arXiv:2505.01584v4 Announce Type: replace Abstract: Adaptive video streaming optimizes Quality of Experience (QoE) metrics by selecting appropriate bitrates according to varying network bandwidth and user demands. In practice, however, real-world network bandwidth often exhibits heterogeneity relative to training environments. Current methods predominantly tackle this problem through learning-based approaches designed to improve generalization performance. While our systematic investigation reveals a critical limitation: neural networks suffer from plasticity loss, significantly impeding their ability to adapt to heterogeneous network conditions. Through theoretical analysis of neural propagation mechanisms, we demonstrate that existing dormant neuron metrics inadequately characterize neural plasticity loss. To address this limitation, we have developed the Silent Neuron theory, which provides a more comprehensive framework for understanding plasticity degradation. Based on these theoretical insights, we propose the Reset Silent Neuron (ReSiN), which preserves neural plasticity through strategic neuron resets guided by both forward and backward propagation states. Moreover, we establish a tighter performance bound for ReSiN under non-stationary network conditions. In our implementation of an adaptive video streaming system, ReSiN has shown significant improvements over existing solutions, achieving up to 168% higher bitrate and 108% better quality of experience (QoE) while maintaining comparable smoothness. Furthermore, ReSiN consistently outperforms in stationary environments, demonstrating its robust adaptability across different network conditions.
Science Journals
arXiv:2505.14215v3 Announce Type: replace Abstract: Despite the growing integration of retrieval-enabled AI agents into society, their safety and ethical behavior remain inadequately understood. In particular, the integration of LLMs and AI agents with external information sources and real-world environments raises critical questions about how they engage with and are influenced by these external data sources and interactive contexts. This study investigates how expanding retrieval access -- from no external sources to Wikipedia-based retrieval and open web search -- affects model reliability, bias propagation, and harmful content generation. Through extensive benchmarking of censored and uncensored LLMs and AI agents, our findings reveal a consistent degradation in refusal rates, bias sensitivity, and harmfulness safeguards as models gain broader access to external sources, culminating in a phenomenon we term safety degradation. Notably, retrieval-enabled agents built on aligned LLMs often behave more unsafely than uncensored models without retrieval. This effect persists even under strong retrieval accuracy and prompt-based mitigation, suggesting that the mere presence of retrieved content reshapes model behavior in structurally unsafe ways. These findings underscore the need for robust mitigation strategies to ensure fairness and reliability in retrieval-enabled and increasingly autonomous AI systems.
arXiv:2604.04809v2 Announce Type: replace Abstract: As software proliferates across domains, its aggregate energy footprint has become a major concern. To reduce software's growing environmental footprint, developers need to identify and refactor energy smells: source code implementations, design choices, or programming practices that lead to inefficient use of computing resources. Existing catalogs of such smells are either domain-specific, limited to performance anti-patterns, lack fine-grained root cause classification, or remain unvalidated against measured energy data. In this paper, we present a comprehensive, language-agnostic, taxonomy of software energy smells. Through a systematic literature review of 60 papers and exhaustive snowballing, we coded 320 inefficiency patterns into 12 primary energy smells and 65 root causes mapped to the primary smells. To empirically validate this taxonomy, we profile over 21,000 functionally equivalent Python code pairs for energy, time, and memory, and classified the top 3000 pairs by energy difference using a multi-step LLM pipeline, mapping 55 of the 65 root causes to real code. The analysis reveals that 71% of samples exhibit multiple co-occurring smells, memory-related smells yield the highest per-fix energy savings, while power draw variation across patterns confirms that energy optimization cannot be reduced to performance optimization alone. Along with the taxonomy, we release the labeled dataset, including energy profiles and reasoning traces, to the community. Together, they provide a shared vocabulary, actionable refactoring guidelines, and an empirical foundation for energy smell detection, energy-efficient code generation, and green software engineering at large.
arXiv:2607.07518v1 Announce Type: new Abstract: Deformable shape representations have proven to be robust complements to texture features in cardiac image classification, offering geometric priors that are invariant to imaging artifacts and intensity variations. However, existing deep networks perform simple concatenation to combine these distinct feature representations, which neither fully exploits their complementary nature nor learns cross-modal feature dependencies. Furthermore, this results in uniform attention across all timepoints; hence ignoring the varying diagnostic importance across the cardiac phases. In this paper, we propose a novel cardiac video classification model that, for the first time, learns temporal features in an integrated space of deformable shape and image texture representations. In particular, we design a bi-directional cross-attention in the latent space to fuse latent deformable shape and image features, allowing each modality to adaptively weight the other based on spatio-temporal correspondence. In contrast to current methods that apply uniform weighting across all the cardiac phases, our approach learns to dynamically adjust the contributions of shape and texture representations, derived from images, over time. We demonstrate state-of-the-art classification performance on a cine cardiac magnetic resonance (CMR) video dataset, achieving improved interpretability from attention mechanisms that identify diagnostically critical cardiac phases and modality contributions.
arXiv:2607.07533v1 Announce Type: new Abstract: Personalizing exoskeleton assistance requires user-specific gait data across many locomotor tasks, yet collecting this data demands repeated motion capture sessions that are costly, time-intensive, and especially burdensome for clinical populations. This challenge is most acute across walking speeds, where gait changes substantially and deviates further in clinical gait. This work introduces a subject-conditioned residual diffusion framework that generates personalized lower-limb kinematics at unseen walking speeds from a subject's gait sequence at a single seen speed. Given sagittal-plane hip, knee, and ankle trajectories at a seen speed and a desired unseen speed, the model generates a residual that transforms the seen trajectory into the unseen one, using a transformer denoiser conditioned on the subject's gait and the two speeds through feature-wise linear modulation. Trained only on able-bodied data, the model achieved a mean absolute error (MAE) of 3.4{\deg} on held-out able-bodied subjects. Without any stroke-specific fine-tuning, it achieved a 6.0{\deg} MAE on out-of-training-distribution stroke subjects, retaining subject identity for clinical gait. The framework reduced the MAE by over 70% relative to supervised feed-forward baselines, and a single seen speed matched the accuracy of four speeds within 0.4{\deg}. These results demonstrate that subject-conditioned residual diffusion can synthesize personalized gait across speeds from minimal data, reducing the collection burden for downstream exoskeleton personalization.
arXiv:2606.29358v2 Announce Type: replace Abstract: Multi-robot manipulation requires jointly reasoning about contact formations, robot motions under coupled dynamics, and collision avoidance. Systematically searching over this large space is difficult and becomes increasingly intractable as the number of robots grows, the task horizon lengthens, or the scene becomes more densely cluttered. Existing approaches therefore either learn to solve the problem end-to-end via reinforcement learning or restrict planning to a simpler surrogate problem, such as planning object motions while learning short-horizon contact primitives. However, neither paradigm scales to the problem instances we target: long-horizon multi-robot manipulation in extremely dense environments. In this paper, we propose Long-horizon Adaptive Manipulation Planning (LAMP), a framework combining a generative model for manipulation with classical planning for long-horizon reasoning. We instantiate our framework with two algorithms leveraging insights from established planning techniques, A* and lazy search: LAMP-A*, which systematically searches over the coupled object-robot space, and LAMP-Lazy, a lazy planner that enables real-time replanning through deferred evaluation. Experiments in challenging simulated environments demonstrate that our approach solves complex long-horizon tasks in highly cluttered environments that prior methods cannot handle.
arXiv:2509.20778v2 Announce Type: replace Abstract: Advancements in remote sensing have led to development of several satellite-derived precipitation products; however, their accuracy must be evaluated before use in scientific and operational studies. This study comprehensively assesses six widely used datasets PERSIANN CCS, CHIRPS, MSWEP, IMERG, AgERA5, and GSMaP ISRO against a dense rain gauge network across Karnataka, a southern Indian state characterized by diverse climatic conditions and complex topography. The analysis focuses on the Indian summer monsoon season for 2011 to 2022. To complement traditional metrics, tools from complex network theory were applied to investigate spatial organization and connectivity patterns of rainfall. A functional climate network approach was used to construct rainfall correlation networks, while event synchronization, a nonlinear measure, quantified the co occurrence of extreme events. Most products reproduced large scale monsoon features, yet their ability to represent intensity categories and extremes varied. GSMaP ISRO showed the highest correlation, lowest bias, and RMSE across subregions, whereas PERSIANN CCS exhibited systematic errors, particularly in Western Ghats, though correlations improved over interior plains. Network-based analysis reaffirmed GSMaP ISROs skill in replicating spatial correlation structures, capturing high coherence in regions dominated by large-scale processes and lower coherence in areas influenced by localized dynamics. The observed rainfall network revealed strong synchronization between the coastal region and central Karnataka, indicating broad spatial co occurrence of extremes, while the Malnad region showed weaker connectivity, suggesting localized events. GSMaP ISRO closely reproduced this degree distribution, reflecting corrections using IMD gridded dataset. Future work should improve sub-daily and localized rainfall estimates, especially in complex terrain.
arXiv:2607.06925v1 Announce Type: new Abstract: Compact world models that condition on a language goal promise to ground relations such as ``put the red block left of the blue block'' using a sparse set of explicit \emph{reference anchors}. We ask when such references actually ground a relation, and identify a trap: a goal-conditioned predictor reaches a striking $0.90$ relation-readout accuracy, yet this is \emph{instruction transcription}, not perception. Withholding the goal collapses it to chance ($0.90\!\to\!0.27$, three seeds) and a counterfactual instruction makes the predicted anchors follow the \emph{false} instruction $94.5\%$ of the time (true scene $2.3\%$; $N{=}256$). Tested across three settings and a within-task ablation, our central claim characterizes the confound: \textbf{instruction leakage occurs when the scored quantity is transcribable from the instruction (when the instruction names the answer) and is essentially independent of how predictive the non-instruction inputs are.} Our tabletop and the external BabyAI benchmark leak, whereas a Language-Table forward-dynamics world model whose instruction names \emph{referents} does not, until the instruction is augmented to name the direction; and degrading the action never increases leakage, the opposite of what predictor-competition predicts. The diagnosis prescribes the fix: keep the goal out of the dynamics (it belongs to the planner's cost) and supervise the \emph{read} path, recovering genuine, instruction-independent grounding ($0.88$, identical with and without the goal). The detection protocol and remedy apply to any goal-conditioned world model whose instruction names the scored quantity.
arXiv:2607.07537v1 Announce Type: new Abstract: In the Euclidean travelling salesman problem (Euclidean TSP), a salesman must visit $n$ points in Euclidean space, while minimizing the travel distance, according to the Euclidean distance function. In online Euclidean TSP, introduced by Abrahamsen, Bercea, Beretta, Klausen and Kozma [ESA 2024], the points are revealed one at a time, and a time slot must be assigned before the next is revealed. Once a point is assigned to a time slot, it can never be reassigned to another time slot. There are $n$ time slots. Euclidean online TSP is a high-dimensional generalization of online sorting, introduced by Aamand, Abrahamsen, Beretta and Kleist [SODA 2023]. Bertram [ESA 2025] showed an algorithm that achieves a competitive ratio of $O(\sqrt{n})$ in the worst case. In stochastic online Euclidean TSP, the points are sampled uniformly and independently in the unit $d$-cube. Kalavas, Platanos and Tolias [STACS 2026] presented an algorithm achieving a competitive ratio of $O(\log^2 n)$ with high probability for stochastic online Euclidean TSP. We present a simple algorithm that for $d \geq 2$ achieves an expected competitive ratio of $O(1)$, and for $d=1$ achieves an expected competitive ratio of $O(\log n)$, matching the lower bound by Hu [SODA 2026] for $d=1$. The algorithm is deterministic, and the expectation is due to the stochastic input. We also show that in the variant where there are more time slots than points, i.e., $\left\lceil(1 + \varepsilon)n\right\rceil$ time slots and $d=1$, our algorithm achieves an expected competitive ratio of $O\left(1 + \log \varepsilon^{-1}\right)$. We also survey algorithms from the literature. We experimentally evaluate our algorithm, which reveals that in all variants the constant factor hidden asymptotically is small. We also evaluate the algorithms from the literature.
arXiv:2508.15516v4 Announce Type: replace Abstract: Urban parks play a key role in supporting public health. Landscape architecture typically considers parks through the lens of form and function. While past research on equitable access has focused mainly on park form, studies addressing functional uses have been constrained by limited scale and coarse measurement techniques. Existing efforts have partially quantified park functions through small-scale surveys and movement data or general usage data, but have not effectively captured the specific activities and motivations underlying park visits. As a result, our understanding of the functional roles urban parks play remains incomplete. We introduce a novel method that refines mobile base station coverage using antenna azimuths, enabling more precise distinction of mobile traffic within parks versus surrounding areas. Using Paris as a case study, we analyze a large-scale dataset of passively collected per-app mobile network traffic across 45 urban parks. We test two hypotheses: the Central-City hypothesis, which posits that multifunctional parks emerge in dense, high-rent areas due to land scarcity; and the Socio-Spatial hypothesis, which views parks as reflections of neighborhood routines and preferences. Our analysis shows that parks have distinctive mobile traffic signatures, differing from both their urban surroundings and from each other and identify three distinct functional types, Lunchbreak, Cultural, and Recreational parks, and analyze the traffic usage toward different motivations for visitation. Centrally located parks display more diverse app usage and pronounced temporal variation, while suburban parks reflect the digital behaviors of nearby communities, with app preferences aligned to neighborhood income. These findings demonstrate the value of mobile traffic as a proxy for studying the diversity of usage and activities within urban green spaces.
arXiv:2607.06970v1 Announce Type: new Abstract: This paper investigates the subspace consensus problem of matrix-weighted multi-agent networks, where each agent possesses a vector-valued state in $\mathbb{R}^{d}$ and interactions between neighboring agents are characterized by matrix-valued edge weights. Besides all dimensions of the agent states achieve full-state consensus, many practical applications appeal that agents are required to agree only on certain dimensions while maintaining desired relative configurations in the remaining ones. To address this gap, we introduce the concept of subspace consensus. A matrix-weighted network is said to achieve subspace consensus on a subspace $\mathbb{V}\subseteq\mathbb{R}^{d}$ if the projection of the agents' state differences onto $\mathbb{V}$ asymptotically converges to zero. This definition renders the traditional consensus as a special case when $\mathbb{V}=\mathbb{R}^{d}$. From an algebraic perspective, we derive necessary and sufficient conditions for subspace consensus by analyzing the interplay between the null spaces of edge weights. From a topological perspective, we present sufficient conditions characterized by $\mathbb{V}$-connectivity and the existence of a $\mathbb{V}$-spanning tree, as well as necessary conditions based on graph cuts. Furthermore, we provide refined necessary and sufficient conditions specifically for tree networks. This work uncovers a fundamental capability inherent to matrix-weighted networks and establishes a systematic framework for analyzing agreement behaviors on prescribed subspaces.
arXiv:2607.06888v1 Announce Type: cross Abstract: We study the atom-atom correlation length in an atomic array coupled to a waveguide under the Bragg condition with delayed non-Markovian interactions caused by a finite photon propagation time. Starting from a single excited atom, the excitation partially spreads among all atoms, reaching a steady state. The remaining excitation localizes near the initially excited atom, and the atom-atom correlation length decreases as a power law with the interaction delay. This localization phenomenon reveals how the delay-induced non-Markovian behavior affects the correlation transport in waveguide QED systems.
arXiv:2607.04758v2 Announce Type: replace Abstract: Physical design quality-of-results~(QoR) optimization is hard and expensive. Choices made at one stage can help or hurt later stages. Each evaluation requires a costly EDA run through the full flow. While existing methods still treat optimization as flat parameter tuning or a LLM-based script generation task, we present AgenticPD, a stage-aware agentic framework for physical design QoR optimization. Instead of re-running the full flow after every trial, AgenticPD is organized around the stage boundaries of the physical design flow, where a Judge Agent navigates the search and stage-specialized agents make local decisions within their own stage using stage-local tools. Additionally, the agent harness in AgenticPD provides structured observations, execution history, and agent context management. As a result, the system can branch from prior intermediate states and reuse checkpoints to continue the optimization procedure, and every candidate is evaluated at the post-route signoff. Across these baselines, AgenticPD achieves strong post-route timing while remaining competitive in power and area.
arXiv:2607.07548v1 Announce Type: new Abstract: Large language model based search agents increasingly adopt multi-agent architectures in which a main agent decomposes a complex question into sub-queries and dispatches them to parallel sub-agents. However, existing systems instantiate all roles from a single model of identical scale, leaving open how model capacity should be distributed across roles. We factorize hierarchical search into three roles: a delegation role responsible for task decomposition, an execution role responsible for retrieval and evidence extraction, and an answer generation role held fixed as a confound control. We then conduct controlled capacity sweeps along the delegation and execution axes on five multi-hop QA benchmarks. The experiments yield three findings. First, role factorization consistently outperforms a single-agent baseline, improving exact match from 4.5 to 8.6 points across six model scales. Second, capacity sensitivity is asymmetric: scaling the delegation backbone improves EM by ~11 points, whereas scaling the execution sub-agent moves EM by only ~2.6 points, identifying decomposition as the capability bottleneck. Third, a 1.7B-parameter executor trained via quality-filtered trajectory distillation matches a frontier sub-agent in accuracy while consuming 37% fewer sub-agent tokens, advancing the Pareto frontier. These results suggest a concrete recipe for building hierarchical search agents: concentrate capacity at delegation and downsize execution without sacrificing accuracy. Our code is available at https://github.com/QinnanCai0115/role-factorized-search.
arXiv:2607.06601v1 Announce Type: new Abstract: Conditional computation can decouple language model quality from per-token inference cost, yet leading techniques act on a single axis in isolation: Mixture-of-Experts (MoE) sparsifies the FFN, Mixture-of-Depths (MoD) skips whole transformer blocks, and KV-cache quantization compresses attention memory. We argue these three decisions (attention resolution, expert selection, and cache bit-width) are strongly coupled and should be made jointly: a token rare enough to warrant full attention may also need high-precision caching regardless of which expert processes it. We introduce TriRoute, a single lightweight controller shared across all three axes that, for every token at every layer, emits a coordinated policy: (i) an attention mode (skip/local/full), (ii) a sparse set of FFN experts (with a null expert recovering MoD), and (iii) a KV-cache bit-width. The controller trains end-to-end via a heterogeneous relaxation (Gumbel-Softmax with straight-through estimation for categorical decisions and load-balanced top-k gating for experts) under a Lagrangian budget constraint that turns the average compute and memory cost into a controllable knob. We identify a cross-axis routing-collapse cascade in naive joint training, where collapse on one axis propagates to the others, and address it with per-axis normalization and a coupling-aware balancing loss. On decoder-only models from 160M to 1.3B parameters at compute-optimal token counts, TriRoute Pareto-dominates the best independent MoD+MoE+KV-quantization combination at matched inference FLOPs and memory, while better preserving tail-case robustness on rare entities, code, and arithmetic that pure perplexity optimization erodes. Post-hoc analysis reveals interpretable structure: the controller allocates full attention and high-precision cache to sentence-initial positions, rare subwords, and named entities, while cheaply routing function words.
arXiv:2607.06829v1 Announce Type: new Abstract: Rail track extraction is essential for effective railway asset management and maintenance, especially in automated inspection and mapping workflows. This paper introduces a novel method for extracting rail tracks from classified 3D point clouds using a fully convolutional recurrent neural network that preserves full spatial resolution and is trained exclusively on synthetically generated data. This approach enhances per-pixel quality and is particularly suited for rail track extraction. The proposed method begins by rasterizing points corresponding to railroad tracks, then applies the neural network to reduce noise and yield a cleaner track representation suitable for vectorization [1]. Subsequent morphological operations further refine the resultant data, enabling accurate track centerline extraction. Next, the extracted centerlines undergo smoothing to eliminate residual irregularities [2, 3]. Finally, the algorithm transfers 3D information from lidar points onto 2D polylines and applies additional vertical smoothing. A single centerline for both tracks is found using the Dynamic Time Warping (DTW) algorithm [4]. The final outcome consists of rail top centerlines and track centerlines derived for rail pairs, with minimal manual intervention. Experimental validation confirms the effectiveness of this method in yielding high-quality rail track extraction.
arXiv:2607.07046v1 Announce Type: new Abstract: Large language models (LLMs) are widely used in intelligent services due to their remarkable capability in generative tasks. Typically, LLM-based services process the inference requests of the users in a centralized data center. Unfortunately, such centralized execution has limitations for end-users, such as increased response latency with communication overhead and privacy leakage risk. To alleviate the aforementioned limitations, there have been increasing pushes to execute LLM inference locally on user-end devices. However, the limited resources of a single edge device impose restrictions on achievable accuracy of LLMs. To overcome the issue, we first propose to leverage multiple user-end devices available at the edge for LLM inference, enabling the execution of larger models. Specifically, we propose Voltron, a novel on-device LLM inference framework that elastically utilizes multiple user-end devices for LLM inference execution while adapting to diverse real-world edge environments. In our evaluation, Voltron achieves up to 16.5% higher accuracy than state-of-the-art LLMs that can be executed on a single edge device, satisfying user QoS requirements.
arXiv:2601.14590v3 Announce Type: replace Abstract: Counterfactual explanations (CFEs) provide human-centric interpretability by identifying the minimal, actionable changes required to alter a machine learning model's prediction. Therefore, CFs can be used as (i) interventions for abnormality prevention and (ii) augmented data for training robust models. We conduct a comprehensive evaluation of CF generation using large language models (LLMs), including GPT-4 (zero-shot and few-shot) and two open-source models-BioMistral-7B and LLaMA-3.1-8B, in both pretrained and fine-tuned configurations. Using the multimodal AI-READI clinical dataset, we assess CFs across three dimensions: intervention quality, feature diversity, and augmentation effectiveness. Fine-tuned LLMs, particularly LLaMA-3.1-8B, produce CFs with high plausibility (up to 99%), strong validity (up to 0.99), and realistic, behaviorally modifiable feature adjustments. When used for data augmentation under controlled label-scarcity settings, LLM-generated CFs substantially restore classifier performance, yielding an average 20% F1 recovery across three scarcity scenarios. Compared with optimization-based baselines such as DiCE, CFNOW, and NICE, LLMs offer a flexible, model-agnostic approach that generates more clinically actionable and semantically coherent counterfactuals. Overall, this work demonstrates the promise of LLM-driven counterfactuals for both interpretable intervention design and data-efficient model training in sensor-based digital health. Impact: SenseCF fine-tunes an LLM to generate valid, representative counterfactual explanations and supplement minority class in an imbalanced dataset for improving model training and boosting model robustness and predictive performance
arXiv:2607.07183v1 Announce Type: new Abstract: This working paper describes a pipeline for turning historical sources into structured data organised around the principle of foregrounding action as the basic and constitutive unit of analysis. It is rooted in a desire for pipelines that suit a granular approach to social history. The pipeline rests on the principles developed in the GRAM-framework (Graph of Roles and Actions Model), but leverages a range of machine learning tools to allow for an automated, skeletal graphing of actions. Ideally, such auto-GRAMS would integrate with close readings, including extensive manual graphing. Finally, we provide an example of how this approach might work in practice by graphing actions of pretending across four separate archival collections, relating to runaways and itinerants in eighteenth and nineteenth-century Denmark.
arXiv:2601.14271v2 Announce Type: replace Abstract: Shared accountability records are often used by parties who may never agree about causation, responsibility, or normative interpretation. For such records, neutrality cannot be achieved by omitting contested information, because accountability requires preserving the claims parties made, with their sources and provenance. Nor can neutrality be achieved by asserting one contested interpretation as the shared base. This paper defines a neutral substrate as a shared representational layer that provides stable reference while making no object-level substrate-layer commitments to causal or normative propositions. The central design constraint is that, when causal and normative propositions are contestable across admissible frameworks and the substrate's referential commitments are common ground, the substrate's neutrality is guaranteed at design time if and only if its foundational layer is restricted to those referential commitments and attribution propositions whose attributional basis is fixed by them. Causal and normative content may still be represented, but not as object-level foundational-layer commitments: it may appear there only as the content of attributed assertions with provenance, made by some identified framework, source, agent, institution, record, or document. The representational machinery used here is standard: reification, attribution, and provenance. The contribution is the constraint: a checkable condition on the foundational layer of a shared record, stated together with the assumptions it depends on and the boundary condition under which the constraint does not apply. A neutral substrate says enough to preserve accountability, but it does not turn one party's interpretation into an object-level substrate-layer commitment. The constraint does not apply at that layer when the referential regime or attributional basis is contested among the frameworks in play.
arXiv:2604.21586v3 Announce Type: replace Abstract: Autocatalytic processes underlie diverse systems in which replication is triggered at interfaces, including heterogeneous catalysis on solid substrates, enzyme activity at membranes, viral infections, biofilm growth, and spatially structured ecosystems. In a typical scenario, particles move in a bulk medium and interact with surface regions, where they may either disappear or reproduce through branching, splitting or fission. Here, we develop a general theoretical framework to understand such surface-mediated autocatalytic processes. We show that the interplay between loss and replication at surfaces gives rise to rich population dynamics. For this purpose, we derive a renewal-type nonlinear integral equation for the generating function of the population size, providing access to its full probability distribution and statistical moments. We further establish an equivalent description in terms of a Fokker-Planck equation with nonlinear Robin-type boundary conditions that encode surface reactions. Our results identify distinct dynamical regimes and universal scaling laws, and provide a unified framework to predict when surface activity promotes extinction or explosive growth. These findings offer quantitative insight into catalytic efficiency, metabolic regulation, and population persistence in spatially heterogeneous environments.
arXiv:2607.07523v1 Announce Type: new Abstract: The whole-system impedance model has proven a powerful tool for assessing the small-signal stability of multi-inverter power systems; however, its application is limited to a small range around a steady-state operating point due to the inherent assumptions of time invariance and linearisation. In this paper, a dedicated physics-informed neural network (PINN) for small-signal stability analysis in high-dimensional multi-inverter power systems is developed. The PINN is trained with step-response data produced from limited sets of system electromagnetic transient (EMT) simulations, and the trained model can predict the poles and residues of the whole-system impedance/admittance model, i.e., the transfer functions, across the full operating space. Such a PINN offers unique insights into system stability that surpass what conventional analytical methods or EMT simulations can achieve. By characterising how the impedance model evolves with power flow variations, it predicts the dynamic behaviour of the time-varying system and reveals oscillation risks that may emerge while identifying their root causes. It also provides direct visualisation of the possible range of oscillatory modes under a given power flow condition, enabling an optimal generation distribution while maintaining safe operation of the system. The proposed PINN is fully validated on a 2-IBR system and a 4-IBR system, with its application details presented.
arXiv:2607.07665v1 Announce Type: new Abstract: Classifier-free guidance (CFG) is the standard way to strengthen class-conditioning in diffusion and flow-matching samplers, yet at large guidance it oversaturates and destabilizes, symptoms practitioners suppress with more steps or limited-interval schedules. We analyze CFG through an asymptotic-preserving, numerical-analysis lens. Building on a recent result that the deterministic DDIM step is the unique fitted operator for the unguided terminal layer, exact on the final small-sigma stretch of sampling, we show that guidance re-stiffens exactly the discriminative subspace to an anomalous exponent 1+w. DDIM is therefore no longer fitted there, and on coarse meshes its guided residual diverges as sigma_min goes to zero. We prove a guided clock barrier with three ordered step-size thresholds, and read one-step oversaturation as its endpoint: a solver artifact on the calibration model rather than the continuous guided law. The same analysis yields a one-coefficient, zero-extra-NFE repair: replace CFG's w(r-1) by r^(1+w)-r on the guidance direction. On the calibration model's discriminative crossover, this removes CFG's sigma_min-divergent blow-up and is first-order accurate against the exact guided flow as sigma_min goes to zero. On learned CIFAR-10 checkpoints, and as a cross-domain smoke test on Stable Diffusion 1.5 DDIM, it acts as a high-guidance stabilizer at no extra cost rather than a universal quality knob: it cuts residual amplification and saturation, gives 9/9 point-FID wins over CFG on the tested grid, and preserves classifier-proxy target accuracy in the hard-cell blocks. We report the limits alongside: it is not a universal image-quality win, and against a dense vanilla-CFG reference it is not a uniformly better integrator of that field.
arXiv:2607.07627v1 Announce Type: new Abstract: We revisit the complexity of deciding whether a graphical game admits a pure Nash equilibrium (PNE) parameterized by standard measures of the input graph, such as treewidth. The natural dynamic programming algorithm for this problem has parameter dependence $\alpha^{(\Delta+1)\text{tw}}$ where $\alpha$ is the maximum number of strategies available to each player, each player's utility depends on at most $\Delta$ other players, and the input graph has width $\text{tw}$. Our first contribution is to point out that an algorithm by Thomas and van Leeuwen [Algorithmica 2015] claiming to improve this dependence to $\alpha^{O(\text{tw})}$ is flawed and, more strongly, such an algorithm would imply that FPT=W[1]. We then set out to pinpoint the fine-grained complexity of this problem with respect to standard parameters and show that the natural DP algorithm is not optimal, as the problem can be solved with dependence $\alpha^{\lfloor \frac{2\Delta}{3} + 1 \rfloor \text{tw}}$, $\alpha^{\lfloor \frac{\Delta}{2} + 1 \rfloor \text{pw}}$, and $\alpha^{\text{ctw}}$, where $\text{pw}, \text{ctw}$ are the pathwidth and cutwidth of the input respectively. Our main algorithmic tool is a tightening of the relationship between the width of a graph $G$, its maximum degree, and the width of $G^2$, which may be of independent interest. Complementing these results, we show that our algorithms for pathwidth and cutwidth are likely to be optimal, as improving them is equivalent to falsifying the pw-SETH.
From Custom-Fit to Portable: Bridging the Gap Between Synthesized and Engineered GPU Query Execution
arXiv:2607.07632v1 Announce Type: new Abstract: GPUs are increasingly used for analytical query processing, but developing GPU-based database engines that achieve the peak performance of the underlying hardware requires substantial research and engineering effort. A recent line of work argues that query processing should be synthesized, not engineered. In this scenario, instead of tuning a general-purpose engine to fit a workload, a large language model (LLM) generates code specialized to one query, one dataset, and one machine, thereby achieving an order-of-magnitude improvement in performance. This thesis, however, has so far been tested only on CPUs. In this work, we revisit the synthesize-versus-engineer debate for GPU analytics by answering three questions: (i) how good is synthesized GPU code?, (ii) why is it faster than engineered engines?, and (iii) how much of its advantage can be transferred back into a single, performance-portable engine? To answer the first question, we present SHADB, an LLM-based synthesis framework that generates optimized CUDA or HIP kernels using an automated, profile-guided optimization loop. Using SHADB, we show that the synthesized code approaches the memory-bandwidth ceiling and outperforms a state-of-the-art JIT-compiled GPU database engine (HeavyDB) by 7.4$\times$ on SSB SF100. To answer the second question, we decompose this performance gap and systematically classify optimizations as generalizable or workload-specific. Finally, to answer the third question, we integrate these generalizable optimizations into SYCLDB, a performance-portable engine written entirely in the open SYCL programming model. Using optimized SYCLDB, we show that it is possible to substantially bridge the gap to synthesized code (within 1.27$\times$ total execution time) while retaining workload-level generality and hardware-level performance portability.