression segmentation hallucinations using controlled visual counterfactuals, alongside new evaluation metrics that measure hallucination severity and disentangle vision- and language-driven failure modes. We further introduce RobustSeg, a segmentation VLM trained with counterfactual fine-tuning (CFT) to learn when to segment and when to abstain. Experimental results confirm RobustSeg reduces hallucinations by 30%, while improving segmentation performance on FP-RefCOCO(+/g).
Authors: Woonsang Kang, Joohyung Lee, Seungjun Kim, Jungchan Cho, Yoonseon Oh
Abstract: Grasp pose detection (GPD) is a fundamental capability for robotic autonomy, but its reliance on large, diverse datasets creates significant data privacy and centralization challenges. Federated Learning (FL) offers a privacy-preserving solution, but its application to GPD is hindered by the substantial communication overhead of large models, a key issue for resource-constrained robots. To address this, we propose a novel module-wise FL framework that begins by analyzing the learning dynamics of the GPD model's functional components. This analysis identifies slower-converging modules, to which our framework then allocates additional communication effort. This is realized through a two-phase process: a standard full-model training phase is followed by a communication-efficient phase where only the identified subset of slower-converging modules is trained and their partial updates are aggregated. Extensive experiments on the GraspNet-1B dataset demonstrate that our method outperforms standard FedAvg and other baselines, achieving higher accuracy for a given communication budget. Furthermore, real-world experiments on a physical robot validate our approach, showing a superior grasp success rate compared to baseline methods in cluttered scenes. Our work presents a communication-efficient framework for training robust, generalized GPD models in a decentralized manner, effectively improving the trade-off between communication cost and model performance.
Authors: Xintao Xia, Linjun Zhang, Zhanrui Cai
Abstract: Privacy preservation in machine learning, particularly through Differentially Private Stochastic Gradient Descent (DP-SGD), is critical for sensitive data analysis. However, existing statistical inference methods for SGD predominantly focus on cyclic subsampling, while DP-SGD requires randomized subsampling. This paper first bridges this gap by establishing the asymptotic properties of SGD under the randomized rule and extending these results to DP-SGD. For the output of DP-SGD, we show that the asymptotic variance decomposes into statistical, sampling, and privacy-induced components. Two methods are proposed for constructing valid confidence intervals: the plug-in method and the random scaling method. We also perform extensive numerical analysis, which shows that the proposed confidence intervals achieve nominal coverage rates while maintaining privacy.
Authors: Sungguk Cha, DongWook Kim, Taeseung Hahn, Mintae Kim, Youngsub Han, Byoung-Ki Jeon
Abstract: Optimizing queries for Retrieval-Augmented Generation (RAG) systems poses a significant challenge, particularly across diverse modal indices. We introduce RL-QR, a novel annotation-free reinforcement learning framework for query rewriting that eliminates the need for costly human-annotated data. By leveraging verifiable search rewards derived from index-aligned synthetic queries, RL-QR overcomes human-annotation dependencies, extending its applicability to various modalities and index domains. Experimental results demonstrate the framework's robustness, achieving substantial retrieval performance gains of up to 3.9$\times$ on lexical retrievers and 3.5$\times$ on semantic retrievers on the MTEB VIDORE V2 benchmark for unstructured visual documents, along with consistent 5\% to 10\% improvements on MS MARCO v2.1 and internal industrial datasets.
Authors: Zihao Wang, Junming Zhang
Abstract: Large Language Models (LLMs) have shown promising potential in business applications, particularly in enterprise decision support and strategic planning, yet current approaches often struggle to reconcile intricate operational analyses with overarching strategic goals across diverse market environments, leading to fragmented workflows and reduced collaboration across organizational levels. This paper introduces BusiAgent, a novel multi-agent framework leveraging LLMs for advanced decision-making in complex corporate environments. BusiAgent integrates three core innovations: an extended Continuous Time Markov Decision Process (CTMDP) for dynamic agent modeling, a generalized entropy measure to optimize collaborative efficiency, and a multi-level Stackelberg game to handle hierarchical decision processes. Additionally, contextual Thompson sampling is employed for prompt optimization, supported by a comprehensive quality assurance system to mitigate errors. Extensive empirical evaluations across diverse business scenarios validate BusiAgent's efficacy, demonstrating its capacity to generate coherent, client-focused solutions that smoothly integrate granular insights with high-level strategy, significantly outperforming established approaches in both solution quality and user satisfaction. By fusing cutting-edge AI technologies with deep business insights, BusiAgent marks a substantial step forward in AI-driven enterprise decision-making, empowering organizations to navigate complex business landscapes more effectively.
Authors: Juan-Pablo Ortega, Florian Rossmannek
Abstract: Recurrent neural networks (RNNs) have become increasingly popular in information processing tasks involving time series and temporal data. A fundamental property of RNNs is their ability to create reliable input/output responses, often linked to how the network handles its memory of the information it processed. Various notions have been proposed to conceptualize the behavior of memory in RNNs, including steady states, echo states, state forgetting, input forgetting, and fading memory. Although these notions are often used interchangeably, their precise relationships remain unclear. This work aims to unify these notions in a common language, derive new implications and equivalences between them, and provide alternative proofs to some existing results. By clarifying the relationships between these concepts, this research contributes to a deeper understanding of RNNs and their temporal information processing capabilities.
Authors: Matthew O'Callaghan, Kaisey S. Mandel, Gerry Gilmore
Abstract: Recent advances in neural density estimation have enabled powerful simulation-based inference (SBI) methods that can flexibly approximate Bayesian inference for intractable stochastic models. Although these methods have demonstrated reliable posterior estimation when the simulator accurately represents the underlying data generative process (DGP), recent work has shown that they perform poorly in the presence of model misspecification. This poses a significant issue for their use in real-world problems, due to simulators always misrepresenting the true DGP to a certain degree. In this paper, we introduce robust variational neural posterior estimation (RVNP), a method which addresses the problem of misspecification in amortised SBI by bridging the simulation-to-reality gap using variational inference and error modelling. We test RVNP on multiple benchmark tasks, including using real data from astronomy, and show that it can recover robust posterior inference in a data-driven manner without adopting hyperparameters or priors governing the misspecification influence.
Authors: Yu Gu, Jingjing Fu, Xiaodong Liu, Jeya Maria Jose Valanarasu, Noel CF Codella, Reuben Tan, Qianchu Liu, Ying Jin, Sheng Zhang, Jinyu Wang, Rui Wang, Lei Song, Guanghui Qin, Naoto Usuyama, Cliff Wong, Hao Cheng, HoHin Lee, Praneeth Sanapathi, Sarah Hilado, Tristan Naumann, Javier Alvarez-Valle, Jiang Bian, Mu Wei, Khalil Malik, Lidong Zhou, Jianfeng Gao, Eric Horvitz, Matthew P. Lungren, Doug Burger, Eric Topol, Hoifung Poon, Paul Vozila
Abstract: Large language models have demonstrated remarkable performance in a wide range of medical benchmarks. Yet underneath the seemingly promising results lie salient growth areas, especially in cutting-edge frontiers such as multimodal reasoning. In this paper, we introduce a series of adversarial stress tests to systematically assess the robustness of flagship models and medical benchmarks. Our study reveals prevalent brittleness in the presence of simple adversarial transformations: leading systems can guess the right answer even with key inputs removed, yet may get confused by the slightest prompt alterations, while fabricating convincing yet flawed reasoning traces. Using clinician-guided rubrics, we demonstrate that popular medical benchmarks vary widely in what they truly measure. Our study reveals significant competency gaps of frontier AI in attaining real-world readiness for health applications. If we want AI to earn trust in healthcare, we must demand more than leaderboard wins and must hold AI systems accountable to ensure robustness, sound reasoning, and alignment with real medical demands.
Authors: Minghui Wu, Zhen Gao
Abstract: This paper investigates multimodal semantic non-orthogonal transmission and fusion in hybrid analog-digital massive multiple-input multiple-output (MIMO). A Transformer-based cross-modal source-channel semantic-aware network (CSC-SA-Net) framework is conceived, where channel state information (CSI) reference signal (RS), feedback, analog-beamforming/combining, and baseband semantic processing are data-driven end-to-end (E2E) optimized at the base station (BS) and user equipments (UEs). CSC-SA-Net comprises five sub-networks: BS-side CSI-RS network (BS-CSIRS-Net), UE-side channel semantic-aware network (UE-CSANet), BS-CSANet, UE-side multimodal semantic fusion network (UE-MSFNet), and BS-MSFNet. Specifically, we firstly E2E train BS-CSIRS-Net, UE-CSANet, and BS-CSANet to jointly design CSI-RS, feedback, analog-beamforming/combining with maximum {\emph{physical-layer's}} spectral-efficiency. Meanwhile, we E2E train UE-MSFNet and BS-MSFNet for optimizing {\emph{application-layer's}} source semantic downstream tasks. On these pre-trained models, we further integrate application-layer semantic processing with physical-layer tasks to E2E train five subnetworks. Extensive simulations show that the proposed CSC-SA-Net outperforms traditional separated designs, revealing the advantage of cross-modal channel-source semantic fusion.
Authors: Huu Nguyen (Sonny), Victor May (Sonny), Harsh Raj (Sonny), Marianna Nezhurina (Sonny), Yishan Wang (Sonny), Yanqi Luo (Sonny), Minh Chien Vu (Sonny), Taishi Nakamura (Sonny), Ken Tsui (Sonny), Van Khue Nguyen (Sonny), David Salinas (Sonny), Aleksandra Krasnod\k{e}bska (Sonny), Christoph Schuhmann (Sonny), Mats Leon Richter (Sonny), Xuan-Son (Sonny), Vu, Jenia Jitsev
Abstract: We present MixtureVitae, an open-access pretraining corpus built to minimize legal risk while providing strong model performance. MixtureVitae follows a risk-mitigated sourcing strategy that combines public-domain and permissively licensed text (e.g., CC-BY/Apache) with carefully justified low-risk additions (e.g., government works and EU TDM-eligible sources), alongside targeted instruction, reasoning and synthetic data with documented provenance. We detail a transparent, multi-stage pipeline for license-aware filtering, safety and quality screening, and domain-aware mixing, and we release the dataset and curation recipes to support reproducible research. In controlled experiments using the open-sci-ref training protocol (fixed architectures at 130M/400M/1.3B/1.7B parameters; training budgets of 50B and 300B tokens), models trained on MixtureVitae consistently outperform other permissive datasets across a suite of standard benchmarks, and at the 1.7B/300B setting they surpass FineWeb-Edu and approach DCLM in the later stages of training. Performance is particularly strong on math/code and competitive on QA tasks. These results demonstrate that permissive-first, risk-mitigated data provides a practical and legally mitigated foundation for training capable LLMs, reducing reliance on indiscriminate web scraping without sacrificing competitiveness. Code: https://github.com/ontocord/mixturevitae
Authors: Yuki Yasuda, Ryo Onishi
Abstract: This study employs a neural network that represents the solution to a Schr\"odinger bridge problem to perform super-resolution of 2-m temperature in an urban area. Schr\"odinger bridges generally describe transformations between two data distributions based on diffusion processes. We use a specific Schr\"odinger-bridge model (SM) that directly transforms low-resolution data into high-resolution data, unlike denoising diffusion probabilistic models (simply, diffusion models; DMs) that generate high-resolution data from Gaussian noise. Low-resolution and high-resolution data were obtained from separate numerical simulations with a physics-based model under common initial and boundary conditions. Compared with a DM, the SM attains comparable accuracy at one-fifth the computational cost, requiring 50 neural-network evaluations per datum for the DM and only 10 for the SM. Furthermore, high-resolution samples generated by the SM exhibit larger variance, implying superior uncertainty quantification relative to the DM. Owing to the reduced computational cost of the SM, our results suggest the feasibility of real-time ensemble micrometeorological prediction using SM-based super-resolution.
Authors: Rohit Goswami (Science Institute and Faculty of Physical Sciences, University of Iceland, Reykjav\'ik, Iceland)
Abstract: Estimating reaction rates and chemical stability is fundamental, yet efficient methods for large-scale simulations remain out of reach despite advances in modeling and exascale computing. Direct simulation is limited by short timescales; machine-learned potentials require large data sets and struggle with transition state regions essential for reaction rates. Reaction network exploration with sufficient accuracy is hampered by the computational cost of electronic structure calculations, and even simplifications like harmonic transition state theory rely on prohibitively expensive saddle point searches. Surrogate model-based acceleration has been promising but hampered by overhead and numerical instability. This dissertation presents a holistic solution, co-designing physical representations, statistical models, and systems architecture in the Optimal Transport Gaussian Process (OT-GP) framework. Using physics-aware optimal transport metrics, OT-GP creates compact, chemically relevant surrogates of the potential energy surface, underpinned by statistically robust sampling. Alongside EON software rewrites for long timescale simulations, we introduce reinforcement learning approaches for both minimum-mode following (when the final state is unknown) and nudged elastic band methods (when endpoints are specified). Collectively, these advances establish a representation-first, modular approach to chemical kinetics simulation. Large-scale benchmarks and Bayesian hierarchical validation demonstrate state-of-the-art performance and practical exploration of chemical kinetics, transforming a longstanding theoretical promise into a working engine for discovery.
Authors: Andrij Vasylenko, Federico Ottomano, Christopher M. Collins, Rahul Savani, Matthew S. Dyer, Matthew J. Rosseinsky
Abstract: Discovering materials with new structural chemistry is key to achieving transformative functionality. Generative artificial intelligence offers a scalable route to propose candidate crystal structures. We introduce a reliable low-cost proxy for structural novelty as a conditioning property to steer generation towards novel yet physically plausible structures. We then develop a physics-informed diffusion model that embeds this descriptor of local environment diversity together with compactness as a stability metric to balance physical plausibility with structural novelty. Conditioning on these metrics improves generative performance across diffusion models, shifting generation away from structural motifs that dominate the training data. A chemically grounded validation protocol isolates those candidates that combine plausibility with structural novelty for physics-based calculation of energetic stability. Both the stability and the novelty of candidates emerging from this workflow can however change when the full potential energy surface at a candidate composition is evaluated with crystal structure prediction (CSP). This suggests a practical generative-CSP synergy for discovery-oriented exploration, where AI targets physically viable yet structurally distinct regions of chemical space for detailed physics-based assessment of novelty and stability.
Authors: Zhaoyang Yu, Jiayi Zhang, Huixue Su, Yufan Zhao, Yifan Wu, Mingyi Deng, Jinyu Xiang, Yizhang Lin, Lingxiao Tang, Yuyu Luo, Bang Liu, Chenglin Wu
Abstract: Real-world tasks require decisions at varying granularities, and humans excel at this by leveraging a unified cognitive representation where planning is fundamentally understood as a high-level form of action. However, current Large Language Model (LLM)-based agents lack this crucial capability to operate fluidly across decision granularities. This limitation stems from existing paradigms that enforce a rigid separation between high-level planning and low-level action, which impairs dynamic adaptability and limits generalization. We propose ReCode (Recursive Code Generation), a novel paradigm that addresses this limitation by unifying planning and action within a single code representation. In this representation, ReCode treats high-level plans as abstract placeholder functions, which the agent then recursively decomposes into finer-grained sub-functions until reaching primitive actions. This recursive approach dissolves the rigid boundary between plan and action, enabling the agent to dynamically control its decision granularity. Furthermore, the recursive structure inherently generates rich, multi-granularity training data, enabling models to learn hierarchical decision-making processes. Extensive experiments show ReCode significantly surpasses advanced baselines in inference performance and demonstrates exceptional data efficiency in training, validating our core insight that unifying planning and action through recursive code generation is a powerful and effective approach to achieving universal granularity control. The code is available at https://github.com/FoundationAgents/ReCode.
Authors: Jean Barbier, Francesco Camilli, Minh-Toan Nguyen, Mauro Pastore, Rudy Skerk
Abstract: For four decades statistical physics has been providing a framework to analyse neural networks. A long-standing question remained on its capacity to tackle deep learning models capturing rich feature learning effects, thus going beyond the narrow networks or kernel methods analysed until now. We positively answer through the study of the supervised learning of a multi-layer perceptron. Importantly, (i) its width scales as the input dimension, making it more prone to feature learning than ultra wide networks, and more expressive than narrow ones or ones with fixed embedding layers; and (ii) we focus on the challenging interpolation regime where the number of trainable parameters and data are comparable, which forces the model to adapt to the task. We consider the matched teacher-student setting. Therefore, we provide the fundamental limits of learning random deep neural network targets and identify the sufficient statistics describing what is learnt by an optimally trained network as the data budget increases. A rich phenomenology emerges with various learning transitions. With enough data, optimal performance is attained through the model's "specialisation" towards the target, but it can be hard to reach for training algorithms which get attracted by sub-optimal solutions predicted by the theory. Specialisation occurs inhomogeneously across layers, propagating from shallow towards deep ones, but also across neurons in each layer. Furthermore, deeper targets are harder to learn. Despite its simplicity, the Bayes-optimal setting provides insights on how the depth, non-linearity and finite (proportional) width influence neural networks in the feature learning regime that are potentially relevant in much more general settings.
Authors: Rafael Oliveira, Daniel M. Steinberg, Edwin V. Bonilla
Abstract: We present a general strategy for turning generative models into candidate solution samplers for batch Bayesian optimization (BO). The use of generative models for BO enables large batch scaling as generative sampling, optimization of non-continuous design spaces, and high-dimensional and combinatorial design. Inspired by the success of direct preference optimization (DPO), we show that one can train a generative model with noisy, simple utility values directly computed from observations to then form proposal distributions whose densities are proportional to the expected utility, i.e., BO's acquisition function values. Furthermore, this approach is generalizable beyond preference-based feedback to general types of reward signals and loss functions. This perspective avoids the construction of surrogate (regression or classification) models, common in previous methods that have used generative models for black-box optimization. Theoretically, we show that the generative models within the BO process approximately follow a sequence of distributions which asymptotically concentrate at the global optima under certain conditions. We also demonstrate this effect through experiments on challenging optimization problems involving large batches in high dimensions.
Authors: Qiang Wang, Liying Yang, Jiayun Song, Yifan Bai, Jingtao Du
Abstract: Cross-subject electroencephalography (EEG) emotion recognition remains a major challenge in brain-computer interfaces (BCIs) due to substantial inter-subject variability. Multi-Source Domain Adaptation (MSDA) offers a potential solution, but existing MSDA frameworks typically assume equal source quality, leading to negative transfer from low-reliability domains and prohibitive computational overhead due to multi-branch model designs. To address these limitations, we propose the Hard-Easy Dual Network (HEDN), a lightweight reliability-aware MSDA framework. HEDN introduces a novel Source Reliability Assessment (SRA) mechanism that dynamically evaluates the structural integrity of each source domain during training. Based on this assessment, sources are routed to two specialized branches: an Easy Network that exploits high-quality sources to construct fine-grained, structure-aware prototypes for reliable pseudo-label generation, and a Hard Network that utilizes adversarial training to refine and align low-quality sources. Furthermore, a cross-network consistency loss aligns predictions between branches to preserve semantic coherence. Extensive experiments conducted on SEED, SEED-IV, and DEAP datasets demonstrate that HEDN achieves state-of-the-art performance across both cross-subject and cross-dataset evaluation protocols while reducing adaptation complexity.
Authors: Chieh-Yun Chen, Zhonghao Wang, Qi Chen, Zhifan Ye, Min Shi, Yue Zhao, Yinan Zhao, Hui Qu, Wei-An Lin, Yiru Shen, Ajinkya Kale, Irfan Essa, Humphrey Shi
Abstract: Reinforcement learning from human feedback (RLHF) with reward models has advanced alignment of generative models to human aesthetic and perceptual preferences. However, jointly optimizing multiple rewards often incurs an alignment tax, improving one dimension while degrading others. To address this, we introduce two complementary methods: MapReduce LoRA and Reward-aware Token Embedding (RaTE). MapReduce LoRA trains preference-specific LoRA experts in parallel and iteratively merges them to refine a shared base model; RaTE learns reward-specific token embeddings that compose at inference for flexible preference control. Experiments on Text-to-Image generation (Stable Diffusion 3.5 Medium and FLUX.1-dev) show improvements of 36.1%, 4.6%, and 55.7%, and 32.7%, 4.3%, and 67.1% on GenEval, PickScore, and OCR, respectively. On Text-to-Video generation (HunyuanVideo), visual and motion quality improve by 48.1% and 90.0%, respectively. On the language task, Helpful Assistant, with Llama-2 7B, helpful and harmless improve by 43.4% and 136.7%, respectively. Our framework sets a new state-of-the-art multi-preference alignment recipe across modalities.
Authors: Thomas Krug, Fabian Raisch, Dominik Aimer, Markus Wirnsberger, Ferdinand Sigg, Felix Koch, Benjamin Sch\"afer, Benjamin Tischler
Abstract: Data-driven modeling of building thermal dynamics is emerging as an increasingly important field of research for large-scale intelligent building control. However, research in data-driven modeling using machine learning (ML) techniques requires massive amounts of thermal building data, which is not easily available. Neither empirical public datasets nor existing data generators meet the needs of ML research in terms of data quality and quantity. Moreover, existing data generation approaches typically require expert knowledge in building simulation. To fill this gap, we present a thermal building data generation framework which we call BuilDa. BuilDa is designed to produce synthetic data of adequate quality and quantity for ML research. The framework does not require profound building simulation knowledge to generate large volumes of data. BuilDa uses a single-zone Modelica model that is exported as a Functional Mock-up Unit (FMU) and simulated in Python. We demonstrate BuilDa by generating data and utilizing it for a transfer learning study involving the fine-tuning of 486 data-driven models.
Authors: Qiming Bao, Xiaoxuan Fu
Abstract: Large language models (LLMs) achieve strong performance on many natural language tasks, yet their generalisation under structured perturbations of logical rule systems remains insufficiently characterised. We present a controlled evaluation framework that probes reasoning reliability through four stress tests: (1) rule deletion, removing redundant versus essential rules from a multi-step inference chain; (2) contradictory evidence injection; (3) logic-preserving rewrites based on equivalence laws (contraposition, double negation, implication-to-disjunction, De Morgan, identity, and commutativity); and (4) multi-law equivalence stacking that composes 2--5 transformations. Across three representative model families -- BERT, Qwen2, and LLaMA-like models -- all models attain Acc$=1.0000$ on the base split and show no degradation under redundant rule deletion. In contrast, essential rule deletion yields a pronounced decrease to near-chance performance, and injecting explicit contradictions reduces accuracy to 0.0000. Under logic-preserving rewrites, accuracy is largely preserved for single-law transformations with only small degradations in a few cases, whereas multi-law stacking exposes model-dependent sensitivity: BERT matches the base condition, TinyLlama shows only marginal degradation, and Qwen2 exhibits a substantial drop. Overall, the results indicate that contemporary LLMs are generally stable under semantic-preserving reformulations, yet remain brittle to missing or inconsistent evidence and may degrade under composed logical transformations depending on the model family. The proposed framework provides a concise diagnostic tool for isolating these failure modes and for evaluating logical generalisation beyond surface-form variation.
Authors: Trung-Kiet Huynh, Duy-Minh Dao-Sy, Thanh-Bang Cao, Phong-Hao Le, Hong-Dan Nguyen, Phu-Quy Nguyen-Lam, Minh-Luan Nguyen-Vo, Hong-Phat Pham, Phu-Hoa Pham, Thien-Kim Than, Chi-Nguyen Tran, Huy Tran, Gia-Thoai Tran-Le, Alessio Buscemi, Le Hong Trang, The Anh Han
Abstract: As Large Language Models (LLMs) increasingly operate as autonomous decision-makers in interactive and multi-agent systems and human societies, understanding their strategic behaviour has profound implications for safety, coordination, and the design of AI-driven social and economic infrastructures. Assessing such behaviour requires methods that capture not only what LLMs output, but the underlying intentions that guide their decisions. In this work, we extend the FAIRGAME framework to systematically evaluate LLM behaviour in repeated social dilemmas through two complementary advances: a payoff-scaled Prisoners Dilemma isolating sensitivity to incentive magnitude, and an integrated multi-agent Public Goods Game with dynamic payoffs and multi-agent histories. These environments reveal consistent behavioural signatures across models and languages, including incentive-sensitive cooperation, cross-linguistic divergence and end-game alignment toward defection. To interpret these patterns, we train traditional supervised classification models on canonical repeated-game strategies and apply them to FAIRGAME trajectories, showing that LLMs exhibit systematic, model- and language-dependent behavioural intentions, with linguistic framing at times exerting effects as strong as architectural differences. Together, these findings provide a unified methodological foundation for auditing LLMs as strategic agents and reveal systematic cooperation biases with direct implications for AI governance, collective decision-making, and the design of safe multi-agent systems.
Authors: Honglian Wang, Haoyun Chen, Aristides Gionis
Abstract: Link-analysis algorithms, such as PageRank, are instrumental in understanding the structural dynamics of networks by evaluating the importance of individual vertices based on their connectivity. Recently, with the rising importance of responsible AI, the question of fairness in link-analysis algorithms has gained traction. In this paper, we present a new approach for incorporating group fairness into the PageRank algorithm by reweighting the transition probabilities in the underlying transition matrix. We formulate the problem of achieving fair PageRank by seeking to minimize the fairness loss, which is the difference between the original group-wise PageRank distribution and a target PageRank distribution. We further define a group-adapted fairness notion, which accounts for group homophily by considering random walks with group-biased restart for each group. Since the fairness loss is non-convex, we propose an efficient projected gradient-descent method for computing locally-optimal edge weights. Unlike earlier approaches, we do not recommend adding new edges to the network, nor do we adjust the restart vector. Instead, we keep the topology of the underlying network unchanged and only modify the relative importance of existing edges. We empirically compare our approach with state-of-the-art baselines and demonstrate the efficacy of our method, where very small changes in the transition matrix lead to significant improvement in the fairness of the PageRank algorithm.
Authors: Daniele Sorini, Sownak Bose, Mathilda Denison, Romeel Dav\'e
Abstract: Stellar and AGN-driven feedback processes affect the distribution of gas on a wide range of scales, from within galaxies well into the intergalactic medium. Yet, it remains unclear how feedback, through its connection to key galaxy properties, shapes the radial gas density profile in the host halo. We tackle this question using suites of the EAGLE, IllustrisTNG, and Simba cosmological hydrodynamical simulations, which span a variety of feedback models. We develop a random forest algorithm that predicts the radial gas density profile within haloes from the total halo mass and five global properties of the central galaxy: gas and stellar mass; star formation rate; mass and accretion rate of the central black hole (BH). The algorithm reproduces the simulated gas density profiles with an average accuracy of $\sim$80-90% over the halo mass range $10^{9.5} \, \mathrm{M}_{\odot} < M_{\rm 200c} < 10^{15} \, \mathrm{M}_{\odot}$ and redshift interval $0
Authors: Uisang Lee, Changhoon Chung, Junmo Lee, Soo-Mook Moon
Abstract: The rapid growth of Ethereum has made it more important to quickly and accurately detect smart contract vulnerabilities. While machine-learning-based methods have shown some promise, many still rely on rule-based preprocessing designed by domain experts. Rule-based preprocessing methods often discard crucial context from the source code, potentially causing certain vulnerabilities to be overlooked and limiting adaptability to newly emerging threats. We introduce BugSweeper, an end-to-end deep learning framework that detects vulnerabilities directly from the source code without manual engineering. BugSweeper represents each Solidity function as a Function-Level Abstract Syntax Graph (FLAG), a novel graph that combines its Abstract Syntax Tree (AST) with enriched control-flow and data-flow semantics. Then, our two-stage Graph Neural Network (GNN) analyzes these graphs. The first-stage GNN filters noise from the syntax graphs, while the second-stage GNN conducts high-level reasoning to detect diverse vulnerabilities. Extensive experiments on real-world contracts show that BugSweeper significantly outperforms all state-of-the-art detection methods. By removing the need for handcrafted rules, our approach offers a robust, automated, and scalable solution for securing smart contracts without any dependence on security experts.
Authors: Waleed Razzaq, Izis Kanjaraway, Yun-Bo Zhao
Abstract: Attention improves representation learning over RNNs, but its discrete nature limits continuous-time (CT) modeling. We introduce Neuronal Attention Circuit (NAC), a novel, biologically plausible CT-Attention mechanism that reformulates attention logits computation as the solution to a linear first-order ODE with nonlinear interlinked gates derived from repurposing \textit{C. elegans} Neuronal Circuit Policies (NCPs) wiring mechanism. NAC replaces dense projections with sparse sensory gates for key-query projections and a sparse backbone network with two heads for computing \textit{content-target} and \textit{learnable time-constant} gates, enabling efficient adaptive dynamics. NAC supports three attention logit computation modes: (i) explicit Euler integration, (ii) exact closed-form solution, and (iii) steady-state approximation. To improve memory intensity, we implemented a sparse Top-\emph{K} pairwise concatenation scheme that selectively curates key-query interactions. We provide rigorous theoretical guarantees, including state stability, bounded approximation errors, and universal approximation. Empirically, we implemented NAC in diverse domains, including irregular time-series classification, lane-keeping for autonomous vehicles, and industrial prognostics. We observed that NAC matches or outperforms competing baselines in accuracy and occupies an intermediate position in runtime and memory efficiency compared with several CT baselines.
Authors: Zhaodong Wang, Zhenting Qi, Sherman Wong, Nathan Hu, Samuel Lin, Jun Ge, Erwin Gao, Yining Yang, Ben Maurer, Wenlin Chen, David Recordon, Yilun Du, Minlan Yu, Ying Zhang
Abstract: Real-world AI software engineering demands coding agents that can reason over massive repositories, maintain durable memory across and within long sessions, and robustly coordinate complex toolchains at test time. Existing open-source coding agents provide transparency but frequently fall short when pushed to these industrial-scale workloads, while proprietary coding agents offer strong practical performance but limited extensibility, interpretability, and controllability. We present the Confucius Code Agent (CCA), an open-sourced AI software engineer that can operate at an industrial scale. CCA is built atop the Confucius SDK, an open-sourced agent development platform designed around three complementary perspectives: Agent Experience (AX), User Experience (UX), and Developer Experience (DX). The SDK introduces a unified orchestrator with hierarchical working memory for long-context reasoning, a persistent note-taking system for cross-session continual learning, and a modular extension module for robust tool use. Moreover, a meta-agent automates the synthesis, evaluation, and refinement of agent configurations through a build-test-improve loop, enabling rapid agent development on new tasks, environments, and tool stacks. Instantiated on Confucius SDK with these mechanisms, CCA delivers strong performance on real-world software engineering tasks. On SWE-Bench-Pro, CCA achieves a state-of-the-art Resolve@1 performance of 54.3%, substantially improving over prior coding agents. Together, the Confucius SDK and CCA provide a transparent, extensible, and reproducible foundation for AI agents, bridge gaps between research prototypes and production-grade systems, and support agent development and deployment at industrial scale.
Authors: Anh-Quan Pham, Marcel Hussing, Shubhankar P. Patankar, Dani S. Bassett, Jorge Mendez-Mendez, Eric Eaton
Abstract: Collecting robotic manipulation data is expensive, making it impractical to acquire demonstrations for the combinatorially large space of tasks that arise in multi-object, multi-robot, and multi-environment settings. While recent generative models can synthesize useful data for individual tasks, they do not exploit the compositional structure of robotic domains and struggle to generalize to unseen task combinations. We propose a semantic compositional diffusion transformer that factorizes transitions into robot-, object-, obstacle-, and objective-specific components and learns their interactions through attention. Once trained on a limited subset of tasks, we show that our model can zero-shot generate high-quality transitions from which we can learn control policies for unseen task combinations. Then, we introduce an iterative self-improvement procedure in which synthetic data is validated via offline reinforcement learning and incorporated into subsequent training rounds. Our approach substantially improves zero-shot performance over monolithic and hard-coded compositional baselines, ultimately solving nearly all held-out tasks and demonstrating the emergence of meaningful compositional structure in the learned representations.