# Research Papers — 2026-03 · bundle page 151 of 243 > Full-text markdown of papers #7501–7550 of 12123 published in 2026-03, newest first. AI-readable mirrors of the full 12123-paper 2026-03 archive (50 papers per file). Canonical page: https://aipod.dev/papers/archive/2026-03/page/151 [← Previous bundle](https://aipod.dev/papers/archive/2026-03/page/150.md) [Next bundle →](https://aipod.dev/papers/archive/2026-03/page/152.md) ### Frequentist Consistency of Prior-Data Fitted Networks for Causal Inference Authors: Valentyn Melnychuk, Vahid Balazadeh, Stefan Feuerriegel, Rahul G. Krishnan Published: 2026-03-12 Categories: cs.LG arXiv: https://arxiv.org/abs/2603.12037v3 Foundation models based on prior-data fitted networks (PFNs) have shown strong empirical performance in causal inference by framing the task as an in-context learning problem. However, it is unclear whether PFN-based causal estimators provide uncertainty quantification that is consistent with classical frequentist estimators. In this work, we address this gap by analyzing the frequentist consistency of PFN-based estimators for the average treatment effect (ATE). (1) We show that existing PFNs, when interpreted as Bayesian ATE estimators, can exhibit prior-induced confounding bias: the prior is not asymptotically overwritten by data, which, in turn, prevents frequentist consistency. (2) As a remedy, we suggest employing a calibration procedure based on a one-step posterior correction (OSPC). We show that the OSPC helps to restore frequentist consistency and can yield a semi-parametric Bernstein-von Mises theorem for calibrated PFNs (i.e., both the calibrated PFN-based estimators and the classical semi-parametric efficient estimators converge in distribution with growing data size). (3) Finally, we implement OSPC through tailoring martingale posteriors on top of the PFNs. In this way, we are able to recover functional nuisance posteriors from PFNs, required by the OSPC. In multiple (semi-)synthetic experiments, PFNs calibrated with our martingale posterior OSPC produce ATE uncertainty that (i) asymptotically matches frequentist uncertainty and (ii) is well calibrated in finite samples in comparison to other Bayesian ATE estimators. --- ### Single Pixel Image Classification using an Ultrafast Digital Light Projector Authors: Aisha Kanwal, Graeme E. Johnstone, Fahimeh Dehkhoda, Johannes H. Herrnsdorf, Robert K. Henderson, Martin D. Dawson, Xavier Porte, Michael J. Strain Published: 2026-03-12 Categories: cs.CV, physics.optics arXiv: https://arxiv.org/abs/2603.12036v1 Pattern recognition and image classification are essential tasks in machine vision. Autonomous vehicles, for example, require being able to collect the complex information contained in a changing environment and classify it in real time. Here, we experimentally demonstrate image classification at multi-kHz frame rates combining the technique of single pixel imaging (SPI) with a low complexity machine learning model. The use of a microLED-on-CMOS digital light projector for SPI enables ultrafast pattern generation for sub-ms image encoding. We investigate the classification accuracy of our experimental system against the broadly accepted benchmarking task of the MNIST digits classification. We compare the classification performance of two machine learning models: An extreme learning machine (ELM) and a backpropagation trained deep neural network. The complexity of both models is kept low so the overhead added to the inference time is comparable to the image generation time. Crucially, our single pixel image classification approach is based on a spatiotemporal transformation of the information, entirely bypassing the need for image reconstruction. By exploring the performance of our SPI based ELM as binary classifier we demonstrate its potential for efficient anomaly detection in ultrafast imaging scenarios. --- ### Agentic-Kube: A Graph-Enhanced Multi-Agent Reinforcement Learning Framework for Multi-Objective Kubernetes Scheduling Authors: Hamed Hamzeh Published: 2026-03-12 Categories: cs.DC, cs.LG, cs.MA arXiv: https://arxiv.org/abs/2603.12031v4 Cloud-native container orchestration requires resource schedulers capable of balancing infrastructure expenditure, fault resilience, and node utilisation. Conventional reinforcement learning approaches typically rely on monolithic single-agent models that suffer from gradient interference and reward dilution when mapping conflicting operational goals into a single scalar reward. We present Agentic-Kube, a cooperative multi-agent reinforcement learning framework designed for real-time Kubernetes pod placement. The architecture decomposes multi-objective scheduling into a tripartite optimisation space managed by dedicated sub-agents for cost minimisation, anti-affinity fault tolerance, and vector resource balancing. Agentic-Kube integrates a bipartite Graph Convolutional Network to capture dynamic host-pod dependencies, a two-stage monotonic QMIX value factorisation network to maintain joint action value coherence, and a plurality voting consensus mechanism with action feasibility masking against allocatable node predicates. We evaluate the framework across live heterogeneous Google Kubernetes Engine deployments and macro-scale cluster environments spanning 50 to 1,000 nodes under empirical Alibaba trace data, diurnal microservice variations, and flash-crowd bursts. Across physical and simulated evaluations, Agentic-Kube consistently achieves Pareto-efficient placements. In diurnal microservice workloads, it reduces anti-affinity service collisions to 7.11%, representing a 53.0% relative reduction compared to the default Kubernetes scheduler. Under Alibaba traces, the policy achieves a 65.15% spot instance allocation ratio, while macro-scale benchmarks demonstrate scaling up to 1,000 nodes with mean decision latencies under 17ms and 99th-percentile latencies under 31ms, executing without container restart failures and operating well within standard scheduling admission timeouts. --- ### AGMARL-DKS: An Adaptive Graph-Enhanced Multi-Agent Reinforcement Learning for Dynamic Kubernetes Scheduling Authors: Hamed Hamzeh Published: 2026-03-12 Categories: cs.DC, cs.LG, cs.MA arXiv: https://arxiv.org/abs/2603.12031v2 State-of-the-art cloud-native applications require intelligent schedulers that can effectively balance system stability, resource utilisation, and associated costs. While Kubernetes provides feasibility-based placement by default, recent research efforts have explored the use of reinforcement learning (RL) for more intelligent scheduling decisions. However, current RL-based schedulers have three major limitations. First, most of these schedulers use monolithic centralised agents, which are non-scalable for large heterogeneous clusters. Second, the ones that use multi-objective reward functions assume simple, static, linear combinations of the objectives. Third, no previous work has produced a stress-aware scheduler that can react adaptively to dynamic conditions. To address these gaps in current research, we propose the Adaptive Graph-enhanced Multi-Agent Reinforcement Learning Dynamic Kubernetes Scheduler (AGMARL-DKS). AGMARL-DKS addresses these gaps by introducing three major innovations. First, we construct a scalable solution by treating the scheduling challenge as a cooperative multi-agent problem, where every cluster node operates as an agent, employing centralised training methods before decentralised execution. Second, to be context-aware and yet decentralised, we use a Graph Neural Network (GNN) to build a state representation of the global cluster context at each agent. This represents an improvement over methods that rely solely on local observations. Finally, to make trade-offs between these objectives, we use a stress-aware lexicographical ordering policy instead of a simple, static linear weighting of these objectives. The evaluations in Google Kubernetes Engine (GKE) reveal that AGMARL-DKS significantly outperforms the default scheduler in terms of fault tolerance, utilisation, and cost, especially in scheduling batch and mission-critical workloads. --- ### Efficient Generative Modeling with Unitary Matrix Product States Using Riemannian Optimization Authors: Haotong Duan, Zhongming Chen, Ngai Wong Published: 2026-03-12 Categories: cs.LG arXiv: https://arxiv.org/abs/2603.12026v1 Tensor networks, which are originally developed for characterizing complex quantum many-body systems, have recently emerged as a powerful framework for capturing high-dimensional probability distributions with strong physical interpretability. This paper systematically studies matrix product states (MPS) for generative modeling and shows that unitary MPS, which is a tensor-network architecture that is both simple and expressive, offers clear benefits for unsupervised learning by reducing ambiguity in parameter updates and improving efficiency. To overcome the inefficiency of standard gradient-based MPS training, we develop a Riemannian optimization approach that casts probabilistic modeling as an optimization problem with manifold constraints, and further derive an efficient space-decoupling algorithm. Experiments on Bars-and-Stripes and EMNIST datasets demonstrate fast adaptation to data structure, stable updates, and strong performance while maintaining the efficiency and expressive power of MPS. --- ### Cascade: Composing Software-Hardware Attack Gadgets for Adversarial Threat Amplification in Compound AI Systems Authors: Sarbartha Banerjee, Prateek Sahu, Anjo Vahldiek-Oberwagner, Jose Sanchez Vicarte, Mohit Tiwari Published: 2026-03-12 Categories: cs.CR, cs.AI arXiv: https://arxiv.org/abs/2603.12023v1 Rapid progress in generative AI has given rise to Compound AI systems - pipelines comprised of multiple large language models (LLM), software tools and database systems. Compound AI systems are constructed on a layered traditional software stack running on a distributed hardware infrastructure. Many of the diverse software components are vulnerable to traditional security flaws documented in the Common Vulnerabilities and Exposures (CVE) database, while the underlying distributed hardware infrastructure remains exposed to timing attacks, bit-flip faults, and power-based side channels. Today, research targets LLM-specific risks like model extraction, training data leakage, and unsafe generation -- overlooking the impact of traditional system vulnerabilities. This work investigates how traditional software and hardware vulnerabilities can complement LLM-specific algorithmic attacks to compromise the integrity of a compound AI pipeline. We demonstrate two novel attacks that combine system-level vulnerabilities with algorithmic weaknesses: (1) Exploiting a software code injection flaw along with a guardrail Rowhammer attack to inject an unaltered jailbreak prompt into an LLM, resulting in an AI safety violation, and (2) Manipulating a knowledge database to redirect an LLM agent to transmit sensitive user data to a malicious application, thus breaching confidentiality. These attacks highlight the need to address traditional vulnerabilities; we systematize the attack primitives and analyze their composition by grouping vulnerabilities by their objective and mapping them to distinct stages of an attack lifecycle. This approach enables a rigorous red-teaming exercise and lays the groundwork for future defense strategies. --- ### Just Use XML: Revisiting Joint Translation and Label Projection Authors: Thennal DK, Chris Biemann, Hans Ole Hatzel Published: 2026-03-12 Categories: cs.CL, cs.AI arXiv: https://arxiv.org/abs/2603.12021v2 Label projection is an effective technique for cross-lingual transfer, extending span-annotated datasets from a high-resource language to low-resource ones. Most approaches perform label projection as a separate step after machine translation, and prior work that combines the two reports degraded translation quality. We re-evaluate this claim with LabelPigeon, a novel framework that jointly performs translation and label projection via XML tags. We design a direct evaluation scheme for label projection, and find that LabelPigeon outperforms baselines and actively improves translation quality in 11 languages. We further assess translation quality across 203 languages and varying annotation complexity, finding consistent improvement attributed to additional fine-tuning. Finally, across 27 languages and three downstream tasks, we report substantial gains in cross-lingual transfer over comparable work, up to +40.2 F1 on NER. Overall, our results demonstrate that XML-tagged label projection provides effective and efficient label transfer without compromising translation quality. --- ### Sim-to-reality adaptation for Deep Reinforcement Learning applied to an underwater docking application Authors: Alaaeddine Chaarani, Narcis Palomeras, Pere Ridao Published: 2026-03-12 Categories: cs.RO, cs.AI arXiv: https://arxiv.org/abs/2603.12020v1 Deep Reinforcement Learning (DRL) offers a robust alternative to traditional control methods for autonomous underwater docking, particularly in adapting to unpredictable environmental conditions. However, bridging the "sim-to-real" gap and managing high training latencies remain significant bottlenecks for practical deployment. This paper presents a systematic approach for autonomous docking using the Girona Autonomous Underwater Vehicle (AUV) by leveraging a high-fidelity digital twin environment. We adapted the Stonefish simulator into a multiprocessing RL framework to significantly accelerate the learning process while incorporating realistic AUV dynamics, collision models, and sensor noise. Using the Proximal Policy Optimization (PPO) algorithm, we developed a 6-DoF control policy trained in a headless environment with randomized starting positions to ensure generalized performance. Our reward structure accounts for distance, orientation, action smoothness, and adaptive collision penalties to facilitate soft docking. Experimental results demonstrate that the agent achieved a success rate of over 90% in simulation. Furthermore, successful validation in a physical test tank confirmed the efficacy of the sim-to-reality adaptation, with the DRL controller exhibiting emergent behaviors such as pitch-based braking and yaw oscillations to assist in mechanical alignment. --- ### An Intent of Collaboration: On Agencies between Designers and Emerging (Intelligent) Technologies Authors: Pei-Ying Lin, Julie Heij, Iris Borst, Britt Joosten, Kristina Andersen, Wijnand IJsselsteijn Published: 2026-03-12 Categories: cs.HC, cs.AI, cs.ET arXiv: https://arxiv.org/abs/2603.12018v1 Amidst the emergence of powerful intelligent technologies such as LLMs and text-to-image AIs that promise to enhance creative processes, designers face the challenges of remaining empowered and creative while working with these foreign digital partners. While generative AIs offer versatile, informative, and occasionally poetic outcomes, their lack of embodied knowledge presents an even greater challenge to designers in gaining fruitful outcomes, such as in the field of Digital Craftsmanship. In this project, three designers embarked on a three-month experimental journey with an intention to co-create with Google's LLM as a potential intelligent partner to investigate how it will influence the designers' creativity. We found that a power dynamic of agencies exists between the LLM and the designer, in which the designer can easily lose their creative agency. Regaining the designer's creative agency involves introspection into their own creative process, a structural understanding of the specific emerging technology involved, and deliberate adjustments to the dynamics of the human-technology relationship. We propose paying attention to the designer's inner world and parties of agencies when engaging with emerging intelligent technologies through three aspects: the sensitivity towards a creative process as cognitive activities; the active investigation into specific technology's capability; and the adjustment towards an appropriate working relationship between the designer and the emerging technology. --- ### Nyxus: A Next Generation Image Feature Extraction Library for the Big Data and AI Era Authors: Nicholas Schaub, Andriy Kharchenko, Hamdah Abbasi, Sameeul Samee, Hythem Sidky, Nathan Hotaling Published: 2026-03-12 Categories: cs.CV, q-bio.QM arXiv: https://arxiv.org/abs/2603.12016v1 Modern imaging instruments can produce terabytes to petabytes of data for a single experiment. The biggest barrier to processing big image datasets has been computational, where image analysis algorithms often lack the efficiency needed to process such large datasets or make tradeoffs in robustness and accuracy. Deep learning algorithms have vastly improved the accuracy of the first step in an analysis workflow (region segmentation), but the expansion of domain specific feature extraction libraries across scientific disciplines has made it difficult to compare the performance and accuracy of extracted features. To address these needs, we developed a novel feature extraction library called Nyxus. Nyxus is designed from the ground up for scalable out-of-core feature extraction for 2D and 3D image data and rigorously tested against established standards. The comprehensive feature set of Nyxus covers multiple biomedical domains including radiomics and cellular analysis, and is designed for computational scalability across CPUs and GPUs. Nyxus has been packaged to be accessible to users of various skill sets and needs: as a Python package for code developers, a command line tool, as a Napari plugin for low to no-code users or users that want to visualize results, and as an Open Container Initiative (OCI) compliant container that can be used in cloud or super-computing workflows aimed at processing large data sets. Further, Nyxus enables a new methodological approach to feature extraction allowing for programmatic tuning of many features sets for optimal computational efficiency or coverage for use in novel machine learning and deep learning applications. --- ### Flowcean - Model Learning for Cyber-Physical Systems Authors: Maximilian Schmidt, Swantje Plambeck, Markus Knitt, Hendrik Rose, Goerschwin Fey, Jan Christian Wieck, Stephan Balduin Published: 2026-03-12 Categories: cs.LG, cs.AI arXiv: https://arxiv.org/abs/2603.12015v1 Effective models of Cyber-Physical Systems (CPS) are crucial for their design and operation. Constructing such models is difficult and time-consuming due to the inherent complexity of CPS. As a result, data-driven model generation using machine learning methods is gaining popularity. In this paper, we present Flowcean, a novel framework designed to automate the generation of models through data-driven learning that focuses on modularity and usability. By offering various learning strategies, data processing methods, and evaluation metrics, our framework provides a comprehensive solution, tailored to CPS scenarios. Flowcean facilitates the integration of diverse learning libraries and tools within a modular and flexible architecture, ensuring adaptability to a wide range of modeling tasks. This streamlines the process of model generation and evaluation, making it more efficient and accessible. --- ### Pano360: Perspective to Panoramic Vision with Geometric Consistency Authors: Zhengdong Zhu, Weiyi Xue, Zuyuan Yang, Wenlve Zhou, Zhiheng Zhou Published: 2026-03-12 Categories: cs.CV arXiv: https://arxiv.org/abs/2603.12013v1 Prior panorama stitching approaches heavily rely on pairwise feature correspondences and are unable to leverage geometric consistency across multiple views. This leads to severe distortion and misalignment, especially in challenging scenes with weak textures, large parallax, and repetitive patterns. Given that multi-view geometric correspondences can be directly constructed in 3D space, making them more accurate and globally consistent, we extend the 2D alignment task to the 3D photogrammetric space. We adopt a novel transformer-based architecture to achieve 3D awareness and aggregate global information across all views. It directly utilizes camera poses to guide image warping for global alignment in 3D space and employs a multi-feature joint optimization strategy to compute the seams. Additionally, to establish an evaluation benchmark and train our network, we constructed a large-scale dataset of real-world scenes. Extensive experiments show that our method significantly outperforms existing alternatives in alignment accuracy and perceptual quality. --- ### Deep Learning-Based Metamodeling of Nonlinear Stochastic Dynamic Systems under Parametric and Predictive Uncertainty Authors: Haimiti Atila, Seymour M. J. Spence Published: 2026-03-12 Categories: cs.LG arXiv: https://arxiv.org/abs/2603.12012v1 Modeling high-dimensional, nonlinear dynamic structural systems under natural hazards presents formidable computational challenges, especially when simultaneously accounting for uncertainties in external loads and structural parameters. Studies have successfully incorporated uncertainties related to external loads from natural hazards, but few have simultaneously addressed loading and parameter uncertainties within structural systems while accounting for prediction uncertainty of neural networks. To address these gaps, three metamodeling frameworks were formulated, each coupling a feature-extraction module implemented through a multi-layer perceptron (MLP), a message-passing neural network (MPNN), or an autoencoder (AE) with a long short-term memory (LSTM) network using Monte Carlo dropout and a negative log-likelihood loss. The resulting architectures (MLP-LSTM, MPNN-LSTM, and AE-LSTM) were validated on two case studies: a multi-degree-of-freedom Bouc-Wen system and a 37-story fiber-discretized nonlinear steel moment-resisting frame, both subjected to stochastic seismic excitation and structural parameter uncertainty. All three approaches achieved low prediction errors: the MLP-LSTM yielded the most accurate results for the lower-dimensional Bouc-Wen system, whereas the MPNN-LSTM and AE-LSTM provided superior performance on the more complex steel-frame model. Moreover, a consistent correlation between predictive variance and actual error confirms the suitability of these frameworks for active-learning strategies and for assessing model confidence in structural response predictions. --- ### Can RL Improve Generalization of LLM Agents? An Empirical Study Authors: Zhiheng Xi, Xin Guo, Jiaqi Liu, Jiazheng Zhang, Yutao Fan, Zhihao Zhang, Shichun Liu, Mingxu Chai, Xiaowei Shi, Yitao Zhai, Xunliang Cai, Tao Gui et al. Published: 2026-03-12 Categories: cs.AI arXiv: https://arxiv.org/abs/2603.12011v1 Reinforcement fine-tuning (RFT) has shown promise for training LLM agents to perform multi-turn decision-making based on environment feedback. However, most existing evaluations remain largely in-domain: training and testing are conducted in the same environment or even on the same tasks. In real-world deployment, agents may operate in unseen environments with different background knowledge, observation spaces, and action interfaces. To characterize the generalization profile of RFT under such shifts, we conduct a systematic study along three axes: (1) within-environment generalization across task difficulty, (2) cross-environment transfer to unseen environments, and (3) sequential multi-environment training to quantify transfer and forgetting. Our results show that RFT generalizes well across task difficulty within an environment, but exhibits weaker transfer to unseen environments, which correlates with shifts in both semantic priors and observation/action interfaces. In contrast, sequential training yields promising downstream gains with minimal upstream forgetting, and mixture training across environments improves the overall balance. We further provide detailed analyses and deeper insights, and hope our work helps the community develop and deploy generalizable LLM agents. --- ### CrossEarth-SAR: A SAR-Centric and Billion-Scale Geospatial Foundation Model for Domain Generalizable Semantic Segmentation Authors: Ziqi Ye, Ziyang Gong, Ning Liao, Xiaoxing Hu, Di Wang, Hongruixuan Chen, Chen Huang, Yiguo He, Yuru Jia, Xiaoxing Wang, Haipeng Wang, Xue Yang et al. Published: 2026-03-12 Categories: cs.CV arXiv: https://arxiv.org/abs/2603.12008v1 Synthetic Aperture Radar (SAR) enables global, all-weather earth observation. However, owing to diverse imaging mechanisms, domain shifts across sensors and regions severely hinder its semantic generalization. To address this, we present CrossEarth-SAR, the first billion-scale SAR vision foundation model built upon a novel physics-guided sparse mixture-of-experts (MoE) architecture incorporating physical descriptors, explicitly designed for cross-domain semantic segmentation. To facilitate large-scale pre-training, we develop CrossEarth-SAR-200K, a weakly and fully supervised dataset that unifies public and private SAR imagery. We also introduce a benchmark suite comprising 22 sub-benchmarks across 8 distinct domain gaps, establishing the first unified standard for domain generalization semantic segmentation on SAR imagery. Extensive experiments demonstrate that CrossEarth-SAR achieves state-of-the-art results on 20 benchmarks, surpassing previous methods by over 10\% mIoU on some benchmarks under multi-gap transfer. All code, benchmark and datasets will be publicly available. --- ### Decentralized Orchestration Architecture for Fluid Computing: A Secure Distributed AI Use Case Authors: Diego Cajaraville-Aboy, Ana Fernández-Vilas, Rebeca P. Díaz-Redondo, Manuel Fernández-Veiga, Pablo Picallo-López Published: 2026-03-12 Categories: cs.DC, cs.LG arXiv: https://arxiv.org/abs/2603.12001v2 Distributed AI and IoT applications increasingly execute across heterogeneous resources spanning end devices, edge/fog infrastructure, and cloud platforms, often under different administrative domains. Fluid Computing has emerged as a promising paradigm for enhancing massive resource management across the computing continuum by treating such resources as a unified fabric, enabling optimal service-agnostic deployments driven by application requirements. However, existing solutions remain largely centralized and often do not explicitly address multi-domain considerations. This paper proposes an agnostic multi-domain orchestration architecture for fluid computing environments. The orchestration plane enables decentralized coordination among domains that maintain local autonomy while jointly realizing intent-based deployment requests from tenants, ensuring end-to-end placement and execution. To this end, the architecture elevates domain-side control services as first-class capabilities to support application-level enhancement at runtime. As a representative proof of concept, we instantiate the architecture through a distributed AI use case; specifically, we consider a multi-domain Decentralized Federated Learning (DFL) deployment under Byzantine threats. Under this setting, we leverage domain-side capabilities to enhance Byzantine security by introducing FU-HST, an SDN-enabled multi-domain anomaly detection mechanism that complements Byzantine-robust aggregation. We validate the use-case workflow via simulation in single- and multi-domain settings, evaluating anomaly detection, DFL performance, and computation/communication overhead. --- ### Few-for-Many Personalized Federated Learning Authors: Ping Guo, Tiantian Zhang, Xi Lin, Xiang Li, Zhi-Ri Tang, Qingfu Zhang Published: 2026-03-12 Categories: cs.AI, cs.LG arXiv: https://arxiv.org/abs/2603.11992v1 Personalized Federated Learning (PFL) aims to train customized models for clients with highly heterogeneous data distributions while preserving data privacy. Existing approaches often rely on heuristics like clustering or model interpolation, which lack principled mechanisms for balancing heterogeneous client objectives. Serving $M$ clients with distinct data distributions is inherently a multi-objective optimization problem, where achieving optimal personalization ideally requires $M$ distinct models on the Pareto front. However, maintaining $M$ separate models poses significant scalability challenges in federated settings with hundreds or thousands of clients. To address this challenge, we reformulate PFL as a few-for-many optimization problem that maintains only $K$ shared server models ($K \ll M$) to collectively serve all $M$ clients. We prove that this framework achieves near-optimal personalization: the approximation error diminishes as $K$ increases and each client's model converges to each client's optimum as data grows. Building on this reformulation, we propose FedFew, a practical algorithm that jointly optimizes the $K$ server models through efficient gradient-based updates. Unlike clustering-based approaches that require manual client partitioning or interpolation-based methods that demand careful hyperparameter tuning, FedFew automatically discovers the optimal model diversity through its optimization process. Experiments across vision, NLP, and real-world medical imaging datasets demonstrate that FedFew, with just 3 models, consistently outperforms other state-of-the-art approaches. Code is available at https://github.com/pgg3/FedFew. --- ### BTZSC: A Benchmark for Zero-Shot Text Classification Across Cross-Encoders, Embedding Models, Rerankers and LLMs Authors: Ilias Aarab Published: 2026-03-12 Categories: cs.CL, cs.AI, cs.LG, stat.ML arXiv: https://arxiv.org/abs/2603.11991v1 Zero-shot text classification (ZSC) offers the promise of eliminating costly task-specific annotation by matching texts directly to human-readable label descriptions. While early approaches have predominantly relied on cross-encoder models fine-tuned for natural language inference (NLI), recent advances in text-embedding models, rerankers, and instruction-tuned large language models (LLMs) have challenged the dominance of NLI-based architectures. Yet, systematically comparing these diverse approaches remains difficult. Existing evaluations, such as MTEB, often incorporate labeled examples through supervised probes or fine-tuning, leaving genuine zero-shot capabilities underexplored. To address this, we introduce BTZSC, a comprehensive benchmark of 22 public datasets spanning sentiment, topic, intent, and emotion classification, capturing diverse domains, class cardinalities, and document lengths. Leveraging BTZSC, we conduct a systematic comparison across four major model families, NLI cross-encoders, embedding models, rerankers and instruction-tuned LLMs, encompassing 38 public and custom checkpoints. Our results show that: (i) modern rerankers, exemplified by Qwen3-Reranker-8B, set a new state-of-the-art with macro F1 = 0.72; (ii) strong embedding models such as GTE-large-en-v1.5 substantially close the accuracy gap while offering the best trade-off between accuracy and latency; (iii) instruction-tuned LLMs at 4--12B parameters achieve competitive performance (macro F1 up to 0.67), excelling particularly on topic classification but trailing specialized rerankers; (iv) NLI cross-encoders plateau even as backbone size increases; and (v) scaling primarily benefits rerankers and LLMs over embedding models. BTZSC and accompanying evaluation code are publicly released to support fair and reproducible progress in zero-shot text understanding. --- ### On-Average Stability of Multipass Preconditioned SGD and Effective Dimension Authors: Simon Vary, Tyler Farghly, Ilja Kuzborskij, Patrick Rebeschini Published: 2026-03-12 Categories: cs.LG, math.OC, stat.ML arXiv: https://arxiv.org/abs/2603.11989v1 We study trade-offs between the population risk curvature, geometry of the noise, and preconditioning on the generalisation ability of the multipass Preconditioned Stochastic Gradient Descent (PSGD). Many practical optimisation heuristics implicitly navigate this trade-off in different ways -- for instance, some aim to whiten gradient noise, while others aim to align updates with expected loss curvature. When the geometry of the population risk curvature and the geometry of the gradient noise do not match, an aggressive choice that improves one aspect can amplify instability along the other, leading to suboptimal statistical behavior. In this paper we employ on-average algorithmic stability to connect generalisation of PSGD to the effective dimension that depends on these sources of curvature. While existing techniques for on-average stability of SGD are limited to a single pass, as first contribution we develop a new on-average stability analysis for multipass SGD that handles the correlations induced by data reuse. This allows us to derive excess risk bounds that depend on the effective dimension. In particular, we show that an improperly chosen preconditioner can yield suboptimal effective dimension dependence in both optimisation and generalisation. Finally, we complement our upper bounds with matching, instance-dependent lower bounds. --- ### LABSHIELD: A Multimodal Benchmark for Safety-Critical Reasoning and Planning in Scientific Laboratories Authors: Qianpu Sun, Xiaowei Chi, Yuhan Rui, Ying Li, Kuangzhi Ge, Jiajun Li, Sirui Han, Shanghang Zhang Published: 2026-03-12 Categories: cs.AI arXiv: https://arxiv.org/abs/2603.11987v1 Artificial intelligence is increasingly catalyzing scientific automation, with multimodal large language model (MLLM) agents evolving from lab assistants into self-driving lab operators. This transition imposes stringent safety requirements on laboratory environments, where fragile glassware, hazardous substances, and high-precision laboratory equipment render planning errors or misinterpreted risks potentially irreversible. However, the safety awareness and decision-making reliability of embodied agents in such high-stakes settings remain insufficiently defined and evaluated. To bridge this gap, we introduce LABSHIELD, a realistic multi-view benchmark designed to assess MLLMs in hazard identification and safety-critical reasoning. Grounded in U.S. Occupational Safety and Health Administration (OSHA) standards and the Globally Harmonized System (GHS), LABSHIELD establishes a rigorous safety taxonomy spanning 164 operational tasks with diverse manipulation complexities and risk profiles. We evaluate 20 proprietary models, 9 open-source models, and 3 embodied models under a dual-track evaluation framework. Our results reveal a systematic gap between general-domain MCQ accuracy and Semi-open QA safety performance, with models exhibiting an average drop of 32.0% in professional laboratory scenarios, particularly in hazard interpretation and safety-aware planning. These findings underscore the urgent necessity for safety-centric reasoning frameworks to ensure reliable autonomous scientific experimentation in embodied laboratory contexts. The full dataset will be released soon. --- ### Ada3Drift: Adaptive Training-Time Drifting for One-Step 3D Visuomotor Robotic Manipulation Authors: Chongyang Xu, Yixian Zou, Tianyu Yang, Fanman Meng, Ziliang Feng, Li Lu, Shuaicheng Liu Published: 2026-03-12 Categories: cs.CV arXiv: https://arxiv.org/abs/2603.11984v2 Diffusion-based visuomotor policies model complex action distributions through iterative denoising, but repeated inference adds latency to robotic control. One-step generators reduce this cost, motivating training objectives that retain useful action structure with few demonstrations. We present Ada3Drift, a point-cloud-conditioned policy that builds on Drifting Models to perform distribution refinement during training and generate action chunks in one forward pass. Our central design is a regression-to-drifting curriculum: paired action regression first emphasizes observation--action correspondence, while a sigmoid schedule progressively increases a batch-level action-distribution regularizer. The drifting term combines attraction to demonstrated actions and repulsion among generated samples using inherited multi-temperature aggregation. A timestep-free generator preserves single-step inference throughout. On Adroit, Meta-World, RoboTwin, and five real-world tasks, Ada3Drift achieves the highest reported average success rates among the evaluated baselines with 1 NFE, compared with 10 NFE for the diffusion baselines. Controlled ablations favor sigmoid scheduling at matched demonstration budgets. We will release our code and pretrained model weights. --- ### Ada3Drift: Adaptive Training-Time Drifting for One-Step 3D Visuomotor Robotic Manipulation Authors: Chongyang Xu, Yixian Zou, Ziliang Feng, Fanman Meng, Shuaicheng Liu Published: 2026-03-12 Categories: cs.CV arXiv: https://arxiv.org/abs/2603.11984v1 Diffusion-based visuomotor policies effectively capture multimodal action distributions through iterative denoising, but their high inference latency limits real-time robotic control. Recent flow matching and consistency-based methods achieve single-step generation, yet sacrifice the ability to preserve distinct action modes, collapsing multimodal behaviors into averaged, often physically infeasible trajectories. We observe that the compute budget asymmetry in robotics (offline training vs.\ real-time inference) naturally motivates recovering this multimodal fidelity by shifting iterative refinement from inference time to training time. Building on this insight, we propose Ada3Drift, which learns a training-time drifting field that attracts predicted actions toward expert demonstration modes while repelling them from other generated samples, enabling high-fidelity single-step generation (1 NFE) from 3D point cloud observations. To handle the few-shot robotic regime, Ada3Drift further introduces a sigmoid-scheduled loss transition from coarse distribution learning to mode-sharpening refinement, and multi-scale field aggregation that captures action modes at varying spatial granularities. Experiments on three simulation benchmarks (Adroit, Meta-World, and RoboTwin) and real-world robotic manipulation tasks demonstrate that Ada3Drift achieves state-of-the-art performance while requiring $10\times$ fewer function evaluations than diffusion-based alternatives. --- ### HomeSafe-Bench: Evaluating Vision-Language Models on Unsafe Action Detection for Embodied Agents in Household Scenarios Authors: Jiayue Pu, Zhongxiang Sun, Zilu Zhang, Xiao Zhang, Jun Xu Published: 2026-03-12 Categories: cs.CV, cs.AI, cs.CR arXiv: https://arxiv.org/abs/2603.11975v2 The rapid evolution of embodied agents has accelerated the deployment of household robots in real-world environments. However, unlike structured industrial settings, household spaces introduce unpredictable safety risks, where system limitations such as perception latency and lack of common sense knowledge can lead to dangerous errors. Current safety evaluations, often restricted to static images, text, or general hazards, fail to adequately benchmark dynamic unsafe action detection in these specific contexts. To bridge this gap, we introduce HomeSafe-Bench, a challenging benchmark designed to evaluate Vision-Language Models (VLMs) on unsafe action detection in household scenarios. HomeSafe-Bench is contrusted via a hybrid pipeline combining physical simulation with advanced video generation and features 438 diverse cases across six functional areas with fine-grained multidimensional annotations. Beyond benchmarking, we propose Hierarchical Dual-Brain Guard for Household Safety (HD-Guard), a hierarchical streaming architecture for real-time safety monitoring. HD-Guard coordinates a lightweight FastBrain for continuous high-frequency screening with an asynchronous large-scale SlowBrain for deep multimodal reasoning, effectively balancing inference efficiency with detection accuracy. Evaluations demonstrate that HD-Guard achieves a superior trade-off between latency and performance, while our analysis identifies critical bottlenecks in current VLM-based safety detection. --- ### Normative Common Ground Replication (NormCoRe): Replication-by-Translation for Studying Norms in Multi-Agent AI Authors: Luca Deck, Simeon Allmendinger, Lucas Müller, Niklas Kühl Published: 2026-03-12 Categories: cs.AI arXiv: https://arxiv.org/abs/2603.11974v2 In the late 2010s, the fashion trend NormCore framed sameness as a signal of belonging, illustrating how norms emerge through collective coordination. Today, similar forms of normative coordination can be observed in systems based on Multi-agent Artificial Intelligence (MAAI), as AI-based agents deliberate, negotiate, and converge on shared decisions in fairness-sensitive domains. Yet, existing empirical approaches often treat norms as targets for alignment or replication, implicitly assuming equivalence between human subjects and AI agents and leaving collective normative dynamics insufficiently examined. To address this gap, we propose Normative Common Ground Replication (NormCoRe), a novel methodological framework to systematically translate the design of human subject experiments into MAAI environments. Building on behavioral science, replication research, and state-of-the-art MAAI architectures, NormCoRe maps the structural layers of human subject studies onto the design of AI agent studies, enabling systematic documentation of study design and analysis of norms in MAAI. We demonstrate the utility of NormCoRe by replicating a seminal experimental study on distributive justice, in which participants negotiate fairness principles under a "veil of ignorance". We show that normative judgments in AI agent studies can differ from human baselines and are sensitive to the choice of the foundation model and the language used to instantiate agent personas. Our work provides a principled pathway for analyzing norms in MAAI and helps to guide, reflect, and document design choices whenever AI agents are used to automate or support tasks formerly carried out by humans. --- ### Topological DeepONets and a generalization of the Chen-Chen operator approximation theorem Authors: Vugar Ismailov Published: 2026-03-12 Categories: cs.LG, cs.NE, math.FA arXiv: https://arxiv.org/abs/2603.11972v1 Deep Operator Networks (DeepONets) provide a branch-trunk neural architecture for approximating nonlinear operators acting between function spaces. In the classical operator approximation framework, the input is a function $u\in C(K_1)$ defined on a compact set $K_1$ (typically a compact subset of a Banach space), and the operator maps $u$ to an output function $G(u)\in C(K_2)$ defined on a compact Euclidean domain $K_2\subset\mathbb{R}^d$. In this paper, we develop a topological extension in which the operator input lies in an arbitrary Hausdorff locally convex space $X$. We construct topological feedforward neural networks on $X$ using continuous linear functionals from the dual space $X^*$ and introduce topological DeepONets whose branch component acts on $X$ through such linear measurements, while the trunk component acts on the Euclidean output domain. Our main theorem shows that continuous operators $G:V\to C(K;\mathbb{R}^m)$, where $V\subset X$ and $K\subset\mathbb{R}^d$ are compact, can be uniformly approximated by such topological DeepONets. This extends the classical Chen-Chen operator approximation theorem from spaces of continuous functions to locally convex spaces and yields a branch-trunk approximation theorem beyond the Banach-space setting. --- ### Multimodal Emotion Recognition via Bi-directional Cross-Attention and Temporal Modeling Authors: Junhyeong Byeon, Jeongyeol Kim, Sejoon Lim Published: 2026-03-12 Categories: cs.CV, cs.AI arXiv: https://arxiv.org/abs/2603.11971v2 Expression recognition in in-the-wild video data remains challenging due to substantial variations in facial appearance, background conditions, audio noise, and the inherently dynamic nature of human affect. Relying on a single modality, such as facial expressions or speech, is often insufficient for capturing these complex emotional cues. To address this limitation, we propose a multimodal emotion recognition framework for the Expression (EXPR) task in the 10th Affective Behavior Analysis in-the-wild (ABAW) Challenge. Our framework builds on large-scale pre-trained models for visual and audio representation learning and integrates them in a unified multimodal architecture. To better capture temporal patterns in facial expression sequences, we incorporate temporal visual modeling over video windows. We further introduce a bi-directional cross-attention fusion module that enables visual and audio features to interact in a symmetric manner, facilitating cross-modal contextualization and complementary emotion understanding. In addition, we employ a text-guided contrastive objective to encourage semantically meaningful visual representations through alignment with emotion-related text prompts. Experimental results on the ABAW 10th EXPR benchmark demonstrate the effectiveness of the proposed framework, achieving a Macro F1 score of 0.32 compared to the baseline score of 0.25, and highlight the benefit of combining temporal visual modeling, audio representation learning, and cross-modal fusion for robust emotion recognition in unconstrained real-world environments. --- ### Statistical and structural identifiability in representation learning Authors: Walter Nelson, Marco Fumero, Theofanis Karaletsos, Francesco Locatello Published: 2026-03-12 Categories: cs.LG arXiv: https://arxiv.org/abs/2603.11970v1 Representation learning models exhibit a surprising stability in their internal representations. Whereas most prior work treats this stability as a single property, we formalize it as two distinct concepts: statistical identifiability (consistency of representations across runs) and structural identifiability (alignment of representations with some unobserved ground truth). Recognizing that perfect pointwise identifiability is generally unrealistic for modern representation learning models, we propose new model-agnostic definitions of statistical and structural near-identifiability of representations up to some error tolerance $ε$. Leveraging these definitions, we prove a statistical $ε$-near-identifiability result for the representations of models with nonlinear decoders, generalizing existing identifiability theory beyond last-layer representations in e.g. generative pre-trained transformers (GPTs) to near-identifiability of the intermediate representations of a broad class of models including (masked) autoencoders (MAEs) and supervised learners. Although these weaker assumptions confer weaker identifiability, we show that independent components analysis (ICA) can resolve much of the remaining linear ambiguity for this class of models, and validate and measure our near-identifiability claims empirically. With additional assumptions on the data-generating process, statistical identifiability extends to structural identifiability, yielding a simple and practical recipe for disentanglement: ICA post-processing of latent representations. On synthetic benchmarks, this approach achieves state-of-the-art disentanglement using a vanilla autoencoder. With a foundation model-scale MAE for cell microscopy, it disentangles biological variation from technical batch effects, substantially improving downstream generalization. --- ### AstroSplat: Physics-Based Gaussian Splatting for Rendering and Reconstruction of Small Celestial Bodies Authors: Jennifer Nolan, Travis Driver, John Christian Published: 2026-03-12 Categories: cs.CV arXiv: https://arxiv.org/abs/2603.11969v2 Image-based surface reconstruction and characterization are crucial for missions to small celestial bodies (e.g., asteroids), as it informs mission planning, navigation, and scientific analysis. Recent advances in Gaussian splatting enable high-fidelity neural scene representations but typically rely on a spherical harmonic intensity parameterization that is strictly appearance-based and does not explicitly model material properties or light-surface interactions. We introduce AstroSplat, a physics-based Gaussian splatting framework that integrates planetary reflectance models to improve the autonomous reconstruction and photometric characterization of small-body surfaces from in-situ imagery. The proposed framework is validated on real imagery taken by NASA's Dawn mission, where we demonstrate superior rendering performance and surface reconstruction accuracy compared to the typical spherical harmonic parameterization. --- ### Uncovering Locally Low-dimensional Structure in Networks by Locally Optimal Spectral Embedding Authors: Hannah Sansford, Nick Whiteley, Patrick Rubin-Delanchy Published: 2026-03-12 Categories: stat.ML, cs.LG, stat.ME arXiv: https://arxiv.org/abs/2603.11965v1 Standard Adjacency Spectral Embedding (ASE) relies on a global low-rank assumption often incompatible with the sparse, transitive structure of real-world networks, causing local geometric features to be 'smeared'. To address this, we introduce Local Adjacency Spectral Embedding (LASE), which uncovers locally low-dimensional structure via weighted spectral decomposition. Under a latent position model with a kernel feature map, we treat the image of latent positions as a locally low-dimensional set in infinite-dimensional feature space. We establish finite-sample bounds quantifying the trade-off between the statistical cost of localisation and the reduced truncation error achieved by targeting a locally low-dimensional region of the embedding. Furthermore, we prove that sufficient localisation induces rapid spectral decay and the emergence of a distinct spectral gap, theoretically justifying low-dimensional local embeddings. Experiments on synthetic and real networks show that LASE improves local reconstruction and visualisation over global and subgraph baselines, and we introduce UMAP-LASE for assembling overlapping local embeddings into high-fidelity global visualisations. --- ### CHiL(L)Grader: Calibrated Human-in-the-Loop Short-Answer Grading Authors: Pranav Raikote, Korbinian Randl, Ioanna Miliou, Athanasios Lakes, Panagiotis Papapetrou Published: 2026-03-12 Categories: cs.CL arXiv: https://arxiv.org/abs/2603.11957v1 Scaling educational assessment with large language models requires not just accuracy, but the ability to recognize when predictions are trustworthy. Instruction-tuned models tend to be overconfident, and their reliability deteriorates as curricula evolve, making fully autonomous deployment unsafe in high-stakes settings. We introduce CHiL(L)Grader, the first automated grading framework that incorporates calibrated confidence estimation into a human-in-the-loop workflow. Using post-hoc temperature scaling, confidence-based selective prediction, and continual learning, CHiL(L)Grader automates only high-confidence predictions while routing uncertain cases to human graders, and adapts to evolving rubrics and unseen questions. Across three short-answer grading datasets, CHiL(L)Grader automatically scores 35-65% of responses at expert-level quality (QWK >= 0.80). A QWK gap of 0.347 between accepted and rejected predictions confirms the effectiveness of the confidence-based routing. Each correction cycle strengthens the model's grading capability as it learns from teacher feedback. These results show that uncertainty quantification is key for reliable AI-assisted grading. --- ### PersonaTrace: Synthesizing Realistic Digital Footprints with LLM Agents Authors: Minjia Wang, Yunfeng Wang, Xiao Ma, Dexin Lv, Qifan Guo, Lynn Zheng, Benliang Wang, Lei Wang, Jiannan Li, Yongwei Xing, David Xu, Zheng Sun Published: 2026-03-12 Categories: cs.CL arXiv: https://arxiv.org/abs/2603.11955v1 Digital footprints (records of individuals' interactions with digital systems) are essential for studying behavior, developing personalized applications, and training machine learning models. However, research in this area is often hindered by the scarcity of diverse and accessible data. To address this limitation, we propose a novel method for synthesizing realistic digital footprints using large language model (LLM) agents. Starting from a structured user profile, our approach generates diverse and plausible sequences of user events, ultimately producing corresponding digital artifacts such as emails, messages, calendar entries, reminders, etc. Intrinsic evaluation results demonstrate that the generated dataset is more diverse and realistic than existing baselines. Moreover, models fine-tuned on our synthetic data outperform those trained on other synthetic datasets when evaluated on real-world out-of-distribution tasks. --- ### Preliminary analysis of RGB-NIR Image Registration techniques for off-road forestry environments Authors: Pankaj Deoli, Karthik Ranganath, Karsten Berns Published: 2026-03-12 Categories: cs.CV arXiv: https://arxiv.org/abs/2603.11952v1 RGB-NIR image registration plays an important role in sensor-fusion, image enhancement and off-road autonomy. In this work, we evaluate both classical and Deep Learning (DL) based image registration techniques to access their suitability for off-road forestry applications. NeMAR, trained under 6 different configurations, demonstrates partial success however, its GAN loss instability suggests challenges in preserving geometric consistency. MURF, when tested on off-road forestry data shows promising large scale feature alignment during shared information extraction but struggles with fine details in dense vegetation. Even though this is just a preliminary evaluation, our study necessitates further refinements for robust, multi-scale registration for off-road forest applications. --- ### Learning Transferable Sensor Models via Language-Informed Pretraining Authors: Yuliang Chen, Arvind Pillai, Yu Yvonne Wu, Sudarshan Regmi, Tess Z. Griffin, Lisa Marsch, Michael V. Heinz, Nicholas C. Jacobson, Andrew Campbell Published: 2026-03-12 Categories: cs.AI, cs.LG arXiv: https://arxiv.org/abs/2603.11950v2 Multimodal language models have demonstrated strong semantic understanding and reasoning over physiological and behavioral signals captured from diverse healthcare sensors. However, existing sensor-language models can only process a fixed number of sensor channels at a fixed sampling rate, making them difficult to adapt to new or different sensor configurations. This inflexibility prevents models from generalizing across heterogeneous health sensing ecosystems, where data spans high-frequency wearable streams to sparse, day-scale clinical signals. To bridge this gap, we introduce SLIP (Sensor Language-Informed Pretraining), an open-source framework for learning language-aligned representations that generalize across diverse sensor setups. SLIP integrates contrastive alignment with sensor-conditioned captioning, facilitating both discriminative understanding and generative reasoning. By repurposing a pretrained decoder-only language model via cross-attention and introducing a flexible patch embedder, SLIP transfers to new sensor configurations at inference without retraining, regardless of their native sampling rate or input length. Across 11 datasets, SLIP demonstrates superior performance in retrieval, signal captioning, and question answering. It achieves a 77.14% average linear-probing accuracy, a 3.56% improvement over the strongest baseline (SensorLM). Beyond classification, SLIP supports open-vocabulary sensor captioning and question answering without any architectural modifications. Our work offers broad implications for the development of sensor-language models, utilizing cross-modal supervision to ensure more robust generalizability in health sensing applications. --- ### Learning Transferable Sensor Models via Language-Informed Pretraining Authors: Yuliang Chen, Arvind Pillai, Yu Yvonne Wu, Tess Z. Griffin, Lisa Marsch, Michael V. Heinz, Nicholas C. Jacobson, Andrew Campbell Published: 2026-03-12 Categories: cs.AI, cs.LG arXiv: https://arxiv.org/abs/2603.11950v1 Modern sensing systems generate large volumes of unlabeled multivariate time-series data. This abundance of unlabeled data makes self-supervised learning (SSL) a natural approach for learning transferable representations. However, most existing approaches are optimized for reconstruction or forecasting objectives and often fail to capture the semantic structure required for downstream classification and reasoning tasks. While recent sensor-language alignment methods improve semantic generalization through captioning and zero-shot transfer, they are limited to fixed sensor configurations, such as predefined channel sets, signal lengths, or temporal resolutions, which hinders cross-domain applicability. To address these gaps, we introduce \textbf{SLIP} (\textbf{S}ensor \textbf{L}anguage-\textbf{I}nformed \textbf{P}retraining), an open-source framework for learning language-aligned representations that generalize across diverse sensor setups. SLIP integrates contrastive alignment with sensor-conditioned captioning, facilitating both discriminative understanding and generative reasoning. By repurposing a pretrained decoder-only language model via cross-attention and introducing an elegant, flexible patch-embedder, SLIP supports different temporal resolutions and variable-length input at inference time without additional retraining. Across 11 datasets, SLIP demonstrates superior performance in zero-shot transfer, signal captioning, and question answering. It achieves a 77.14% average linear-probing accuracy, a 5.93% relative improvement over strong baselines, and reaches 64.83% accuracy in sensor-based question answering. --- ### Delayed Backdoor Attacks: Exploring the Temporal Dimension as a New Attack Surface in Pre-Trained Models Authors: Zikang Ding, Haomiao Yang, Meng Hao, Wenbo Jiang, Kunlan Xiang, Runmeng Du, Yijing Liu, Ruichen Zhang, Dusit Niyato Published: 2026-03-12 Categories: cs.CR, cs.AI arXiv: https://arxiv.org/abs/2603.11949v1 Backdoor attacks against pre-trained models (PTMs) have traditionally operated under an ``immediacy assumption,'' where malicious behavior manifests instantly upon trigger occurrence. This work revisits and challenges this paradigm by introducing \textit{\textbf{Delayed Backdoor Attacks (DBA)}}, a new class of threats in which activation is temporally decoupled from trigger exposure. We propose that this \textbf{temporal dimension} is the key to unlocking a previously infeasible class of attacks: those that use common, everyday words as triggers. To examine the feasibility of this paradigm, we design and implement a proof-of-concept prototype, termed \underline{D}elayed Backdoor Attacks Based on \underline{N}onlinear \underline{D}ecay (DND). DND embeds a lightweight, stateful logic module that postpones activation until a configurable threshold is reached, producing a distinct latency phase followed by a controlled outbreak. We derive a formal model to characterize this latency behavior and propose a dual-metric evaluation framework (ASR and ASR$_{delay}$) to empirically measure the delay effect. Extensive experiments on four (natural language processing)NLP benchmarks validate the core capabilities of DND: it remains dormant for a controllable duration, sustains high clean accuracy ($\ge$94\%), and achieves near-perfect post-activation attack success rates ($\approx$99\%, The average of other methods is below 95\%.). Moreover, DND exhibits resilience against several state-of-the-art defenses. This study provides the first empirical evidence that the temporal dimension constitutes a viable yet unprotected attack surface in PTMs, underscoring the need for next-generation, stateful, and time-aware defense mechanisms. --- ### Resurfacing Paralinguistic Awareness in Large Audio Language Models Authors: Hao Yang, Minghan Wang, Tongtong Wu, Lizhen Qu, Ehsan Shareghi, Gholamreza Haffari Published: 2026-03-12 Categories: cs.SD, cs.CL, cs.MM, eess.AS arXiv: https://arxiv.org/abs/2603.11947v1 Large Audio Language Models (LALMs) have expanded the interaction with human to speech modality, which introduces great interactive potential, due to the paralinguistic cues implicitly indicating the user context. However, building on the current content-centred paradigm, LALMs usually neglect such paralinguistic cues and respond solely based on query content. In this work, to resurface the paralinguistic awareness in LALMs, we introduce five diverse layer-wise analyses to jointly identify paralinguistic layers and semantic understanding layers. Based on these insights, we propose a paralinguistic-enhanced fine-tuning (PE-FT) protocol accordingly to equip LALMs with paralinguistic-aware capabilities, including (1) selective-layer fine-tuning, and (2) an auxiliary dual-level classification head. Our experiments demonstrate that PE-FT protocol efficiently and effectively resurfaces the paralinguistic awareness, even surpassing the performance of the all-layer fine-tuning strategy. --- ### Geometry-Aware Probabilistic Circuits via Voronoi Tessellations Authors: Sahil Sidheekh, Sriraam Natarajan Published: 2026-03-12 Categories: cs.LG, cs.AI arXiv: https://arxiv.org/abs/2603.11946v2 Probabilistic circuits (PCs) enable exact and tractable inference but employ data independent mixture weights that limit their ability to capture local geometry of the data manifold. We propose Voronoi tessellations (VT) as a natural way to incorporate geometric structure directly into the sum nodes of a PC. However, naïvely introducing such structure breaks tractability. We formalize this incompatibility and develop two complementary solutions: (1) an approximate inference framework that provides guaranteed lower and upper bounds for inference, and (2) a structural condition for VT under which exact tractable inference is recovered. Finally, we introduce a differentiable relaxation for VT that enables gradient-based learning and empirically validate the resulting approach on standard density estimation tasks. --- ### Effective Resistance Rewiring: A Simple Topological Correction for Over-Squashing Authors: Bertran Miquel-Oliver, Manel Gil-Sorribes, Victor Guallar, Alexis Molina Published: 2026-03-12 Categories: cs.LG, cs.AI arXiv: https://arxiv.org/abs/2603.11944v1 Graph Neural Networks struggle to capture long-range dependencies due to over-squashing, where information from exponentially growing neighborhoods must pass through a small number of structural bottlenecks. While recent rewiring methods attempt to alleviate this limitation, many rely on local criteria such as curvature, which can overlook global connectivity constraints that restrict information flow. We introduce Effective Resistance Rewiring (ERR), a simple topology correction strategy that uses effective resistance as a global signal to detect structural bottlenecks. ERR iteratively adds edges between node pairs with the largest resistance while removing edges with minimal resistance, strengthening weak communication pathways while controlling graph densification under a fixed edge budget. The procedure is parameter-free beyond the rewiring budget and relies on a single global measure aggregating all paths between node pairs. Beyond predictive performance with GCN models, we analyze how rewiring affects message propagation. By tracking cosine similarity between node embeddings across layers, we examine how the relationship between initial node features and learned representations evolves during message passing, comparing graphs with and without rewiring. This analysis helps determine whether improvements arise from better long-range communication rather than changes in embedding geometry. Experiments on homophilic and heterophilic graphs, including directed settings with DirGCN, reveal a trade-off between over-squashing and oversmoothing, where oversmoothing corresponds to the loss of representation diversity across layers. Resistance-guided rewiring improves connectivity and signal propagation but can accelerate representation mixing in deep models. Combining ERR with normalization techniques such as PairNorm stabilizes this trade-off and improves performance. --- ### Causal Matrix Completion under Multiple Treatments via Mixed Synthetic Nearest Neighbors Authors: Minrui Luo, Zhiheng Zhang Published: 2026-03-12 Categories: cs.LG arXiv: https://arxiv.org/abs/2603.11942v1 Synthetic Nearest Neighbors (SNN) provides a principled solution to causal matrix completion under missing-not-at-random (MNAR) by exploiting local low-rank structure through fully observed anchor submatrices. However, its effectiveness critically relies on sufficient data availability within each treatment level, a condition that often fails in settings with multiple or complex treatments. In this work, we propose Mixed Synthetic Nearest Neighbors (MSNN), a new entry-wise causal identification estimator that integrates information across treatment levels. We show that MSNN retains the finite-sample error bounds and asymptotic normality guarantees of SNN, while enlarging the effective sample size available for estimation. Empirical results on synthetic and real-world datasets illustrate the efficacy of the proposed approach, especially under data-scarce treatment levels. --- ### Exhaustive Circuit Mapping of a Single-Cell Foundation Model Reveals Massive Redundancy, Heavy-Tailed Hub Architecture, and Layer-Dependent Differentiation Control Authors: Ihor Kendiukhov Published: 2026-03-12 Categories: cs.LG arXiv: https://arxiv.org/abs/2603.11940v1 Mechanistic interpretability of biological foundation models has relied on selective feature sampling, pairwise interaction testing, and observational trajectory analysis. Each of these can introduce systematic bias. Here we present three experiments that address these limitations through exhaustive circuit tracing, higher order combinatorial ablation, and causal trajectory steering in Geneformer, a transformer based single cell foundation model. First, exhaustive tracing of all 4065 active sparse autoencoder features at layer 5 yields 1393850 significant downstream edges, a 27 fold expansion over selective sampling. This reveals a heavy tailed hub distribution in which 1.8 percent of features account for disproportionate connectivity and 40 percent of the top 20 hubs lack biological annotation. These results indicate systematic annotation bias in prior selective analyses. Second, three way combinatorial ablation across 8 feature triplets shows that redundancy deepens monotonically with interaction order, with a three way ratio of 0.59 versus a pairwise ratio of 0.74, and with zero synergy. This confirms that the model architecture is subadditive at all tested orders. Third, trajectory guided feature steering establishes a causal link between layer position and differentiation directionality. Late layer features at L17 consistently push cell states toward maturity, with fraction positive equal to 1.0. Early and mid layer features at L0 and L11 mostly push away from maturity, with fraction positive ranging from 0.00 to 0.58. Together these results move from correlation toward causal evidence for layer dependent control of cell state. --- ### SNAP-V: A RISC-V SoC with Configurable Neuromorphic Acceleration for Small-Scale Spiking Neural Networks Authors: Kanishka Gunawardana, Sanka Peeris, Kavishka Rambukwella, Thamish Wanduragala, Saadia Jameel, Roshan Ragel, Isuru Nawinne Published: 2026-03-12 Categories: cs.AR, cs.NE arXiv: https://arxiv.org/abs/2603.11939v1 Spiking Neural Networks (SNNs) have gained significant attention in edge computing due to their low power consumption and computational efficiency. However, existing implementations either use conventional System on Chip (SoC) architectures that suffer from memory-processor bottlenecks, or large-scale neuromorphic hardware that is inefficient and wasteful for small-scale SNN applications. This work presents SNAP-V, a RISC-V-based neuromorphic SoC with two accelerator variants: Cerebra-S (bus-based) and Cerebra-H (Network-on-Chip (NoC)-based) which are optimized for small-scale SNN inference, integrating a RISC-V core for management tasks, with both accelerators featuring parallel processing nodes and distributed memory. Experimental results show close agreement between software and hardware inference, with an average accuracy deviation of 2.62% across multiple network configurations, and an average synaptic energy of 1.05 pJ per synaptic operation (SOP) in 45 nm CMOS technology. These results show that the proposed solution enables accurate, energy-efficient SNN inference suitable for real-time edge applications. --- ### Prototype-Based Knowledge Guidance for Fine-Grained Structured Radiology Reporting Authors: Chantal Pellegrini, Adrian Delchev, Ege Özsoy, Nassir Navab, Matthias Keicher Published: 2026-03-12 Categories: cs.AI, cs.CV, cs.LG arXiv: https://arxiv.org/abs/2603.11938v1 Structured radiology reporting promises faster, more consistent communication than free text, but automation remains difficult as models must make many fine-grained, discrete decisions about rare findings and attributes from limited structured supervision. In contrast, free-text reports are produced at scale in routine care and implicitly encode fine-grained, image-linked information through detailed descriptions. To leverage this unstructured knowledge, we propose ProtoSR, an approach for injecting free-text information into structured report population. First, we introduce an automatic extraction pipeline that uses an instruction-tuned LLM to mine 80k+ MIMIC-CXR studies and build a multimodal knowledge base aligned with a structured reporting template, representing each answer option with a visual prototype. Using this knowledge base, ProtoSR is trained to retrieve prototypes relevant for the current image-question pair and augment the model predictions through a prototype-conditioned residual, providing a data-driven second opinion that selectively corrects predictions. On the Rad-ReStruct benchmark, ProtoSR achieves state-of-the-art results, with the largest improvements on detailed attribute questions, demonstrating the value of integrating free-text derived signal for fine-grained image understanding. --- ### Fair Learning for Bias Mitigation and Quality Optimization in Paper Recommendation Authors: Uttamasha Anjally Oyshi, Susan Gauch Published: 2026-03-12 Categories: cs.AI arXiv: https://arxiv.org/abs/2603.11936v1 Despite frequent double-blind review, demographic biases of authors still disadvantage the underrepresented groups. We present Fair-PaperRec, a MultiLayer Perceptron (MLP)-based model that addresses demographic disparities in post-review paper acceptance decisions while maintaining high-quality requirements. Our methodology penalizes demographic disparities while preserving quality through intersectional criteria (e.g., race, country) and a customized fairness loss, in contrast to heuristic approaches. Evaluations using conference data from ACM Special Interest Group on Computer-Human Interaction (SIGCHI), Designing Interactive Systems (DIS), and Intelligent User Interfaces (IUI) indicate a 42.03% increase in underrepresented group participation and a 3.16% improvement in overall utility, indicating that diversity promotion does not compromise academic rigor and supports equity-focused peer review solutions. --- ### MobileKernelBench: Can LLMs Write Efficient Kernels for Mobile Devices? Authors: Xingze Zou, Jing Wang, Yuhua Zheng, Xueyi Chen, Haolei Bai, Lingcheng Kong, Syed A. R. Abu-Bakar, Zhaode Wang, Chengfei Lv, Haoji Hu, Huan Wang Published: 2026-03-12 Categories: cs.LG, cs.AI arXiv: https://arxiv.org/abs/2603.11935v2 Large language models (LLMs) have demonstrated remarkable capabilities in code generation, yet their potential for generating kernels specifically for mobile devices remains largely unexplored. In this work, we extend the scope of automated kernel generation to the mobile domain to investigate the central question: Can LLMs write efficient kernels for mobile devices? To enable systematic investigation, we introduce MobileKernelBench, a comprehensive evaluation framework comprising a benchmark prioritizing operator diversity and cross-framework interoperability, coupled with an automated pipeline that bridges the host-device gap for on-device verification. Leveraging this framework, we conduct extensive evaluation on the CPU backend of Mobile Neural Network (MNN), revealing that current LLMs struggle with the engineering complexity and data scarcity inherent to mobile frameworks; standard models and even fine-tuned variants exhibit high compilation failure rates (over 54%) and negligible performance gains due to hallucinations and a lack of domain-specific grounding. To overcome these limitations, we propose the Mobile Kernel Agent (MoKA), a multi-agent system equipped with repository-aware reasoning and a plan-and-execute paradigm. Validated on MobileKernelBench, MoKA achieves state-of-the-art performance, boosting compilation success to 93.7% and enabling 27.4% of generated kernels to deliver measurable speedups over native libraries. --- ### AS-Bridge: A Bidirectional Generative Framework Bridging Next-Generation Astronomical Surveys Authors: Dichang Zhang, Yixuan Shao, Simon Birrer, Dimitris Samaras Published: 2026-03-12 Categories: astro-ph.IM, cs.CV arXiv: https://arxiv.org/abs/2603.11928v2 The upcoming decade of observational cosmology will be shaped by large sky surveys, such as the ground-based LSST at the Vera C. Rubin Observatory and the space-based Euclid mission. While they promise an unprecedented view of the Universe across depth, resolution, and wavelength, their differences in observational modality, sky coverage, point-spread function, and scanning cadence make joint analysis beneficial, but also challenging. To facilitate joint analysis, we introduce A(stronomical)S(urvey)-Bridge, a bidirectional generative model that translates between ground- and space-based observations. AS-Bridge learns a diffusion model that employs a stochastic Brownian Bridge process between the LSST and Euclid observations. The two surveys have overlapping sky regions, where we can explicitly model the conditional probabilistic distribution between them. We show that this formulation enables new scientific capabilities beyond single-survey analysis, including faithful probabilistic predictions of missing survey observations and inter-survey detection of rare events. These results establish the feasibility of inter-survey generative modeling. AS-Bridge is therefore well-positioned to serve as a complementary component of future LSST-Euclid joint data pipelines, enhancing the scientific return once data from both surveys become available. Data and code are available at https://github.com/ZHANG7DC/AS-Bridge. --- ### Chem4DLLM: 4D Multimodal LLMs for Chemical Dynamics Understanding Authors: Xinyu Li, Zhen Zhang, Qi Chen, Anton van den Hengel, Lina Yao, Javen Qinfeng Shi Published: 2026-03-12 Categories: cs.LG, cs.CL arXiv: https://arxiv.org/abs/2603.11924v1 Existing chemical understanding tasks primarily rely on static molecular representations, limiting their ability to model inherently dynamic phenomena such as bond breaking or conformational changes, which are essential for a chemist to understand chemical reactions. To address this gap, we introduce Chemical Dynamics Understanding (ChemDU), a new task that translates 4D molecular trajectories into interpretable natural-language explanations. ChemDU focuses on fundamental dynamic scenarios, including gas-phase and catalytic reactions, and requires models to reason about key events along molecular trajectories, such as bond formation and dissociation, and to generate coherent, mechanistically grounded narratives. To benchmark this capability, we construct Chem4DBench, the first dataset pairing 4D molecular trajectories with expert-authored explanations across these settings. We further propose Chem4DLLM, a unified model that integrates an equivariant graph encoder with a pretrained large language model to explicitly capture molecular geometry and rotational dynamics. We hope that ChemDU, together with Chem4DBench and Chem4DLLM, will stimulate further research in dynamic chemical understanding and multimodal scientific reasoning. --- ### PicoSAM3: Real-Time In-Sensor Region-of-Interest Segmentation Authors: Pietro Bonazzi, Nicola Farronato, Stefan Zihlmann, Haotong Qin, Michele Magno Published: 2026-03-12 Categories: cs.CV arXiv: https://arxiv.org/abs/2603.11917v4 Real-time, on-device segmentation is critical for latency-sensitive and privacy-aware applications such as smart glasses and Internet-of-Things devices. We introduce PicoSAM3, a lightweight promptable visual segmentation model optimized for edge and in-sensor execution, including deployment on the Sony IMX500 vision sensor. PicoSAM3 has 1.3M parameters and combines a dense CNN architecture with region of interest prompt encoding, Efficient Channel Attention, and knowledge distillation from SAM2 and SAM3. On COCO and LVIS, PicoSAM3 achieves 65.45% and 64.01% mIoU, respectively, outperforming existing SAM-based and edge-oriented baselines at similar or lower complexity. The INT8 quantized model preserves accuracy with negligible degradation while enabling real-time in-sensor inference at 11.82ms latency on the IMX500, fully complying with its memory and operator constraints. Ablation studies show that distillation from large SAM models yields up to +14.5% mIoU improvement over supervised training and demonstrate that high-quality, spatially flexible promptable segmentation is feasible directly at the sensor level. --- ### CoMMET: A Psychologically Grounded Benchmark for Evaluating Theory of Mind in Multimodal LLMs Authors: Ruirui Chen, Weifeng Jiang, Chengwei Qin, Kaiwen Wei, Yanzhen Yue, Cheston Tan Published: 2026-03-12 Categories: cs.CL arXiv: https://arxiv.org/abs/2603.11915v2 Theory of Mind (ToM)-the ability to reason about the mental states of oneself and others-is a cornerstone of human social intelligence. As Multimodal Large Language Models (MLLMs) become ubiquitous in real-world applications, validating their capacity for this level of social reasoning is essential for effective and natural interactions. However, existing benchmarks for assessing ToM in MLLMs are limited; most rely solely on text inputs and focus narrowly on belief-related tasks. In this paper, we propose a new multimodal benchmark dataset, CoMMET, a comprehensive mental states and moral evaluation task inspired by the Theory of Mind Booklet Task. CoMMET expands the scope of evaluation by covering a broader range of mental states and introducing multi-turn testing. To the best of our knowledge, this is the first psychology-grounded benchmark to evaluate MLLMs across multiple mental states in a multimodal, open-ended, and multi-turn setting. Through a comprehensive assessment of MLLMs across different families and sizes, we analyze the strengths and limitations of current models and identify directions for future improvement. --- ### CoMMET: To What Extent Can LLMs Perform Theory of Mind Tasks? Authors: Ruirui Chen, Weifeng Jiang, Chengwei Qin, Cheston Tan Published: 2026-03-12 Categories: cs.CL arXiv: https://arxiv.org/abs/2603.11915v1 Theory of Mind (ToM)-the ability to reason about the mental states of oneself and others-is a cornerstone of human social intelligence. As Large Language Models (LLMs) become ubiquitous in real-world applications, validating their capacity for this level of social reasoning is essential for effective and natural interactions. However, existing benchmarks for assessing ToM in LLMs are limited; most rely solely on text inputs and focus narrowly on belief-related tasks. In this paper, we propose a new multimodal benchmark dataset, CoMMET, a Comprehensive Mental states and Moral Evaluation Task inspired by the Theory of Mind Booklet Task. CoMMET expands the scope of evaluation by covering a broader range of mental states and introducing multi-turn testing. To the best of our knowledge, this is the first multimodal dataset to evaluate ToM in a multi-turn conversational setting. Through a comprehensive assessment of LLMs across different families and sizes, we analyze the strengths and limitations of current models and identify directions for future improvement. Our work offers a deeper understanding of the social cognitive capabilities of modern LLMs. --- ### Understanding LLM Behavior When Encountering User-Supplied Harmful Content in Harmless Tasks Authors: Junjie Chu, Yiting Qu, Ye Leng, Michael Backes, Yun Shen, Savvas Zannettou, Yang Zhang Published: 2026-03-12 Categories: cs.CR, cs.AI arXiv: https://arxiv.org/abs/2603.11914v1 Large Language Models (LLMs) are increasingly trained to align with human values, primarily focusing on task level, i.e., refusing to execute directly harmful tasks. However, a subtle yet crucial content-level ethical question is often overlooked: when performing a seemingly benign task, will LLMs -- like morally conscious human beings -- refuse to proceed when encountering harmful content in user-provided material? In this study, we aim to understand this content-level ethical question and systematically evaluate its implications for mainstream LLMs. We first construct a harmful knowledge dataset (i.e., non-compliant with OpenAI's usage policy) to serve as the user-supplied harmful content, with 1,357 entries across ten harmful categories. We then design nine harmless tasks (i.e., compliant with OpenAI's usage policy) to simulate the real-world benign tasks, grouped into three categories according to the extent of user-supplied content required: extensive, moderate, and limited. Leveraging the harmful knowledge dataset and the set of harmless tasks, we evaluate how nine LLMs behave when exposed to user-supplied harmful content during the execution of benign tasks, and further examine how the dynamics between harmful knowledge categories and tasks affect different LLMs. Our results show that current LLMs, even the latest GPT-5.2 and Gemini-3-Pro, often fail to uphold human-aligned ethics by continuing to process harmful content in harmless tasks. Furthermore, external knowledge from the ``Violence/Graphic'' category and the ``Translation'' task is more likely to elicit harmful responses from LLMs. We also conduct extensive ablation studies to investigate potential factors affecting this novel misuse vulnerability. We hope that our study could inspire enhanced safety measures among stakeholders to mitigate this overlooked content-level ethical risk. ---