1. FlyTrap: Physical Distance-Pulling Attack Towards Camera-based Autonomous Target Tracking Systems
Authors: Shaoyuan Xie, Mohamad Habib Fakih, Junchi Lu, Fayzah Alshammari, Ningfei Wang, Takami Sato, Halima Bouzidi, Mohammad Abdullah Al Faruque, Qi Alfred Chen β’
Published: 2025-09-24 β’
Source: arXiv
Autonomous Target Tracking (ATT) systems, especially ATT drones, are widely used in applications such as surveillance, border control, and law enforcement, while also being misused in stalking and destructive actions. Thus, the security of ATT is highly critical for real-world applications. Under the scope, we present a new type of attack: distance-pulling attacks (DPA) and a systematic study of it, which exploits vulnerabilities in ATT systems to dangerously reduce tracking distances, leading to drone capturing, increased susceptibility to sensor attacks, or even physical collisions. To achieve these goals, we present FlyTrap, a novel physical-world attack framework that employs an adversarial umbrella as a deployable and domain-specific attack vector. FlyTrap is specifically designed to meet key desired objectives in attacking ATT drones: physical deployability, closed-loop effectiveness, and spatial-temporal consistency. Through novel progressive distance-pulling strategy and controllable spatial-temporal consistency designs, FlyTrap manipulates ATT drones in real-world setups to achieve significant system-level impacts. Our evaluations include new datasets, metrics, and closed-loop experiments on real-world white-box and even commercial ATT drones, including DJI and HoverAir. Results demonstrate FlyTrap's ability to reduce tracking distances within the range to be captured, sensor attacked, or even directly crashed, highlighting urgent security risks and practical implications for the safe deployment of ATT systems.
2. EditVerse: Unifying Image and Video Editing and Generation with In-Context Learning
Authors: Xuan Ju, Tianyu Wang, Yuqian Zhou, He Zhang, Qing Liu, Nanxuan Zhao, Zhifei Zhang, Yijun Li, Yuanhao Cai, Shaoteng Liu, Daniil Pakhomov, Zhe Lin, Soo Ye Kim, Qiang Xu β’
Published: 2025-09-24 β’
Source: arXiv
Recent advances in foundation models highlight a clear trend toward unification and scaling, showing emergent capabilities across diverse domains. While image generation and editing have rapidly transitioned from task-specific to unified frameworks, video generation and editing remain fragmented due to architectural limitations and data scarcity. In this work, we introduce EditVerse, a unified framework for image and video generation and editing within a single model. By representing all modalities, i.e., text, image, and video, as a unified token sequence, EditVerse leverages self-attention to achieve robust in-context learning, natural cross-modal knowledge transfer, and flexible handling of inputs and outputs with arbitrary resolutions and durations. To address the lack of video editing training data, we design a scalable data pipeline that curates 232K video editing samples and combines them with large-scale image and video datasets for joint training. Furthermore, we present EditVerseBench, the first benchmark for instruction-based video editing covering diverse tasks and resolutions. Extensive experiments and user studies demonstrate that EditVerse achieves state-of-the-art performance, surpassing existing open-source and commercial models, while exhibiting emergent editing and generation abilities across modalities.
3. EmbeddingGemma: Powerful and Lightweight Text Representations
Authors: Henrique Schechter Vera, Sahil Dua, Biao Zhang, Daniel Salz, Ryan Mullins, Sindhu Raghuram Panyam, Sara Smoot, Iftekhar Naim, Joe Zou, Feiyang Chen, Daniel Cer, Alice Lisak, Min Choi, Lucas Gonzalez, Omar Sanseviero, Glenn Cameron, Ian Ballantyne, Kat Black, Kaifeng Chen, Weiyi Wang, Zhe Li, Gus Martins, Jinhyuk Lee, Mark Sherwood, Juyeong Ji, Renjie Wu, Jingxiao Zheng, Jyotinder Singh, Abheesht Sharma, Divya Sreepat, Aashi Jain, Adham Elarabawy, AJ Co, Andreas Doumanoglou, Babak Samari, Ben Hora, Brian Potetz, Dahun Kim, Enrique Alfonseca, Fedor Moiseev, Feng Han, Frank Palma Gomez, Gustavo HernΓ‘ndez Γbrego, Hesen Zhang, Hui Hui, Jay Han, Karan Gill, Ke Chen, Koert Chen, Madhuri Shanbhogue, Michael Boratko, Paul Suganthan, Sai Meher Karthik Duddu, Sandeep Mariserla, Setareh Ariafar, Shanfeng Zhang, Shijie Zhang, Simon Baumgartner, Sonam Goenka, Steve Qiu, Tanmaya Dabral, Trevor Walker, Vikram Rao, Waleed Khawaja, Wenlei Zhou, Xiaoqi Ren, Ye Xia, Yichang Chen, Yi-Ting Chen, Zhe Dong, Zhongli Ding, Francesco Visin, GaΓ«l Liu, Jiageng Zhang, Kathleen Kenealy, Michelle Casbon, Ravin Kumar, Thomas Mesnard, Zach Gleicher, Cormac Brick, Olivier Lacombe, Adam Roberts, Yunhsuan Sung, Raphael Hoffmann, Tris Warkentin, Armand Joulin, Tom Duerig, Mojtaba Seyedhosseini β’
Published: 2025-09-24 β’
Source: arXiv
We introduce EmbeddingGemma, a new lightweight, open text embedding model based on the Gemma 3 language model family. Our innovative training recipe strategically captures knowledge from larger models via encoder-decoder initialization and geometric embedding distillation. We improve model robustness and expressiveness with a spread-out regularizer, and ensure generalizability by merging checkpoints from varied, optimized mixtures. Evaluated on the Massive Text Embedding Benchmark (MTEB) across multilingual, English, and code domains, EmbeddingGemma (300M) achieves state-of-the-art results. Notably, it outperforms prior top models, both proprietary and open, with fewer than 500M parameters, and provides performance comparable to models double its size, offering an exceptional performance-to-cost ratio. Remarkably, this lead persists when quantizing model weights or truncating embedding outputs. This makes EmbeddingGemma particularly well-suited for low-latency and high-throughput use cases such as on-device applications. We provide ablation studies exploring our key design choices. We release EmbeddingGemma to the community to promote further research.
4. Process-Informed Forecasting of Complex Thermal Dynamics in Pharmaceutical Manufacturing
Authors: Ramona Rubini, Siavash Khodakarami, Aniruddha Bora, George Em Karniadakis, Michele Dassisti β’
Published: 2025-09-24 β’
Source: arXiv
Accurate time-series forecasting for complex physical systems is the backbone of modern industrial monitoring and control. While deep learning models excel at capturing complex dynamics, currently, their deployment is limited due to physical inconsistency and robustness, hence constraining their reliability in regulated environments. We introduce process-informed forecasting (PIF) models for temperature in pharmaceutical lyophilization. We investigate a wide range of models, from classical ones such as Autoregressive Integrated Moving Average Model (ARIMA) and Exponential Smoothing Model (ETS), to modern deep learning architectures, including Kolmogorov-Arnold Networks (KANs). We compare three different loss function formulations that integrate a process-informed trajectory prior: a fixed-weight loss, a dynamic uncertainty-based loss, and a Residual-Based Attention (RBA) mechanism. We evaluate all models not only for accuracy and physical consistency but also for robustness to sensor noise. Furthermore, we test the practical generalizability of the best model in a transfer learning scenario on a new process. Our results show that PIF models outperform their data-driven counterparts in terms of accuracy, physical plausibility and noise resilience. This work provides a roadmap for developing reliable and generalizable forecasting solutions for critical applications in the pharmaceutical manufacturing landscape.
5. Quantum speed limits based on Jensen-Shannon and Jeffreys divergences for general physical processes
Authors: Jucelino Ferreira de Sousa, Diego Paiva Pires β’
Published: 2025-09-24 β’
Source: arXiv
We discuss quantum speed limits (QSLs) for finite-dimensional quantum systems undergoing a general physical process. These QSLs were obtained using two families of entropic measures, namely the square root of the Jensen-Shannon divergence, which in turn defines a faithful distance of quantum states, and the square root of the quantum Jeffreys divergence. The results apply to both closed and open quantum systems, and are evaluated in terms of the Schatten speed of the evolved state, as well as cost functions that depend on the smallest and largest eigenvalues of both initial and instantaneous states of the quantum system. To illustrate our findings, we focus on the unitary and nonunitary dynamics of mixed single-qubit states. In the first case, we obtain speed limits $\textit{\`{a} la}$ Mandelstam-Tamm that are inversely proportional to the variance of the Hamiltonian driving the evolution. In the second case, we set the nonunitary dynamics to be described by the noisy operations: depolarizing channel, phase damping channel, and generalized amplitude damping channel. We provide analytical results for the two entropic measures, present numerical simulations to support our results on the speed limits, comment on the tightness of the bounds, and provide a comparison with previous QSLs. Our results may find applications in the study of quantum thermodynamics, entropic uncertainty relations, and also complexity of many-body systems.
6. xGFabric: Coupling Sensor Networks and HPC Facilities with Private 5G Wireless Networks for Real-Time Digital Agriculture
Authors: Liubov Kurafeeva, Alan Subedi, Ryan Hartung, Michael Fay, Avhishek Biswas, Shantenu Jha, Ozgur O. Kilic, Chandra Krintz, Andre Merzky, Douglas Thain, Mehmet C. Vuran, Rich Wolski β’
Published: 2025-09-24 β’
Source: arXiv
Advanced scientific applications require coupling distributed sensor networks with centralized high-performance computing facilities. Citrus Under Protective Screening (CUPS) exemplifies this need in digital agriculture, where citrus research facilities are instrumented with numerous sensors monitoring environmental conditions and detecting protective screening damage. CUPS demands access to computational fluid dynamics codes for modeling environmental conditions and guiding real-time interventions like water application or robotic repairs. These computing domains have contrasting properties: sensor networks provide low-performance, limited-capacity, unreliable data access, while high-performance facilities offer enormous computing power through high-latency batch processing. Private 5G networks present novel capabilities addressing this challenge by providing low latency, high throughput, and reliability necessary for near-real-time coupling of edge sensor networks with HPC simulations. This work presents xGFabric, an end-to-end system coupling sensor networks with HPC facilities through Private 5G networks. The prototype connects remote sensors via 5G network slicing to HPC systems, enabling real-time digital agriculture simulation.
7. Feature Dynamics as Implicit Data Augmentation: A Depth-Decomposed View on Deep Neural Network Generalization
Authors: Tianyu Ruan, Kuo Gai, Shihua Zhang β’
Published: 2025-09-24 β’
Source: arXiv
Why do deep networks generalize well? In contrast to classical generalization theory, we approach this fundamental question by examining not only inputs and outputs, but the evolution of internal features. Our study suggests a phenomenon of temporal consistency that predictions remain stable when shallow features from earlier checkpoints combine with deeper features from later ones. This stability is not a trivial convergence artifact. It acts as a form of implicit, structured augmentation that supports generalization. We show that temporal consistency extends to unseen and corrupted data, but collapses when semantic structure is destroyed (e.g., random labels). Statistical tests further reveal that SGD injects anisotropic noise aligned with a few principal directions, reinforcing its role as a source of structured variability. Together, these findings suggest a conceptual perspective that links feature dynamics to generalization, pointing toward future work on practical surrogates for measuring temporal feature evolution.
8. VisualMimic: Visual Humanoid Loco-Manipulation via Motion Tracking and Generation
Authors: Shaofeng Yin, Yanjie Ze, Hong-Xing Yu, C. Karen Liu, Jiajun Wu β’
Published: 2025-09-24 β’
Source: arXiv
Humanoid loco-manipulation in unstructured environments demands tight integration of egocentric perception and whole-body control. However, existing approaches either depend on external motion capture systems or fail to generalize across diverse tasks. We introduce VisualMimic, a visual sim-to-real framework that unifies egocentric vision with hierarchical whole-body control for humanoid robots. VisualMimic combines a task-agnostic low-level keypoint tracker -- trained from human motion data via a teacher-student scheme -- with a task-specific high-level policy that generates keypoint commands from visual and proprioceptive input. To ensure stable training, we inject noise into the low-level policy and clip high-level actions using human motion statistics. VisualMimic enables zero-shot transfer of visuomotor policies trained in simulation to real humanoid robots, accomplishing a wide range of loco-manipulation tasks such as box lifting, pushing, football dribbling, and kicking. Beyond controlled laboratory settings, our policies also generalize robustly to outdoor environments. Videos are available at: https://visualmimic.github.io .
9. A Comprehensive Evaluation of YOLO-based Deer Detection Performance on Edge Devices
Authors: Bishal Adhikari, Jiajia Li, Eric S. Michel, Jacob Dykes, Te-Ming Paul Tseng, Mary Love Tagert, Dong Chen β’
Published: 2025-09-24 β’
Source: arXiv
The escalating economic losses in agriculture due to deer intrusion, estimated to be in the hundreds of millions of dollars annually in the U.S., highlight the inadequacy of traditional mitigation strategies since these methods are often labor-intensive, costly, and ineffective for modern farming systems. To overcome this, there is a critical need for intelligent, autonomous solutions which require accurate and efficient deer detection. But the progress in this field is impeded by a significant gap in the literature, mainly the lack of a domain-specific, practical dataset and limited study on the on-field deployability of deer detection systems. Addressing this gap, this study presents a comprehensive evaluation of state-of-the-art deep learning models for deer detection in challenging real-world scenarios. The contributions of this work are threefold. First, we introduce a curated, publicly available dataset of 3,095 annotated images with bounding-box annotations of deer, derived from the Idaho Cameratraps project. Second, we provide an extensive comparative analysis of 12 model variants across four recent YOLO architectures(v8, v9, v10, and v11). Finally, we benchmarked performance on a high-end NVIDIA RTX 5090 GPU and evaluated on two representative edge computing platforms: Raspberry Pi 5 and NVIDIA Jetson AGX Xavier. Results show that the real-time detection is not feasible in Raspberry Pi without hardware-specific model optimization, while NVIDIA Jetson provides greater than 30 FPS with GPU-accelerated inference on 's' and 'n' series models. This study also reveals that smaller, architecturally advanced models such as YOLOv11n, YOLOv8s, and YOLOv9s offer the optimal balance of high accuracy (AP@.5 > 0.85) and computational efficiency (FPS > 30). To support further research, both the source code and datasets are publicly available at https://github.com/WinnerBishal/track-the-deer.
10. On Robustness of Consensus over Pseudo-Undirected Path Graphs
Authors: Abhinav Sinha, Dwaipayan Mukherjee, Shashi Ranjan Kumar β’
Published: 2025-09-24 β’
Source: arXiv
Consensus over networked agents is typically studied using undirected or directed communication graphs. Undirected graphs enforce symmetry in information exchange, leading to convergence to the average of initial states, while directed graphs permit asymmetry but make consensus dependent on root nodes and their influence. Both paradigms impose inherent restrictions on achievable consensus values and network robustness. This paper introduces a theoretical framework for achieving consensus over a class of network topologies, termed pseudo-undirected graphs, which retains bidirectional connectivity between node pairs but allows the corresponding edge weights to differ, including the possibility of negative values under bounded conditions. The resulting Laplacian is generally non-symmetric, yet it guarantees consensus under connectivity assumptions, to expand the solution space, which enables the system to achieve a stable consensus value that can lie outside the convex hull of the initial state set. We derive admissibility bounds for negative weights for a pseudo-undirected path graph, and show an application in the simultaneous interception of a moving target.
11. 4D-QENS Analysis of Correlated Ionic Conduction in SrCl$_2$
Authors: Jared Coles, Omar Chmaissem, Matthew Krogstad, Daniel M. Pajerowski, Feng Ye, Duck Young Chung, Mercouri G. Kanatzidis, Stephan Rosenkranz, Raymond Osborn β’
Published: 2025-09-24 β’
Source: arXiv
Methods of elucidating the mechanisms of fast-ion conduction in solid-state materials are pivotal for advancements in energy technologies such as batteries, fuel cells, sensors, and supercapacitors. In this study, we examine the ionic conduction pathways in single crystal SrCl$_2$, which is a fast-ion conductor above 900~K, using four-dimensional Quasi-Elastic Neutron Scattering (4D-QENS). We explore both coherent and incoherent neutron scattering at temperatures above the transition temperature into the superionic phase to explore the correlated motion of hopping anions. Refinements of the incoherent QENS yield residence times and jump probabilities between lattice sites in good agreement with previous studies, confirming that ionic hopping along nearest-neighbor directions is the most probable conduction pathway. However, the coherent QENS reveals evidence of de Gennes narrowing, indicating the importance of ionic correlations in the conduction mechanism. This highlights the need for improvements both in the theory of ionic transport in fluorite compounds and the modeling of coherent 4D-QENS in single crystals.
12. FAST: Foreground-aware Diffusion with Accelerated Sampling Trajectory for Segmentation-oriented Anomaly Synthesis
Authors: Xichen Xu, Yanshu Wang, Jinbao Wang, Xiaoning Lei, Guoyang Xie, Guannan Jiang, Zhichao Lu β’
Published: 2025-09-24 β’
Source: arXiv
Industrial anomaly segmentation relies heavily on pixel-level annotations, yet real-world anomalies are often scarce, diverse, and costly to label. Segmentation-oriented industrial anomaly synthesis (SIAS) has emerged as a promising alternative; however, existing methods struggle to balance sampling efficiency and generation quality. Moreover, most approaches treat all spatial regions uniformly, overlooking the distinct statistical differences between anomaly and background areas. This uniform treatment hinders the synthesis of controllable, structure-specific anomalies tailored for segmentation tasks. In this paper, we propose FAST, a foreground-aware diffusion framework featuring two novel modules: the Anomaly-Informed Accelerated Sampling (AIAS) and the Foreground-Aware Reconstruction Module (FARM). AIAS is a training-free sampling algorithm specifically designed for segmentation-oriented industrial anomaly synthesis, which accelerates the reverse process through coarse-to-fine aggregation and enables the synthesis of state-of-the-art segmentation-oriented anomalies in as few as 10 steps. Meanwhile, FARM adaptively adjusts the anomaly-aware noise within the masked foreground regions at each sampling step, preserving localized anomaly signals throughout the denoising trajectory. Extensive experiments on multiple industrial benchmarks demonstrate that FAST consistently outperforms existing anomaly synthesis methods in downstream segmentation tasks. We release the code at: https://anonymous.4open.science/r/NeurIPS-938.
13. When Judgment Becomes Noise: How Design Failures in LLM Judge Benchmarks Silently Undermine Validity
Authors: Benjamin Feuer, Chiung-Yi Tseng, Astitwa Sarthak Lathe, Oussama Elachqar, John P Dickerson β’
Published: 2025-09-24 β’
Source: arXiv
LLM-judged benchmarks are increasingly used to evaluate complex model behaviors, yet their design introduces failure modes absent in conventional ground-truth based benchmarks. We argue that without tight objectives and verifiable constructions, benchmark rankings can produce high-confidence rankings that are in fact largely noise. We introduce two mechanisms to diagnose these issues. Schematic adherence quantifies how much of a judge's overall verdict is explained by the explicit evaluation schema, revealing unexplained variance when judges deviate from their own rubric. Psychometric validity aggregates internal consistency and discriminant validity signals to quantify irreducible uncertainty in any benchmarking run. Applying these tools to Arena-Hard Auto, we find severe schema incoherence and factor collapse across popular judges: for example, unexplained variance exceeding 90 percent for DeepSeek-R1-32B and factor correlations above 0.93 for most criteria. We also show that the ELO-style aggregation used by Arena-Hard Auto collapses and masks genuine ranking uncertainty. Our results highlight design failures that undermine validity and offer actionable principles for building better-scoped, reliability-aware LLM-judged benchmarks. We release our code at https://anonymous.4open.science/r/judgment-to-noise-947D/README.md
14. Studies of beauty hadron and non-prompt charm hadron production in pp collisions at $\sqrt{s}$=13 TeV within a transport model approach
Authors: Jialin He, Xinye Peng, Xiaoming Zhang, Liang Zheng β’
Published: 2025-09-24 β’
Source: arXiv
In high-energy proton proton ($pp$) collisions at the LHC, non-prompt charm hadrons, originating from beauty hadron decays, provide a valuable probe for beauty quark dynamics, particularly at low transverse momentum where direct beauty measurements are challenging. We employ A Multi-Phase Transport Model (AMPT) of string melting version coupled with PYTHIA8 initial conditions to study the beauty hadron and non-prompt charm hadron productions in $pp$ collisions at $\sqrt{s} = 13$ TeV. In this work, the beauty quark mass during the generation stage has been increased to reproduce the measured $b\bar{b}$ cross section, and a beauty flavor specific coalescence parameter $r_{BM}^b$ is introduced to match LHCb measurements of beauty baryon to meson ratios. With these refinements, AMPT achieves a reasonable agreement with experimental data on beauty hadron yields and non-prompt charm hadron production from ALICE and LHCb. We present the transverse momentum and multiplicity dependence of non-prompt to prompt charm hadron ratios, providing new insights into the interplay between beauty quark production and hadronization process. We emphasize that the multiplicity dependence of the non-prompt to prompt charm hadron productions can be useful to constrain the flavor dependences of the coalescence dynamics. This work establishes a unified framework for future studies of heavy quark transport and collective flow behavior in small collision systems.
15. HiPerformer: A High-Performance Global-Local Segmentation Model with Modular Hierarchical Fusion Strategy
Authors: Dayu Tan, Zhenpeng Xu, Yansen Su, Xin Peng, Chunhou Zheng, Weimin Zhong β’
Published: 2025-09-24 β’
Source: arXiv
Both local details and global context are crucial in medical image segmentation, and effectively integrating them is essential for achieving high accuracy. However, existing mainstream methods based on CNN-Transformer hybrid architectures typically employ simple feature fusion techniques such as serial stacking, endpoint concatenation, or pointwise addition, which struggle to address the inconsistencies between features and are prone to information conflict and loss. To address the aforementioned challenges, we innovatively propose HiPerformer. The encoder of HiPerformer employs a novel modular hierarchical architecture that dynamically fuses multi-source features in parallel, enabling layer-wise deep integration of heterogeneous information. The modular hierarchical design not only retains the independent modeling capability of each branch in the encoder, but also ensures sufficient information transfer between layers, effectively avoiding the degradation of features and information loss that come with traditional stacking methods. Furthermore, we design a Local-Global Feature Fusion (LGFF) module to achieve precise and efficient integration of local details and global semantic information, effectively alleviating the feature inconsistency problem and resulting in a more comprehensive feature representation. To further enhance multi-scale feature representation capabilities and suppress noise interference, we also propose a Progressive Pyramid Aggregation (PPA) module to replace traditional skip connections. Experiments on eleven public datasets demonstrate that the proposed method outperforms existing segmentation techniques, demonstrating higher segmentation accuracy and robustness. The code is available at https://github.com/xzphappy/HiPerformer.
16. Instruction Boundary: Quantifying Biases in LLM Reasoning under Various Coverage
Authors: Zipeng Ling, Yuehao Tang, Chen Huang, Shuliang Liu, Gaoyang Jiang, Shenghong Fu, Junqi Yang, Yao Wan, Jiawan Zhang, Kejia Huang, Xuming Hu β’
Published: 2025-09-24 β’
Source: arXiv
Large-language-model (LLM) reasoning has long been regarded as a powerful tool for problem solving across domains, providing non-experts with valuable advice. However, their limitations - especially those stemming from prompt design - remain underexplored. Because users may supply biased or incomplete prompts - often unintentionally - LLMs can be misled, undermining reliability and creating risks. We refer to this vulnerability as the Instruction Boundary. To investigate the phenomenon, we distill it into eight concrete facets and introduce BiasDetector, a framework that measures biases arising from three instruction types: complete, redundant, and insufficient. We evaluate several mainstream LLMs and find that, despite high headline accuracy, substantial biases persist in many downstream tasks as a direct consequence of prompt coverage. Our empirical study confirms that LLM reasoning reliability can still be significantly improved. We analyze the practical impact of these biases and outline mitigation strategies. Our findings underscore the need for developers to tackle biases and for users to craft options carefully.
17. Investigating Security Implications of Automatically Generated Code on the Software Supply Chain
Authors: Xiaofan Li, Xing Gao β’
Published: 2025-09-24 β’
Source: arXiv
In recent years, various software supply chain (SSC) attacks have posed significant risks to the global community. Severe consequences may arise if developers integrate insecure code snippets that are vulnerable to SSC attacks into their products. Particularly, code generation techniques, such as large language models (LLMs), have been widely utilized in the developer community. However, LLMs are known to suffer from inherent issues when generating code, including fabrication, misinformation, and reliance on outdated training data, all of which can result in serious software supply chain threats. In this paper, we investigate the security threats to the SSC that arise from these inherent issues. We examine three categories of threats, including eleven potential SSC-related threats, related to external components in source code, and continuous integration configuration files. We find some threats in LLM-generated code could enable attackers to hijack software and workflows, while some others might cause potential hidden threats that compromise the security of the software over time. To understand these security impacts and severity, we design a tool, SSCGuard, to generate 439,138 prompts based on SSC-related questions collected online, and analyze the responses of four popular LLMs from GPT and Llama. Our results show that all identified SSC-related threats persistently exist. To mitigate these risks, we propose a novel prompt-based defense mechanism, namely Chain-of-Confirmation, to reduce fabrication, and a middleware-based defense that informs users of various SSC threats.
18. Effective bases and notions of effective second countability in computable analysis
Authors: Vasco Brattka, Emmanuel Rauzy β’
Published: 2025-09-24 β’
Source: arXiv
We investigate different notions of "computable topological base" for represented spaces. We show that several non-equivalent notions of bases become equivalent when we consider computably enumerable bases. This indicates the existence of a robust notion of computably second countable represented space. These spaces are precisely those introduced by Grubba and Weihrauch under the name "computable topological spaces". The present work thus clarifies the articulation between Schr\"oder's approach to computable topology based on the Sierpinski representation and other approaches based on notions of computable bases. These other approaches turn out to be compatible with the Sierpinski representation approach, but also strictly less general. We revisit Schr\"oder's Effective Metrization Theorem, by showing that it characterizes those represented spaces that embed into computable metric spaces: those are the computably second countable strongly computably regular represented spaces. Finally, we study different forms of open choice problems. We show that having a computable open choice is equivalent to being computably separable, but that the "non-total open choice problem", i.e., open choice restricted to open sets that have non-empty complement, interacts with effective second countability in a satisfying way.
19. 4D Driving Scene Generation With Stereo Forcing
Authors: Hao Lu, Zhuang Ma, Guangfeng Jiang, Wenhang Ge, Bohan Li, Yuzhan Cai, Wenzhao Zheng, Yunpeng Zhang, Yingcong Chen β’
Published: 2025-09-24 β’
Source: arXiv
Current generative models struggle to synthesize dynamic 4D driving scenes that simultaneously support temporal extrapolation and spatial novel view synthesis (NVS) without per-scene optimization. Bridging generation and novel view synthesis remains a major challenge. We present PhiGenesis, a unified framework for 4D scene generation that extends video generation techniques with geometric and temporal consistency. Given multi-view image sequences and camera parameters, PhiGenesis produces temporally continuous 4D Gaussian splatting representations along target 3D trajectories. In its first stage, PhiGenesis leverages a pre-trained video VAE with a novel range-view adapter to enable feed-forward 4D reconstruction from multi-view images. This architecture supports single-frame or video inputs and outputs complete 4D scenes including geometry, semantics, and motion. In the second stage, PhiGenesis introduces a geometric-guided video diffusion model, using rendered historical 4D scenes as priors to generate future views conditioned on trajectories. To address geometric exposure bias in novel views, we propose Stereo Forcing, a novel conditioning strategy that integrates geometric uncertainty during denoising. This method enhances temporal coherence by dynamically adjusting generative influence based on uncertainty-aware perturbations. Our experimental results demonstrate that our method achieves state-of-the-art performance in both appearance and geometric reconstruction, temporal generation and novel view synthesis (NVS) tasks, while simultaneously delivering competitive performance in downstream evaluations. Homepage is at \href{https://jiangxb98.github.io/PhiGensis}{PhiGensis}.
20. Into the Void: Understanding Online Health Information in Low-Web Data Languages
Authors: Hellina Hailu Nigatu, Nuredin Ali Abdelkadir, Fiker Tewelde, Stevie Chancellor, Daricia Wilkinson β’
Published: 2025-09-24 β’
Source: arXiv
Data voids--areas of the internet where reliable information is scarce or absent--pose significant challenges to online health information seeking, particularly for users operating in low-web data languages. These voids are increasingly encountered not on traditional search engines alone, but on social media platforms, which have gradually morphed into informal search engines for millions of people. In this paper, we introduce the phenomenon of data horizons: a critical boundary where algorithmic structures begin to degrade the relevance and reliability of search results. Unlike the core of a data void, which is often exploited by bad actors to spread misinformation, the data horizon marks the critical space where systemic factors, such as linguistic underrepresentation, algorithmic amplification, and socio-cultural mismatch, create conditions of informational instability. Focusing on Tigrinya and Amharic as languages of study, we evaluate (1) the common characteristics of search results for health queries, (2) the quality and credibility of health information, and (3) characteristics of search results that diverge from their queries. We find that search results for health queries in low-web data languages may not always be in the language of search and may be dominated by nutritional and religious advice. We show that search results that diverge from their queries in low-resourced languages are due to algorithmic failures, (un)intentional manipulation, or active manipulation by content creators. We use our findings to illustrate how a data horizon manifests under several interacting constraints on information availability.
21. Time-adaptive HΓ©nonNets for separable Hamiltonian systems
Authors: Konrad Janik, Peter Benner β’
Published: 2025-09-24 β’
Source: arXiv
Measurement data is often sampled irregularly, i.e., not on equidistant time grids. This is also true for Hamiltonian systems. However, existing machine learning methods, which learn symplectic integrators, such as SympNets [1] and H\'enonNets [2] still require training data generated by fixed step sizes. To learn time-adaptive symplectic integrators, an extension to SympNets called TSympNets is introduced in [3]. The aim of this work is to do a similar extension for H\'enonNets. We propose a novel neural network architecture called T-H\'enonNets, which is symplectic by design and can handle adaptive time steps. We also extend the T-H\'enonNet architecture to non-autonomous Hamiltonian systems. Additionally, we provide universal approximation theorems for both new architectures for separable Hamiltonian systems and discuss why it is difficult to handle non-separable Hamiltonian systems with the proposed methods. To investigate these theoretical approximation capabilities, we perform different numerical experiments.
22. Federation of Agents: A Semantics-Aware Communication Fabric for Large-Scale Agentic AI
Authors: Lorenzo Giusti, Ole Anton Werner, Riccardo Taiello, Matilde Carvalho Costa, Emre Tosun, Andrea Protani, Marc Molina, Rodrigo Lopes de Almeida, Paolo Cacace, Diogo Reis Santos, Luigi Serio β’
Published: 2025-09-24 β’
Source: arXiv
We present Federation of Agents (FoA), a distributed orchestration framework that transforms static multi-agent coordination into dynamic, capability-driven collaboration. FoA introduces Versioned Capability Vectors (VCVs): machine-readable profiles that make agent capabilities searchable through semantic embeddings, enabling agents to advertise their capabilities, cost, and limitations. Our aarchitecturecombines three key innovations: (1) semantic routing that matches tasks to agents over sharded HNSW indices while enforcing operational constraints through cost-biased optimization, (2) dynamic task decomposition where compatible agents collaboratively break down complex tasks into DAGs of subtasks through consensus-based merging, and (3) smart clustering that groups agents working on similar subtasks into collaborative channels for k-round refinement before synthesis. Built on top of MQTT,s publish-subscribe semantics for scalable message passing, FoA achieves sub-linear complexity through hierarchical capability matching and efficient index maintenance. Evaluation on HealthBench shows 13x improvements over single-model baselines, with clustering-enhanced laboration particularly effective for complex reasoning tasks requiring multiple perspectives. The system scales horizontally while maintaining consistent performance, demonstrating that semantic orchestration with structured collaboration can unlock the collective intelligence of heterogeneous federations of AI agents.
23. CyberSOCEval: Benchmarking LLMs Capabilities for Malware Analysis and Threat Intelligence Reasoning
Authors: Lauren Deason, Adam Bali, Ciprian Bejean, Diana Bolocan, James Crnkovich, Ioana Croitoru, Krishna Durai, Chase Midler, Calin Miron, David Molnar, Brad Moon, Bruno Ostarcevic, Alberto Peltea, Matt Rosenberg, Catalin Sandu, Arthur Saputkin, Sagar Shah, Daniel Stan, Ernest Szocs, Shengye Wan, Spencer Whitman, Sven Krasser, Joshua Saxe β’
Published: 2025-09-24 β’
Source: arXiv
Today's cyber defenders are overwhelmed by a deluge of security alerts, threat intelligence signals, and shifting business context, creating an urgent need for AI systems to enhance operational security work. While Large Language Models (LLMs) have the potential to automate and scale Security Operations Center (SOC) operations, existing evaluations do not fully assess the scenarios most relevant to real-world defenders. This lack of informed evaluation impacts both AI developers and those applying LLMs to SOC automation. Without clear insight into LLM performance in real-world security scenarios, developers lack a north star for development, and users cannot reliably select the most effective models. Meanwhile, malicious actors are using AI to scale cyber attacks, highlighting the need for open source benchmarks to drive adoption and community-driven improvement among defenders and model developers. To address this, we introduce CyberSOCEval, a new suite of open source benchmarks within CyberSecEval 4. CyberSOCEval includes benchmarks tailored to evaluate LLMs in two tasks: Malware Analysis and Threat Intelligence Reasoning--core defensive domains with inadequate coverage in current benchmarks. Our evaluations show that larger, more modern LLMs tend to perform better, confirming the training scaling laws paradigm. We also find that reasoning models leveraging test time scaling do not achieve the same boost as in coding and math, suggesting these models have not been trained to reason about cybersecurity analysis, and pointing to a key opportunity for improvement. Finally, current LLMs are far from saturating our evaluations, showing that CyberSOCEval presents a significant challenge for AI developers to improve cyber defense capabilities.
24. Affective Computing and Emotional Data: Challenges and Implications in Privacy Regulations, The AI Act, and Ethics in Large Language Models
Authors: Nicola Fabiano β’
Published: 2025-09-24 β’
Source: arXiv
This paper examines the integration of emotional intelligence into artificial intelligence systems, with a focus on affective computing and the growing capabilities of Large Language Models (LLMs), such as ChatGPT and Claude, to recognize and respond to human emotions. Drawing on interdisciplinary research that combines computer science, psychology, and neuroscience, the study analyzes foundational neural architectures - CNNs for processing facial expressions and RNNs for sequential data, such as speech and text - that enable emotion recognition. It examines the transformation of human emotional experiences into structured emotional data, addressing the distinction between explicit emotional data collected with informed consent in research settings and implicit data gathered passively through everyday digital interactions. That raises critical concerns about lawful processing, AI transparency, and individual autonomy over emotional expressions in digital environments. The paper explores implications across various domains, including healthcare, education, and customer service, while addressing challenges of cultural variations in emotional expression and potential biases in emotion recognition systems across different demographic groups. From a regulatory perspective, the paper examines emotional data in the context of the GDPR and the EU AI Act frameworks, highlighting how emotional data may be considered sensitive personal data that requires robust safeguards, including purpose limitation, data minimization, and meaningful consent mechanisms.
25. Smaller is Better: Enhancing Transparency in Vehicle AI Systems via Pruning
Authors: Sanish Suwal, Shaurya Garg, Dipkamal Bhusal, Michael Clifford, Nidhi Rastogi β’
Published: 2025-09-24 β’
Source: arXiv
Connected and autonomous vehicles continue to heavily rely on AI systems, where transparency and security are critical for trust and operational safety. Post-hoc explanations provide transparency to these black-box like AI models but the quality and reliability of these explanations is often questioned due to inconsistencies and lack of faithfulness in representing model decisions. This paper systematically examines the impact of three widely used training approaches, namely natural training, adversarial training, and pruning, affect the quality of post-hoc explanations for traffic sign classifiers. Through extensive empirical evaluation, we demonstrate that pruning significantly enhances the comprehensibility and faithfulness of explanations (using saliency maps). Our findings reveal that pruning not only improves model efficiency but also enforces sparsity in learned representation, leading to more interpretable and reliable decisions. Additionally, these insights suggest that pruning is a promising strategy for developing transparent deep learning models, especially in resource-constrained vehicular AI systems.
26. Can LLMs Forecast Internet Traffic from Social Media?
Authors: Jonatan Langlet, Mariano Scazzariello, Flavio Luciani, Marta Burocchi, Dejan KostiΔ, Marco Chiesa β’
Published: 2025-09-24 β’
Source: arXiv
Societal events shape the Internet's behavior. The death of a prominent public figure, a software launch, or a major sports match can trigger sudden demand surges that overwhelm peering points and content delivery networks. Although these events fall outside regular traffic patterns, forecasting systems still rely solely on those patterns and therefore miss these critical anomalies. Thus, we argue for socio-technical systems that supplement technical measurements with an active understanding of the underlying drivers, including how events and collective behavior shape digital demands. We propose traffic forecasting using signals from public discourse, such as headlines, forums, and social media, as early demand indicators. To validate our intuition, we present a proof-of-concept system that autonomously scrapes online discussions, infers real-world events, clusters and enriches them semantically, and correlates them with traffic measurements at a major Internet Exchange Point. This prototype predicted between 56-92% of society-driven traffic spikes after scraping a moderate amount of online discussions. We believe this approach opens new research opportunities in cross-domain forecasting, scheduling, demand anticipation, and society-informed decision making.
27. Hyperspectral Adapter for Semantic Segmentation with Vision Foundation Models
Authors: JuanaJuana Valeria Hurtado, Rohit Mohan, Abhinav Valada β’
Published: 2025-09-24 β’
Source: arXiv
Hyperspectral imaging (HSI) captures spatial information along with dense spectral measurements across numerous narrow wavelength bands. This rich spectral content has the potential to facilitate robust robotic perception, particularly in environments with complex material compositions, varying illumination, or other visually challenging conditions. However, current HSI semantic segmentation methods underperform due to their reliance on architectures and learning frameworks optimized for RGB inputs. In this work, we propose a novel hyperspectral adapter that leverages pretrained vision foundation models to effectively learn from hyperspectral data. Our architecture incorporates a spectral transformer and a spectrum-aware spatial prior module to extract rich spatial-spectral features. Additionally, we introduce a modality-aware interaction block that facilitates effective integration of hyperspectral representations and frozen vision Transformer features through dedicated extraction and injection mechanisms. Extensive evaluations on three benchmark autonomous driving datasets demonstrate that our architecture achieves state-of-the-art semantic segmentation performance while directly using HSI inputs, outperforming both vision-based and hyperspectral segmentation methods. We make the code available at https://hyperspectraladapter.cs.uni-freiburg.de.
28. Projective Kolmogorov Arnold Neural Networks (P-KANs): Entropy-Driven Functional Space Discovery for Interpretable Machine Learning
Authors: Alastair Poole, Stig McArthur, Saravan Kumar β’
Published: 2025-09-24 β’
Source: arXiv
Kolmogorov-Arnold Networks (KANs) relocate learnable nonlinearities from nodes to edges, demonstrating remarkable capabilities in scientific machine learning and interpretable modeling. However, current KAN implementations suffer from fundamental inefficiencies due to redundancy in high-dimensional spline parameter spaces, where numerous distinct parameterisations yield functionally equivalent behaviors. This redundancy manifests as a "nuisance space" in the model's Jacobian, leading to susceptibility to overfitting and poor generalization. We introduce Projective Kolmogorov-Arnold Networks (P-KANs), a novel training framework that guides edge function discovery towards interpretable functional representations through entropy-minimisation techniques from signal analysis and sparse dictionary learning. Rather than constraining functions to predetermined spaces, our approach maintains spline space flexibility while introducing "gravitational" terms that encourage convergence towards optimal functional representations. Our key insight recognizes that optimal representations can be identified through entropy analysis of projection coefficients, compressing edge functions to lower-parameter projective spaces (Fourier, Chebyshev, Bessel). P-KANs demonstrate superior performance across multiple domains, achieving up to 80% parameter reduction while maintaining representational capacity, significantly improved robustness to noise compared to standard KANs, and successful application to industrial automated fiber placement prediction. Our approach enables automatic discovery of mixed functional representations where different edges converge to different optimal spaces, providing both compression benefits and enhanced interpretability for scientific machine learning applications.
29. Demystifying the Evolution of Neural Networks with BOM Analysis: Insights from a Large-Scale Study of 55,997 GitHub Repositories
Authors: Xiaoning Ren, Yuhang Ye, Xiongfei Wu, Yueming Wu, Yinxing Xue β’
Published: 2025-09-24 β’
Source: arXiv
Neural networks have become integral to many fields due to their exceptional performance. The open-source community has witnessed a rapid influx of neural network (NN) repositories with fast-paced iterations, making it crucial for practitioners to analyze their evolution to guide development and stay ahead of trends. While extensive research has explored traditional software evolution using Software Bill of Materials (SBOMs), these are ill-suited for NN software, which relies on pre-defined modules and pre-trained models (PTMs) with distinct component structures and reuse patterns. Conceptual AI Bills of Materials (AIBOMs) also lack practical implementations for large-scale evolutionary analysis. To fill this gap, we introduce the Neural Network Bill of Material (NNBOM), a comprehensive dataset construct tailored for NN software. We create a large-scale NNBOM database from 55,997 curated PyTorch GitHub repositories, cataloging their TPLs, PTMs, and modules. Leveraging this database, we conduct a comprehensive empirical study of neural network software evolution across software scale, component reuse, and inter-domain dependency, providing maintainers and developers with a holistic view of its long-term trends. Building on these findings, we develop two prototype applications, \textit{Multi repository Evolution Analyzer} and \textit{Single repository Component Assessor and Recommender}, to demonstrate the practical value of our analysis.
30. Learning Robust Penetration-Testing Policies under Partial Observability: A systematic evaluation
Authors: Raphael Simon, Pieter Libin, Wim Mees β’
Published: 2025-09-24 β’
Source: arXiv
Penetration testing, the simulation of cyberattacks to identify security vulnerabilities, presents a sequential decision-making problem well-suited for reinforcement learning (RL) automation. Like many applications of RL to real-world problems, partial observability presents a major challenge, as it invalidates the Markov property present in Markov Decision Processes (MDPs). Partially Observable MDPs require history aggregation or belief state estimation to learn successful policies. We investigate stochastic, partially observable penetration testing scenarios over host networks of varying size, aiming to better reflect real-world complexity through more challenging and representative benchmarks. This approach leads to the development of more robust and transferable policies, which are crucial for ensuring reliable performance across diverse and unpredictable real-world environments. Using vanilla Proximal Policy Optimization (PPO) as a baseline, we compare a selection of PPO variants designed to mitigate partial observability, including frame-stacking, augmenting observations with historical information, and employing recurrent or transformer-based architectures. We conduct a systematic empirical analysis of these algorithms across different host network sizes. We find that this task greatly benefits from history aggregation. Converging three times faster than other approaches. Manual inspection of the learned policies by the algorithms reveals clear distinctions and provides insights that go beyond quantitative results.
31. The Knowledge-Behaviour Disconnect in LLM-based Chatbots
Authors: Jan Broersen β’
Published: 2025-09-24 β’
Source: arXiv
Large language model-based artificial conversational agents (like ChatGPT) give answers to all kinds of questions, and often enough these answers are correct. Just on the basis of that capacity alone, we may attribute knowledge to them. But do these models use this knowledge as a basis for their own conversational behaviour? I argue this is not the case, and I will refer to this failure as a `disconnect'. I further argue this disconnect is fundamental in the sense that with more data and more training of the LLM on which a conversational chatbot is based, it will not disappear. The reason is, as I will claim, that the core technique used to train LLMs does not allow for the establishment of the connection we are after. The disconnect reflects a fundamental limitation on the capacities of LLMs, and explains the source of hallucinations. I will furthermore consider the ethical version of the disconnect (ethical conversational knowledge not being aligned with ethical conversational behaviour), since in this domain researchers have come up with several additional techniques to influence a chatbot's behaviour. I will discuss how these techniques do nothing to solve the disconnect and can make it worse.
32. RAD: Towards Trustworthy Retrieval-Augmented Multi-modal Clinical Diagnosis
Authors: Haolin Li, Tianjie Dai, Zhe Chen, Siyuan Du, Jiangchao Yao, Ya Zhang, Yanfeng Wang β’
Published: 2025-09-24 β’
Source: arXiv
Clinical diagnosis is a highly specialized discipline requiring both domain expertise and strict adherence to rigorous guidelines. While current AI-driven medical research predominantly focuses on knowledge graphs or natural text pretraining paradigms to incorporate medical knowledge, these approaches primarily rely on implicitly encoded knowledge within model parameters, neglecting task-specific knowledge required by diverse downstream tasks. To address this limitation, we propose Retrieval-Augmented Diagnosis (RAD), a novel framework that explicitly injects external knowledge into multimodal models directly on downstream tasks. Specifically, RAD operates through three key mechanisms: retrieval and refinement of disease-centered knowledge from multiple medical sources, a guideline-enhanced contrastive loss that constrains the latent distance between multi-modal features and guideline knowledge, and the dual transformer decoder that employs guidelines as queries to steer cross-modal fusion, aligning the models with clinical diagnostic workflows from guideline acquisition to feature extraction and decision-making. Moreover, recognizing the lack of quantitative evaluation of interpretability for multimodal diagnostic models, we introduce a set of criteria to assess the interpretability from both image and text perspectives. Extensive evaluations across four datasets with different anatomies demonstrate RAD's generalizability, achieving state-of-the-art performance. Furthermore, RAD enables the model to concentrate more precisely on abnormal regions and critical indicators, ensuring evidence-based, trustworthy diagnosis. Our code is available at https://github.com/tdlhl/RAD.
33. When Words Can't Capture It All: Towards Video-Based User Complaint Text Generation with Multimodal Video Complaint Dataset
Authors: Sarmistha Das, R E Zera Marveen Lyngkhoi, Kirtan Jain, Vinayak Goyal, Sriparna Saha, Manish Gupta β’
Published: 2025-09-24 β’
Source: arXiv
While there exists a lot of work on explainable complaint mining, articulating user concerns through text or video remains a significant challenge, often leaving issues unresolved. Users frequently struggle to express their complaints clearly in text but can easily upload videos depicting product defects (e.g., vague text such as `worst product' paired with a 5-second video depicting a broken headphone with the right earcup). This paper formulates a new task in the field of complaint mining to aid the common users' need to write an expressive complaint, which is Complaint Description from Videos (CoD-V) (e.g., to help the above user articulate her complaint about the defective right earcup). To this end, we introduce ComVID, a video complaint dataset containing 1,175 complaint videos and the corresponding descriptions, also annotated with the emotional state of the complainer. Additionally, we present a new complaint retention (CR) evaluation metric that discriminates the proposed (CoD-V) task against standard video summary generation and description tasks. To strengthen this initiative, we introduce a multimodal Retrieval-Augmented Generation (RAG) embedded VideoLLaMA2-7b model, designed to generate complaints while accounting for the user's emotional state. We conduct a comprehensive evaluation of several Video Language Models on several tasks (pre-trained and fine-tuned versions) with a range of established evaluation metrics, including METEOR, perplexity, and the Coleman-Liau readability score, among others. Our study lays the foundation for a new research direction to provide a platform for users to express complaints through video. Dataset and resources are available at: https://github.com/sarmistha-D/CoD-V.
34. Beyond Language Barriers: Multi-Agent Coordination for Multi-Language Code Generation
Authors: Micheline BΓ©nΓ©dicte Moumoula, Serge Lionel Nikiema, AlbΓ©rick Euraste Djire, Abdoul Kader Kabore, Jacques Klein, TegawendΓ© F. Bissyande β’
Published: 2025-09-24 β’
Source: arXiv
Producing high-quality code across multiple programming languages is increasingly important as today's software systems are built on heterogeneous stacks. Large language models (LLMs) have advanced the state of automated programming, yet their proficiency varies sharply between languages, especially those with limited training data such as Rust, Perl, OCaml, and Erlang. Many current solutions including language-specific fine-tuning, multi-agent orchestration, transfer learning, and intermediate-representation pipelines still approach each target language in isolation, missing opportunities to share knowledge or exploit recurring cross-language patterns. XL-CoGen tackles this challenge with a coordinated multi-agent architecture that integrates intermediate representation, code generation, translation, and automated repair. Its distinguishing feature is a data-driven mechanism for selecting bridging languages: empirically derived transfer matrices identify the best intermediate languages based on demonstrated translation success rather than raw generation accuracy. The system performs early output validation, iteratively corrects errors, and reuses intermediate artifacts as contextual scaffolds for subsequent translations. Extensive experiments show that XL-CoGen yields notable improvements with 13 percentage-point gains over the strongest fine-tuned baseline and as much as 30 percentage points over existing single-language multi-agent methods. Ablation studies further demonstrate that compatibility-guided bridging significantly outperforms LLM-based heuristics, confirming the value of cumulative cross-language knowledge transfer.
35. SPARQ: An Optimization Framework for the Distribution of AI-Intensive Applications under Non-Linear Delay Constraints
Authors: Pietro Spadaccino, Paolo Di Lorenzo, Sergio Barbarossa, Antonia M. Tulino, Jaime Llorca β’
Published: 2025-09-24 β’
Source: arXiv
Next-generation real-time compute-intensive applications, such as extended reality, multi-user gaming, and autonomous transportation, are increasingly composed of heterogeneous AI-intensive functions with diverse resource requirements and stringent latency constraints. While recent advances have enabled very efficient algorithms for joint service placement, routing, and resource allocation for increasingly complex applications, current models fail to capture the non-linear relationship between delay and resource usage that becomes especially relevant in AI-intensive workloads. In this paper, we extend the cloud network flow optimization framework to support queuing-delay-aware orchestration of distributed AI applications over edge-cloud infrastructures. We introduce two execution models, Guaranteed-Resource (GR) and Shared-Resource (SR), that more accurately capture how computation and communication delays emerge from system-level resource constraints. These models incorporate M/M/1 and M/G/1 queue dynamics to represent dedicated and shared resource usage, respectively. The resulting optimization problem is non-convex due to the non-linear delay terms. To overcome this, we develop SPARQ, an iterative approximation algorithm that decomposes the problem into two convex sub-problems, enabling joint optimization of service placement, routing, and resource allocation under nonlinear delay constraints. Simulation results demonstrate that the SPARQ not only offers a more faithful representation of system delays, but also substantially improves resource efficiency and the overall cost-delay tradeoff compared to existing state-of-the-art methods.